Read more

The other side have been quickly heralded as the panacea to ‘brain rot’ and the negligence perpetuated by social media marketing. ‘Evidence based’ creators, often with little academic background, mentorship, support or industrial experience, have (rightfully so) seen an opportunity to challenge sensationalist claims with reason, rationalism, and the odd p-number here or there.

I myself have been guilty of posting such content previously. There’s certainly some good to be had in advocating for objective reasoning and the scientific method: you can counter dizzying emotions surrounding the latest ‘biohacking’ claim with assurances that you don’t need millions of pounds (or hours) to achieve decent, baseline health. If you happen to have a background in cognitive science or neuroscience, it can be really powerful to offer a counter to various brain. myths, or cherry-picking minute, nothing-burger-findings that are stretched to their limits in order to fit a creator’s ‘branding needs’. These are positive in that we can empower people to think critically rather than buy into a beautifully crafted Canva post with immediacy.

Yet in an effort to counter flagrant findings, many - to emphasise, well-meaning - individuals have fallen for these very traps, just simply coated in clinical-sounding terms, oddly-interpreted statistical outputs, and a terribly disguised cognitive bias of their own, presented as “research”.

In philosophy circles, we call these kinds of logical errors a type of scientism. Scientism, in a very general, broad nutshell (look at me committing the very errors I present to you, dear reader!), posits that science is the only way to render truth about reality.

I am a scientist, and I love scientist. And, I absolutely hate to see the ways in which this ‘evidence-based’ community has mutated through illogical syllogisms, poor statistical wrangling, and weaponising a weird superiority complex that’s cloaked as ‘concern’.

Scientism As A Logical Fallacy

To make things abundantly clear, here’s some philosophical logic, outlining the pitfalls of arguing that ‘absolutely everything can be argued using science’ (I say this as a scientist!)

(1) Any argument in favour of scientism would necessarily not be a scientific argument or one that proceeds via the scientific method. Since scientism is the idea that science and the scientific method are the only sources of knowledge, then by its own lights, any successful argument in its favour would therefore actually undermine it.

Such an argument would give us knowledge in a way that, according to scientism, isn’t possible.

(2) A simpler way to see the point is to recognise that scientism is a philosophical theory, not a scientific theory.

So if it is true, then it is a piece of knowledge that, according to scientism itself, we can’t possibly have.

(3) Simpler still, scientism is self-refuting.

Yet scientism rarely announces itself that plainly. Nobody sits down and writes “science is the only valid path to truth” on a slide with a stock photo of a brain; you’ll see a smug ‘rebuttal’ video, often mocking influencers who, whilst have nothing better to do with their time, don’t often fully think their original nonsense through anyway.

There exists a real reflex to reach for a study before a thought, in the assumption that if something can’t be quantified it probably isn’t real. If we assume this, then we quietly substitute rigour for certainty.

Any Statistics Can Be A Story

And this is where the statistics come in, because statistics are the costume scientism - or some ‘evidence based influencers - wears when it wants to appear respectable. Nobody asks what the confidence interval [that the influencer feeds us] actually contains, or whether the p-value of 0.049 is doing the load-bearing work of an entire personality theory.

A given, singular paper becomes scripture the moment it becomes content. Whilst the data is not being misused exactly it’s most definitely being venerated, treated as though the act of citing a DOI transforms an opinion into a law of nature, or hushes alternative metrics or testing procedures.

Reductionism, A Nihilist’s Bedfellow

As the name suggests, reductionism is the school of thought that a set of processes or phenomena can be ‘reduced’ down to some other (often, simplistic) theory or explanation.

If reductionism would have one hater in the world, then that hater would be me.

In an effort to counter some of the flagrant claims made by your more commercially-adjacent influencers, these public figures brush with the biggest bloody strokes you’ve seen, so much so Van Gogh would blush.

Cortisol is the reason you’re anxious. Dopamine is the reason you’re addicted to your phone. Your gut bacteria are the reason you’re sad, anxious, tired, uninspired, and apparently unable to hold down a relationship.

Somewhere along the way, “this hormone is involved in this process” got flattened into “this hormone is the cause of your entire inner life”, and the flattening is the whole trick. It’s not wrong in the way a flat-earther is wrong, but it is wrong in the way a caricature is wrong, and that, dear reader, is arguably more dangerous than the satirically off-key claims of ‘SPF bad’ we all know, love, and satirise.

Why is this? Well, by cloying not at the surface and gesturing at something true, you end up missing everything that actually mattered about the original.

What makes this particularly slippery is that it borrows the aesthetics of scepticism while abandoning the substance of it. Real scientific literacy means holding a finding loosely, knowing its limits, understanding that a mechanism existing in a petri dish or a fMRI scanner doesn’t hand you a life philosophy or concrete theory to revere or rattle off. What we’re getting instead is certainty dressed up as nuance. The confident air of “well, actually, this study shows”, when done in such a way we’re seeing now, actually shuts a conversation down rather than opening it up. The scientific method is, yes, about throwing so much crap at your theory in an attempt to break it (we want to know it’s solid, for certain!) but reductionism mistakes the vocabulary of science for the humility - and intellectual curiosity - that’s supposed to come with it.

It’s been funny coming into the world of cognitive and data science. I love the hardness that statistics and quantitative methods brings forth, for sure - but I’m also a writer and bookworm, keen to find any opportunity to delve into ancient forests and castles dangling in the sky baubles. When I work with statistical methods, in the initial process I must strip any attributed “meaning” I might be tempted to assign to a given dataset in order to find signal. In these instances, a forest becomes a land-cover classification. An owl becomes a data point in a species-distribution model. The methodology necessitates disenchantment to function.

There’s stats to the left of me and fae folk to the right, and here I am, stuck in the middle between two worlds which provide equal importance in how us humans attempt to understand the world around ourselves.

A Seed Of Intrigue

The Green Man appears on medieval church carvings across Britain and nobody is quite sure why. A pagan remnant, a symbol of nature’s persistence, perhaps something that crept into sacred spaces through the instincts of stonemasons (who maybe felt that the forest deserved representation in the cathedral, whether or not the theology had room for it).

C.S. Lewis called this quality the Numinous: the sense of encountering something that exceeds your pre-allocated categories, an experience that makes you feel simultaneously very small and very awake. It’s not comfortable. It doesn’t resolve into our predefined vectors neatly. By being there it presses on you.

Scientists would do well to try and sit with this feeling of Luminosity in their work and not just try brute force p-hacking or diluted observational studies. As someone who is stalwart on our quantitative methods, I’m now here to advocate for a wisdom that is deep, profound, and present in our folkloric and fiction works peppered across literature and history’s sense-making.

Wizard Versus Data Scientist

There’s a moment in almost every fantasy narrative where an apprentice asks the wizard to simply explain it. There, our protagonist veers over a precipice; the ragged veil dividing normalcy and call to the unknown. Yet adventure is not a real-world dataset, even if it mirrors its unstandardised chaos. We cannot quantify such experiences. So, the wizard (Gandalf, Merlin, Elminster if you’re of the Dungeons and Dragons variety) invariably refuses, or more precisely, simply can’t explain. There is a deep, profound wisdom that comes with the experience in of itself. It lives in the hands, the gut, the accumulated scar tissue of ten thousand encounters with things that didn’t behave as expected. You don’t learn it. You become it, slowly, through proximity and failure and time.

The philosopher Michael Polanyi called this tacit knowledge. In 1958, he wrote of the uncomfortable fact that “we know more than we can tell.” He was thinking of surgeons, master glassblowers, cyclists who couldn’t explain how they balanced; of whom have a certain je-ne-sais-quoi quality in how they evaluate the thing that stands before them.. In the same vein, if you spend enough time around genuinely brilliant scientists, you’ll recognise the same quality. That razor instinct that a dataset or research methodology smells wrong before they’ve even run a single diagnostic. Perhaps they assert the quiet suspicion that an effect is too clean, too tidy; that something upstream has gone sideways in data collection, or even when we’ve formed our hypotheses (because yes, there very much is such a thing as too good to be true in the veins of research methodology).

In science and research, one is forced to make a judgment call about which variables actually matter. This precedes and shapes every model you build: applying the most complex statistical processes does bugger all if the data demands a broader approach, and this innate ‘nouse’ we try and carve out in our craft cannot be generated by a simple prompt, or found within the first few pages of a textbook. It is earned through various edge cases, practical problems, and discourse with peers and those who have paved the path before us. And let’s be completely honest, these experience points can be rarely justified in a methods section of a paper. Yet, they are fundamental in how we develop a knack for statistics.

It’s kinda hilarious that the methods section is precisely where we pretend none of this happened.

Historiographically, “Western” science inherited from the Enlightenment (which twisted many branches of knowledge-seeking and experimental paradigms from the Abbasid Caliphate, no matter who argues otherwise) its own peculiar fantasy: that knowledge, properly conducted, is fully explicit, reproducible, and transmissible to any sufficiently trained reader.

The methods section in any given paper is the purest expression of this faith. Reading it gives you a linear ‘map’ of the design at hand: here is the procedure; follow it and you will get what we got. It is a noble aspiration. It is also, in practice, a polite fiction. What it leaves out is the third author’s instinct that something was off with the wave two data. It omits a senior researcher’s decision, made in about four seconds, to treat those outliers differently. The years of pattern recognition that make one analyst’s exploratory data analysis look like clairvoyance compared to a graduate student’s.

Suffice to say this isn’t a critique of rigour. Rigour matters enormously, and anyone who’s read my previous piece on hypothesis testing will know I’ve got little patience for its abuse, as influencers attempt to dress up poorly applied statistics as being “evidence-based”. It’s a critique of the pretence of completeness when presenting our findings.

I always chuckle when people present the act of writing up research as though it were executed by a well-documented algorithm, rather than a human being with a history.

And with all that said: a wizard doesn’t pretend to be following a manual. Perhaps we shouldn’t either.

Tom Bombadil Is Not a Bug To Be Fixed

There is a figure in Tolkien’s The Lord of the Rings who has troubled readers and adaptors for decades. Tom Bombadil, ancient, joyful, inexplicable, lives in the Old Forest and cares nothing for the Ring. When Bombadil puts it on, nothing happens at all. He sits entirely outside every system of power and meaning that the rest of the narrative operates within, a character that refuses to be indexed. The Ring, which reduces all things to its own logic, simply has no entry point into Bombadil’s presence within the pages.

Peter Jackson left him out of the films entirely, which is understandable but frankly also telling. Bombadil doesn’t serve a narrative function in any conventional sense. He’s not there to advance plot or illuminate character. He just is, irreducibly, in a way that resists the tidy machinery of a popcorn-flick story structure.

Tolkien, when pressed, defended him as deliberate:

“even in a mythical Age there must be some enigmas, as there always are. Tom Bombadil is one (intentionally)” 1

In essence, Bombadil acts as a reminder that there are things in the world that exist outside all worldly frameworks and predate all systems.

In research, data pipelines have a word for this: noise. We call them anomalies, outliers, edge cases. The implicit argument here is that the world, properly understood, should be classifiable, and anything that resists said classification is a problem with the data rather than a problem with the classification.

Yet anomalies can also be where everything interesting happens. The correlation that appears between two variables you weren’t studying and has no business being there, has the potential to become (years later, with much political fatigue) the seed of an entirely new field. I’ve personally found some of my most interesting findings from proving a null-hypothesis, or uncovering anomalous findings, than I have receiving calculations or evaluation metrics that I had sort’ve expected a priori.

What does this say about a system that has no room for a figure like Bombadil? Whether that absence is a failure of the world, or a failure of the framework you brought to it, it’s worth poking at in the same way we challenge weak statistical power or data modelling methods.

Reality As A Fantasy

Fantasy worlds, for all their strangeness and magical realism, don’t lie about this. They don’t pretend their wizards are just following procedure, nor do they dismiss Tom Bombadil as noise or study anomaly and move on. They hold open space for the things that resist classification, not because they’re being unrigorous, but because they understand that a map which excludes the unmappable isn’t a complete map, but rather a confident one.

The enchantment and joy us scientists should be cultivating is in the quality of attention we chose to bring to our research. There is a version of scientific practice I envision we all try to embody, regardless of our job title or academic background. Incidentally, this vision encapsulates traits which are innate in our favourite fantasy novel characters: careful, curious, genuinely humble about what their world’s frameworks can and cannot hold.

• Carpenter, Humphrey, ed. (2023) [1981]. The Letters of J. R. R. Tolkien: Revised and Expanded Edition. New York: HarperCollins. ISBN 978-0-35-865298-4.

To answer this kind of question, researchers typically reach for what’s known as a hypothesis test. At the centre of almost every hypothesis test sits a small, deceptively simple number: the p-value. Understanding what a p-value actually is, and crucially what it is not, turns out to be one of the most important skills in all of science.

So what is a p-value, really?

The formal definition is this: a p-value is the probability of observing your data, or something more extreme, assuming the null hypothesis is true. The null hypothesis is typically the assumption that nothing interesting is happening that there is no difference between your two conditions, no effect of your intervention, no signal in your noise. So, in the example I gave above, in this case the null hypothesis would be something like "there is no difference in exam performance between students who complete practice tests and students who spend all their time revising."

Crucially, notice what that definition does not say. It does not say the p-value tells you the probability that your hypothesis is true. It does not tell you whether your finding matters in the real world. And it certainly does not hand you a binary verdict of “real effect” or “no effect.” It is a statement about the compatibility of your data with a particular model of the world. Nothing more.

This sounds like a subtle distinction. It isn’t. Getting it wrong (and most people do, including many trained researchers - I know I certainly have) leads to three specific errors that may corrupt scientific literature every single day.

Oopsie 1: Treating p > 0.05 as “no effect”

Perhaps the most common mistake is reading a non-significant result — say, p = 0.34 — and concluding “there was no difference between the groups.” This sounds reasonable. It is wrong.

A non-significant p-value means only that your observed data is not sufficiently surprising under the null hypothesis to justify rejecting it. It does not mean the null hypothesis is true.

There could be a genuine effect that your study simply lacked the statistical power to detect. For example, sample size can interfere with study robustness. A study with too few participants will routinely fail to find real effects, not because those effects do not exist, but because the study was never capable of seeing them in the first place.

The correct statement after p > 0.05 is closer to: “we found no statistically significant difference”. This is genuinely different from “we found no difference.” The first is a statement about your data. The second is a claim about reality. And you have not earned the second one from a single p-value aberration.

Oopsie 2: Treating p < 0.05 as proof

The mirror-image error is equally common. A researcher finds p = 0.03, and writes “we can conclude there is an effect.” The implicit assumption is that the null hypothesis is now false and the alternative is confirmed.

But a p-value below 0.05 simply means that, if the null hypothesis were true, you would see data this extreme less than 5% of the time. That’s it. It does not rule out the possibility that you got unlucky. That unlucky result has a name: a Type 1 error, or false positive. In essence, set your threshold at 5%, run enough studies, and roughly one in twenty results will be a false positive even when absolutely nothing is going on.

This is precisely the problem that came to a head in psychology’s so-called replication crisis. In 2015, the Open Science Collaboration published the results of an ambitious project in which researchers attempted to replicate 100 published psychology studies. Despite the fact that 97% of the original studies had reported statistically significant results, only 36% successfully replicated. One well-known casualty was John Bargh’s influential “elderly walking” priming study from 1996, which had reported that participants walked more slowly after being exposed to words associated with old age. Astonishingly, this was a result that was widely taught in universities and cited hundreds of times.

Here comes the womp-womp. When researchers later attempted direct replications, the effect repeatedly failed to appear. The original p-value had cleared the 0.05 bar. It had not, it turned out, confirmed a truth about human behaviour.

This is not a flaw in the logic of hypothesis testing. In fact, I’d argue that it is a feature of how the system is designed, and it works well when understood correctly. The problem is the leap from “surprising data” to “confirmed hypothesis.” Those are not the same thing, and treating them as equivalent is what fills journals with findings that cannot be reproduced.

Oopsie 3: Lindley’s Paradox; when significance misleads even the careful

Here is where things get genuinely strange, and where even researchers who understand the first two dangers can go wrong.

Imagine you have conducted a very large, well-powered study. The kind that methodologists actively encourage, and the ‘golden star’ type of study they hammer into you at university and within academia. Because of your large sample size, you have extremely high statistical power. You run your test and get p = 0.045. Significant, right?

Not so fast. When statistical power is very high (say, 99%) your study is so sensitive that if a real effect exists, you should be seeing p-values far below 0.01, not hovering just under 0.05.

A p-value of 0.045, in a high-powered study, is actually more consistent with the null hypothesis being true than with a meaningful effect existing. You have a result that clears the conventional significance bar, but is, paradoxically, mild evidence against the alternative hypothesis you are trying to support.

This is Lindley’s Paradox, and it illustrates something important: the 0.05 threshold is not a universal truth. It is a convention, and like all conventions, it breaks down in certain conditions. High-powered studies call for stricter thresholds. Treating 0.05 as sacred regardless of context is the kind of rigid, half-understood rule-following that the paradox exposes.

The real danger of half-knowledge

Each of these errors shares a common root. Someone has learned just enough about p-values to use them, but not necessarily with the depth to watch out for edge cases. It’s fundamental knowledge to run the test, read the output, and report the number, but it’s still not enough to understand what the number genuinely represents. It is far too easy know the label without understanding the contents.

The danger here is not ignorance. Ignorance, after all, at least tends to come with caution. The danger is the false confidence that comes from having a method. A researcher who has never heard of p-values will hesitate before drawing strong conclusions. A researcher who runs a t-test and sees 0.03 will often feel, viscerally, that they have found something; and that feeling is hard to shake, even when the logic demands more humility.

This replication crisis the sciences are facing did not happen because scientists were careless. It happened, in large part, because they were confidently applying a tool they only partially understood.

Statistics is a framework for thinking more clearly in the presence of uncertainty. It works best when the person using it holds that uncertainty front and centre, as opposed to when they treat a single number as the answer.

A beat, until you hear a bleating of sheep, and footsteps of a farmhand on their way to greet the inevitable dawn. The spell is broken, and the world - with all of its squalid chaos and noise and political fatigue - is still there.

No matter how many moments of solitude we manage to squirrel away, it’s very difficult to retain the illusion of being the only human alive for very long. Our virtual worlds, however, paint a different story. When we converse with another over the internet, there’s a level of anonymity that brings forth that same question we ask ourselves on that country lane: am I the only person here? Is there anyone out there, in this big blob of technological mist? And damn, why am I feeling so incensed after visiting the comments section of this cat video? After all, how do we know that Reddit user ItsT1meToDuel420 isn't some Mossad-backed bot, implemented to reap uncertainty and rage-baiting throughout the subreddits they frequent?

It’s not just dusk-touched pathways that have us questioning the existential. Very much in keeping with the fear of the unknown, a tongue-in-cheek conspiracy, (appropriately dubbed the Dark Forest of The Internet), presupposes that the internet is divided by a “good”, human subgroup of users, who are forced to retreat into private, hidden spaces (a la, as per its name, a dark forest), staying silent in an effort to avoid predators (bots, rage-bait, rampant advertisements, and the like).

Dark Forest: Sanctity or Sanitised?

An article which writes on this subject matter wonderfully is a piece by Yancey Strickler. Whilst I will likely butcher the summary piece, the author makes an apt analogy to Liu Cixin’s sci-fi work, The Three Body Problem. At the risk of spoiling, the novel talks about alien civilisations remaining in hiding, not wishing to reveal themselves for fear of predators, much like users and netizens online will actively seek these “dark forests” of eerie calm to hide away from hostile actors.

As Strickler writes:

In response to the ads, the tracking, the trolling, the hype, and other predatory behaviors, we’re retreating to our dark forests of the internet, and away from the mainstream.

Whilst the internet has always been a hub of seedy pop-up ads and angry strangers debating how women should dress or men should act, the AI bubble has bought on a particularly pernicious, detached approach to conversation and ideological dissection. For one, unlike the blatantly obnoxious and bad faith discussions found across Reddit, X, or Instagram, “AI slop” can be an eerily-curated, digital machine regulated by an algorithm to appear, in all aspects, human. Whilst for now there is an ‘uncanny valley’ sense of these Sora-generated videos, or copywriting which reads much like a ‘how do you do, fellow kids’ Buscemi meme, what is most frightening isn’t the idea that AI will reach human-like communication, but rather that people will embrace this.

As someone who works in tech and deals with AI Models in her work, I cannot stress to you enough how the corporatist commodification of AI-as-human-replacement disgusts me. And perhaps a little knowledge is a dangerous thing, because one can feel awfully helpless watching ChatGPT control humans’ lives as easily as an adult bribes a toddler with chocolate. This digital space is deceptively alluring: bot-to-human, or bot-to-bot exchanges temper all that internet trash-talking and vulgarity into something, well, totally stripped. Reminiscent of Huxley’s Brave New World, we are easily lulled by this “prim” semblance of sentience by these AI agents: “ChatGPT it”, “ask ChatGPT”, or being consistently validated for terrible behaviour by an algorithm that has been weighted to, essentially, agree wholeheartedly with you. As a species, human beings define their reality through suffering and overcoming that suffering. AI systems and computers are designed with optimisation in mind: freedom is chaos to an algorithm, and as such, as is humanity’s choice to withstand pain and embrace growth.

As aptly put by Slavoj Zizek:

We are simply not aware of the extent to which our daily lives are already manipulated and regulated by digital algorithms that, in some sense, know us better than we know ourselves and impose on us our “free” choices. In other words, to mention yet again the well-known scene from cartoons (a cat walks in the air above a precipice and only falls when it looks down and realizes there is no ground beneath its feet), we are like a cat refusing to look down.

We cannot, therefore, in good conscience, assume that hiding away from “icky yucky” things equate to our dark glades of privacy and safety. Given our political and economic structures at present, users wish for comfort, gentleness, and in not-so-subtle terms, being told what to do. Instead of rejecting this wooden horse, we welcome AI’s curtailing of obscenities, squalor, and imperfections. In an attempt to swot away the vulgarities of bots, trolls, and downright bizarre interactions with strangers and ads, we’ve turned to the very system which will forever change the landscape of human labour and interactions. We’ve opted for comfortable, algorithmic control over chaotic freedom. Chaotic, yes, but freedom nonetheless; freedom to be disgustingly and utterly human.

Beyond The Woods

So here lies our choice: do we remain in the dark forest, letting algorithms curate our reality into something palatable and safe, or do we step back into the clearing? Undeniably messy, hostile, yet by all accounts, alive?

Humanity’s call to adventure is not to find some mythical, unspoiled corner of the internet. Rather, it's to reclaim the public square with all its ugliness intact. To argue with strangers who frustrate us, to write imperfectly, to be wrong in public, to suffer the indignity of pop-up ads and bot replies because the alternative, sterile efficiency, costs us our humanity.

The farmhand's footsteps breaking that country lane silence aren’t comfortable - indeed, if you’ve watched too much true crime, your imagination might be getting the better of you in that moment - but they are real. They prove the world is still breathing.

We must resist the seductive whisper that tells us to let machines handle the abrasive work of being human. Better to walk through the chaos, to trip over our words, to rage and reconcile and create something flawed and vital, than to retreat into the gentle suffocation of algorithmic perfection.

“But I don't want comfort. I want God, I want poetry, I want real danger, I want freedom, I want goodness. I want sin... I'm claiming the right to be unhappy” - Aldous Huxley, A Brave New World

What an earth could have happened to transform my body into such a useless lump?

My eldest sister had just died. And I was so afraid to live on when she could not.

I am still tethered to that point in time whilst the world trails on. I have a gaping, Emma-shaped hole in my chest, which should be filled with extra memories and stories and adventures with my big sister. Still, bracken and wild things have grown around this void. I tended to a new garden - one that originally had her in mind. Now, its foliage is coloured with a steady knowing of what life can throw.

Since then, I have reconciled with the morbid peace losing a loved one may bring. Everyday tediums or friendship tiffs seem trivial in the face of loss’ abyssal maw. I mentally soothe myself at wince-inducing moments with the adage “well, at least nobody died”. As a neurodivergent individual who can feel quite ‘lost-at-sea’ in this world, loss acts as a surprisingly effective, default comparative when trying to grapple with the obstacle course that is ‘Adulting’.

Still, there have always been moments of dread. Moments in which I picture that darkening maw visiting me once more, invading my lungs, my stomach, incapacitating me to such a degree even breathing feels laborious. What happens if I lose another?

I escaped unscathed from this hurt for a few years. Until Lauren.

Lauren was my best friend. Lauren was young, just like Emma. Lauren loved riding horses and swearing and reminded me to look at the sunnier side of things during periods when all I could see was the dark.

Lauren was also diagnosed with a brain tumour. This brain tumour proved particularly troublesome in the end, as it transmogrified into something no young woman ever wants to consider when she’s taking out the bins, or getting ready for dinner with her partner: terminal brain cancer.

So you could imagine my annoyance when last year, on October 13th, 2024, mere hours after visiting her in hospital, we’d received the news that she was no longer.

How inconsiderate. Damnit, Lauren. I suppose she did always have a flair for the theatrical.

This entire year has been numb, hurt, some more numb, and some more hurt again. Old friendships have felt fraught and difficult to navigate. New ones on the other hand have been oddly balming: fellow “Grief Clubbers” have stepped forward to share their own experiences, making me feel a bit less overwhelmed as I had to face not only the finality of Lauren’s passing, but of course revisit that smarting reminder: my big sister is also not here (oh goodie, oh yay,).

Of course, grief does not have a schedule. It doesn’t wait for the weekend or give you space for important deadlines. One moment it’s dormant, the next…it’s there. It’s quiet and deafening all at once. Yet despite this, we are expected to show up: for my family, other friendships, training, work responsibilities, even myself. I cannot count how many times during this year I’ve logged off a meeting and just let my tears burst from the seams, or days when the brain fog has been so bad I’ve barely fumbled my way to ‘survive’ my expected tasks.

Navigating the world now after another statistically far-fetched loss feels…a lot. I just don’t know how to ‘do’ normalcy.

I say this not to garner sympathy, but honestly, out of irritation. An unquenchable itch, a reminder, that grief cannot be parked or put to the side despite obligations. It’s like that awkward guest who always outstays their welcome. I cannot just ‘stop feeling’ or ‘move on’, as so many have oh-so-helpfully suggested.

But I try. I try to wake up and live for those who have not had the privilege of experiencing today. I try to find glimmers amidst the apathy. I visit my family and we reminisce on Emma and share not only our grief, but our joy and closeness and the relief that we have each other as pillars. My husband restores me with an understanding and love that I’d be hard-pressed to find in fairy tales. I have my jiu-jitsu training partners (and whom I’d like to think I can also call friends) fighting for their lives to choke me out or break my limbs in one breath, and then helping me put my own world to rights over pizza in the next.

It’s a funny, excruciating thing. Grief leaves you in the dullest of lows, but in turn, life’s little highs feel almost giddying.

There is no ‘ideal’ time for grief. No wisdom, no self-help book, no amount of martial arts training that can satiate the hungry maw of grief. Still, it is a part of life. And life is as precious and beautiful as it is heavy and long.

I’m still learning where to place death in the boundaries between these little paradoxes. I think we all are.

I’ve always devoured fantasy books like a nourishing, wintertime broth. As a child, I wore my reading list like a badge of pride. I kept my Tolkien-themed notebook - which now could be labeled as a crude attempt at a commonplace book! - scrappy and loved, full of (frankly bonkers) ideas about the unwritten novel I had lurking in the recesses of my mind. Books were a happy escape from a childhood of loneliness and being the ‘weird one’.

Being labelled the weird kid taught me that being a bookworm, being well-read, being curious about the world, ran the risk of making some people uncomfortable. Being inquisitive wasn’t seen as endearing; you were a know-it-all if you preferred the company of tomes over people. Still, I didn’t care. Reality was grey and cold. I’d much rather take a dive into paper portals, meeting unicorns, dodging deals with impish creatures, or standing in awe of swirling skylines underneath the silhouettes of flitting wyverns. My mind felt electric. I felt alive. There was a susurration in my blood, like something sparkling and sharp and clear that made Sophie, Sophie.

Over the pandemic, I sought this same feeling. Stuck glued to my sofa whilst en route to an adventure was a satisfying paradox. Physically, we were all boxed into quarantines and doomscrolling. Mentally, a book could transport you to dreamy skies, terrifying mountain scapes, or corridors of darkness where only the bravest dare venture. Rather than books being considered a ‘chore’, for me they armed my dreary reality with the tools to regulate my emotions and instil resilience.

Sometimes, we learn far more from fictional tales than we do text books.

Books & Cognitive Health

From a cognition perspective, we’re fully aware of the health benefits on offer through reading books. Whilst we take picking up a book, opening its pages, and getting cracking with a story (or arguments presented, if non-fiction) for granted, there’s a lot more going on under the hood than we initially believe.

Unlike speaking, a primarily ‘biological’ process which is linked with a variety of evolutionary mechanisms, reading is considered largely external (i.e. cultural). As such, people need a few ‘extras’ on top of the neural networks primed for visual processing and language comprehension we pick up in speaking. This includes things like attentional control, managing motivation to learn, and a variety of visual and linguistic tools that we repurpose from your baseline things like ‘seeing’, ‘reading’, or ‘speaking’.

From a neuroscience perspective, reading is quite a complex process, which explains why some neurodivergent individuals may either take to it completely, or feel put off, depending on which aspects of cognitive effort impacts them most.

When we want to do our best to be fully immersed in a good book, several cognitive processes are at play - and not linearly, either. Our working memory tries to integrate characters, language, and the unfolding story into an overlapping, layered ‘cognitive tower’ of sorts, in that we both a) are aware of what’s happening and what has happened, and b), we can critically reflect on our own thoughts and experiences on what has occurred so far. Working memory tends to intersect with processes like attention (ever felt you need to reread a page several times if you need it to ‘sink in’ after a bad night’s sleep?), as well as metacognition (thinking about thinking. Sounds silly, I know.). All of these processes involve both bottom-up and top-down mechanisms. In simple terms, this means we combine the more immediate feedback (decoding letters and grammar) with context-specific information (like character descriptions and background knowledge), respectively, to try and understand what’s at play.

Naturally, because these are cognition-derived processes, things that impact our brain can have an effect on how we feel about reading, and our habits towards book enjoyment: external stress, internal fatigue or executive dysfunction, in addition to how we were taught to read when we were younger, can all impede on our love for reading, and consequently, how we absorb both stories and non-fiction in general media.

There are, however, a couple of key things to incorporate if you want to fall in love with reading. These methods are supported by the existing psychological research, so if you fancy dusting off your once-loved tomes and taking up a new hobby throughout the colder months, consider this a signpost.

Don’t Be A Reading Snob

If you haven’t touched a book in ten years, and think your friends will poke fun at you for picking up the Percy Jackson series (pretty bomb YA stories actually, I won’t lie), you might be tempted to pick up a more ‘appropriate’ read for a fancy pants adult like yourself. Say, I dunno, War and Peace.

The problem is: War and Peace is absolutely dire if you haven’t gotten accustomed to the process and attitude to reading. In fact, even as someone who has enjoyed the story immensely, it’s no easy read, and you run the risk of alienating yourself from books if you think that every story should be that way.

It’s just not the case. Firstly, whilst no books are created equal in terms of writing style or substance, starting to build up a habit requires a baseline of enjoyment. You will never stick to a reading habit if you choose things you hate because they’re ‘good’ stories. I’d much rather you read books critics snub their noses at (which is silly - why yuck someone’s yum?) and build up gradually - if you so choose- so you can recapture the joy and magic a fun story brings.

Literary impact is important. It keeps cultural discussion, virtue ethics, and the pursuit of being a better person all alive. There is, however, always space for more “fun” reads which enable us with both the joy and tools to cope better with a stark, harsh reality.

One must prepare to be useful for tomorrow by engaging with the useless today. These forms of stories and entertainments actually equip us with the emotional toolkit to be happier, better versions of ourselves. We don’t always need a Penguin Classic - though challenging ourselves, slowly over time, is no bad thing.

“Shueyyah, Shueyyah - Or Slowly, Slowly”

Rather than diving in first to an hour-long session and wondering why you feel a burning, practically painful urge to scroll on your phone, build your tolerance for attention by putting your phone somewhere far away, and setting a timer for 10-15 minutes.

Personally, I try make it a bit more of a romanticised period of quiet during my busy life by milking my reading period. This is also known as habit stacking - pairing an initially difficult activity with something a bit more enjoyable. Whack on some lyric-free fantasy music. Grab a hot chocolate. Make a cosy corner and spray lavender on my blanket. Whatever I need to help my sensory system feel grounded and present, which in turn, helps those cognitive processes feel a lot less exhausting - especially after work, or an intense training/research period.

Much like training a muscle, try and extend this timer period, increasing it week by week as you get stuck into the process of reading. You’ll likely forget the passage of time as you start to catch the reading bug - you never know, you may join our ranks and not want to put a book down during those crepuscular moments of 3 A.M..

There’s about fifteen-odd people, cooing over the Turner piece in the National Gallery. They’re not trying to pace, or sit too idle in the room. They’re hovering by the pièce d’oeuvre, weaving between each other, on cue for their debut performance. To their left, a security guard who can’t help but join the hum of acknowledgement. He centres his gaze towards the smorgasbord of hues and pinks strummed across the canvas.

In the room, the ceiling is domed and high. The convent gather for their accord, their sanctuary. Each visitor hangs in the air, eyes fixed towards the piece, and for a moment it feels like they’re trying to mentally grapple with what’s set before them. A moment longer, and they’re sucked in, no longer standing in a London art museum, but dancing with the ebb of coastal waves and fuschia-dipped skyline.

Consider this relationship between a piece of art and human perception. Perception is a cognitive process in which we navigate and understand the world around us, dependent on sensorimotor variables in addition to an initial, visual intake of our environment. When we encounter a novel artwork - one with which we are unfamiliar - our brain may ‘update’ our expectations of what we’re seeing before us, and attempt to make sense of the artwork. This cognitive effort lends a certain immersion to it, in which we are engaging with the aestheticism - “is this beautiful” - before us. Inversely, if we find a piece particularly uninspiring - or we’re not in any type of mood to evaluate creative output after a long day of work, or the morning after - we’re probably less willing to cognitively engage with this dilemma. We disengage with the aesthetics, and see nothing beyond a few embellished paint strokes (and possibly some historical tidbits we can pull out for our next pub quiz). In this sense, engaging in the creative - heck, even the process of perception - is a conscious, purposeful stage, which requires attentional control and consideration of what is before us.

Imagine attending an art show with a dear friend, someone who feels strongly about a specific artist or piece, and is able to indulge you with the particulars one may need to thoroughly evaluate a piece. You receive data on the style, painting technique - even the psychological state of that artist during this time. Suddenly, what was once a boreish favour lights your inner world. The way a canvas turns into a universe proper enthralls you, and for those few moments of looking at the painting, you reflect on your own emotional state and attitudes towards the painting. In short, when aesthetics encourage us to be reflective, we become engaged with both the inner and outer worlds in which we find ourselves.

The Art Entanglement: Aesthetics & Morality

When we stand before that Turner seascape, temporarily forgetting whether we'd hang it in our living room or how much it might fetch at auction, we're practicing something profound: the ability to appreciate something purely for what it is, rather than what it can do for us. Kant believed that this aesthetic experience operates as a kind of gym for the soul; a space where we exercise our capacity for disinterested judgment, training ourselves to perceive beauty without the contamination of personal desire or utility. This aesthetics experience ensures we do not appreciate an object simply as a means to an end. I may not like the taste of my post-workout protein shake, but I can be charitable in that it’s beneficial for my satiety and nutrition choice, thus I am pretty consequence-driven in how I select for it. Art, on the other hand, is not something so frivolous or throwaway. It creates opportunity for us to reflect and engage; for our Selves to be challenged both on a perceptual level (in the cognition sense) and a philosophical one.

This appreciation, Kant argued, mirrors the very foundation of moral reasoning. Just as we must judge the beautiful without reference to our personal interests, moral judgment requires us to act from duty rather than inclination. In other words, to consider what ought to be done regardless of what we might gain. The person who can lose themselves (genuinely, sincerely) in aesthetic contemplation is rehearsing the mental muscles needed for ethical action.

Still, there's something more delicate at work here. In aesthetic experience, Kant suggests we encounter what he called the "supersensible"(“that which transcends both the mechanical world of nature and our own subjective desires”). When we're genuinely moved by beauty, we touch something universal, something that connects us to our shared humanity. The security guard drawn into the Turner's luminous atmosphere experiences the same aesthetic pleasure as the art historian beside him, despite their vastly different backgrounds and knowledge. Beauty, in this moment, becomes a bridge between individual consciousness and universal human experience. Art is not something that divides by class, but rather, unites, through a baseline, transcendental need to engage with aestheticism beyond the world we know. Philosopher Alva Noë pushes this connection even further, arguing that art doesn't merely train us for moral thinking, but rather, it is itself a form of moral investigation. Art, he suggests, is philosophy by other means, a way of thinking through the fundamental questions of human existence without the constraints of propositional logic. When we encounter a challenging work, say, a Francis Bacon painting that forces us to confront violence and vulnerability, we're not just having an aesthetic experience. The brushstrokes and colour palettes are not typically attributed to “dainty” works of art, yet they move us to reflect and engage with the canvas, all the same. In this instance, we may grapple with questions about what it means to be human, what we owe each other, how we should live.

Consider how a piece of art can make strange what we normally take for granted. Noë argues that this is art's essential function: to reorganize our experience, to make us see familiar things anew. When Picasso fractures a face across geometric planes, he's not just creating visual interest, he's challenging our assumptions about identity, perspective; the very nature of seeing another person. This re-organisation of experience, is precisely what moral growth requires. Through the sense-making of the somewhat nonsensical, we become better people not by learning new rules, but by developing new ways of seeing. New capacities for attention and care.

This is why aesthetic education matters morally. Each genuine encounter with beauty expands our capacity for what Kant termed "enlarged thinking”, which is the ability to think from the standpoint of others. When we allow ourselves to be genuinely affected by a piece of music that initially seemed foreign, or a poem that challenges our assumptions, we practice the fundamental skill of moral life: imagining perspectives beyond our own immediate experience. We become less provincial in our judgments, more attuned to the complexity of human experience, more capable of the kind of sympathetic understanding that ethical action requires.

And, hopefully, a little bit aware of manners (and self-restraint) when we do eventually visit our local art museum.

It’s easy to spatter these catchphrases across saccharine-pink and duckblue, pre-made, Canva posts. It’s even easier to digest as a reader; as a consumer of brands, identities, and an image of psychological work, curated by a probability-driven algorithm. And it’s enticing to buy into the notion of a life that appears to be the panacea to our entrenched, spiritual numbness in a landscape that demands wants (feeling good all the time, avoiding suffering, and pretending we’re perfect) versus needs (a deep, rigorous look at our internal worlds; righting our own wrongs before projecting them onto others).

Take for example the demand for vulnerability. There was once a time that such a moment where we can breathe, where we can be, was sought after. An absolute prerequisite to taking some time, to come home to oneself. Now, privacy is viewed as an inconvenience; a barrier to glittering riches and entrepreneurial success. Unless you share every ounce of a memory, or squeeze every last drop of you into an Instagram story, nobody will be interested in your service or your wares, and you’ll be a redundant speck of entropy on this slow-moving rock we call Earth.

Personally, I see the temptation: social media algorithms “reward” narcissistic behaviours with cheaply-mined dopamine. The cascading “heart” of approval on a story; a notification that a random user now deems you worthy of “following”. Up until recently, if someone boasted that they had ~4k followers in a few days, one would - rightfully so - be worried about who is following your travel route after work. Now, we sate for recognition; to be seen as a “public figure” for nothing more than captivating the hearts, and what’s left of the minds, of their salivating audience. It is far easier to beg for pity than it is to earn respect: embellish a somewhat upsetting story from your childhood, perhaps reconfigure a mutually ugly relationship as “toxic” or “traumatic”, and watch the likes roll in on their own merit. This is not to say there isn’t a space for community and counsel in the cases of severe trauma, and incomprehensibly awful events which require such sensitive language to articulate. But we must face the harsh reality that a small subset of very privileged, internet-addicted, mostly-middle-class-women are harnessing the reactive nature of imitation and mimicry. That they can, as any slimy politician does, make their followers feel “seen” and “connected” in an effort to sell a half-hearted “coaching programme”, with little to no accreditation backing their corner.

The more we share, the more we gain, as famed public thinker Brene Brown argues in her work. But what if some individuals took this in a far more literalist vein than once intended? What if people share their “mental health struggles” without the intention of “spreading awareness”, and seeking connection in the digital void, or - even worse - commodifying their moderate anxiety, or self-inflicted introversion, in an effort to gain kudos or address books. To what extent is this addressing the real issues surrounding mental illness - circumstantial, materialist, and spiritual numbness - or simply flogging a financially fortuitous dead horse?

There is a common myth that “self-respect” requires arrogant proclamation, or “calling out” outlandish behaviour, or setting “boundaries” (in newspeak: cutting out people of your life without a thought) whenever things leave you feeling the slightest bit rattled. It instead concerns a humming peace; a reconciliation towards the self and what is sacred and private. It involves understanding our flaws and vulnerabilities in a way where we direct our dialogue to ourselves, and not to a public screen - where our stories are perfectly pruned, curated, and sanitised for the placation of the masses. And what’s more, when we fail to strive for something greater than ourselves - I don’t care how you interpret that statement; be it God, a charity, or a cause bigger than your social media following - we try satiate the gaping hole through emotional junk food.

Rovelli's perspective in quantum physics challenges our fundamental assumptions about reality itself. His relational interpretation suggests that physical properties don't exist in isolation but emerge through interactions between systems. In other words, there are no absolute properties, only relations. This radical view dismantles the notion of an objective, observer-independent reality—a perspective with profound implications for how we conceptualise cognition.

Physics may set itself worlds apart from the domain of neuroscience and psychology, but research demands us to require more interdisciplinary freedom than ever. Computational approaches are vital in deepening our understanding of the universe around us, as complex algorithms from machine learning expose properties in data which the human eye (or even brain) neglects to pick out. Logic persists that non-statistical models may also afford us deeper insight from applying mechanics across a variety of sciences. So, if we apply quantum mechanics - or Rovelli's thinking - to cognitive science, we might ask: What if the mind isn't a computational device processing information about an external world, nor an ecological system detecting pre-existing affordances? What if, instead, cognition itself is fundamentally relational—neither located entirely "in the head" nor wholly determined by environmental structures?

This brings us to Alva Noë's enactive approach to perception. Noë challenges the prevailing assumption that perception involves creating internal representations of an external world. Instead, he argues that perception is a skilled activity—something we do, not something that happens to us. "Perception is not something that happens to us or in us," Noë writes. "It is something we do." Our perceptual experience arises through our active engagement with environments, mediated by our sensorimotor knowledge.

I have a confession: I've long been intrigued by these embodied approaches, yet remained skeptical about abandoning computational frameworks entirely. The mathematical elegance of Bayesian models seems to capture something essential about how we navigate uncertainty. However, Noë's perspective offers a compelling alternative that doesn't necessarily reject computation but situates it within a broader relational context.

Consider visual perception. The traditional computational view treats vision as a process of constructing detailed internal models from retinal input. The ecological approach emphasizes direct perception of environmental affordances. But Noë suggests something different: we perceive by actively exploring our environment through movement. We don't need detailed internal representations because the world itself serves as an "outside memory" that we access through skilled bodily activity.

Worryingly, both computational and ecological approaches inadvertently perpetuate a problematic subject-object dichotomy—either overemphasising internal processing or external determination. This dichotomy collapses under Rovelli's relational perspective, where properties emerge through interaction rather than existing independently. Similarly, Noë's enactive approach dissolves the rigid boundary between perceiver and perceived, emphasizing their co-constitution through dynamic engagement.

The standard metaphors—”brain as computer, perception as detection”—fail because they presuppose a world of discrete, pre-existing objects with fixed properties. But if Rovelli is right, reality itself is relational all the way down. There are no observer-independent facts, only interactions between systems. And if Noë is right, our perceptual experience isn't about representing or detecting a pre-existing world, but about skillfully engaging with a world that reveals itself through our active participation.

This perspective radically reframes how we should approach cognitive science. The question isn't whether brains compute probabilities or detect affordances; it's how organisms and environments co-constitute each other through ongoing, dynamic interaction. Cognition emerges not within neural tissue alone, nor is it determined solely by environmental structures, but exists in the relational space between.

Perhaps more problematically, both computational and ecological approaches struggle to account for the phenomenological dimension of experience—the lived, first-person perspective that remains frustratingly elusive to third-person methodologies. Noë's embodied approach helps bridge this gap by acknowledging that perception isn't something happening "in here" about something "out there," but is itself a mode of active engagement with the world.

Active inference, with its emphasis on organisms as hypothesis-testing agents, moves in the right direction. But even this framework often remains too computationally focused, missing the fundamental insight from Rovelli and Noë: reality itself, including our cognitive processes, is constituted through relations and interactions rather than existing as pre-given entities that interact secondarily.

The circularity in comparing brains to AI systems becomes even more apparent through this lens. We've designed AI systems based on computational metaphors that themselves emerged from particular ways of conceptualizing cognition. These metaphors were never faithful representations of brain function but conceptual frameworks with their own historical and cultural contingencies. Then, in a remarkable feat of circular reasoning, we point to these human-designed systems as evidence for computational theories of mind.

When we apply Noë's insight that perception is a skilled activity to the realm of artificial intelligence, we immediately see what's missing. Language models process vast amounts of text data but lack the embodied, exploratory engagement that characterizes human understanding. They don't perceive or act in any meaningful sense because they exist outside the relational matrix that constitutes conscious experience.

Rovelli's quantum relationalism offers a profound challenge to the metaphysical assumptions underlying both computational and ecological approaches. If properties emerge through interactions rather than existing independently, then the search for the "true nature" of cognition may be fundamentally misguided. There is no absolute fact of the matter about whether brains compute or detect—these are perspectives that become meaningful only in relation to specific experimental contexts and theoretical frameworks.

A more fruitful approach might embrace this relationality, recognizing that cognition isn't reducible to computational processes, ecological detection, or any other singular framework. Instead, it emerges through the dynamic interplay between embodied agents and their environments—a dance of mutual constitution that defies our traditional categories and metaphors.

In this view, the brain isn't a computer running cognitive algorithms, nor is it merely a detector of environmental affordances. It's a participant in relational systems that span traditional boundaries between self and world. Consciousness isn't something that happens within the skull but exists in the spaces between—in the ongoing, dynamic relations that constitute experience itself.

This relational perspective doesn't diminish the importance of neuroscience or computational modeling. Rather, it situates these approaches within a broader framework that acknowledges the irreducibly interactive nature of cognition. The brain matters tremendously, but not as an isolated computational device—as an organ of relation, enabling particular modes of engagement with environments and others.

The ultimate irony may be that our very attempt to understand cognition through technological metaphors—from clockwork to computers to neural networks—reveals something profound about consciousness itself: it is inherently relational, constituted through its engagements rather than existing as a thing-in-itself. Perhaps the most accurate model of consciousness isn't a model at all, but a recognition of its fundamentally participatory, enactive nature—a nature that reveals itself not through detached observation but through embodied involvement in the world we seek to understand.

Worryingly, there seems to be an air of mystique and '“cool” to this circularity of logic. Brains are like LLMs, therefore LLMs are like brains. The idea that the most “in vogue”, cutting edge technologies are capable of reproducing human-like results, and therefore must be like human brains, is a problematic insight that cuts to the root issue of systemic neuroscience, psychology, and statistical research.

I have a confession: I definitely have a soft spot - both in academic writing and personal opinion - for a Bayesian approach. And whilst I am most certainly less fervent in its application now - I’m much more sceptical to how *much* or at least, *what* specifically we can attribute to cognitive function proper - I myself have been a proponent of the “Bayesian Brain” trend. Recently, I had the pleasure of catching up with contemporary challenges to this notion - Kathryn Nave's 'A Drive to Survive' is a crucial piece of criticism, and there's lots of opposing arguments in enactive cognition.

While contemporary models frame brains as statistical prediction machines, this framing itself ironically misses the point. The "brain as computer" metaphor—with its neat inputs and outputs—is seductively simple but fundamentally flawed. It's another technological lens through which we're trying to decode our own cognition, much like how clockwork once informed our understanding of biological mechanisms, or the steam engine helped us understand mitochondrial function.

So, surely the answer lies in the alternative - an environmental, ecological approach? This, my friends, I am perhaps more sceptical of.

J.J. Gibson's ecological psychology emerged as a refreshing counterpoint to computational approaches, emphasizing "affordances"—those possibilities for action that environments present to organisms. Gibson rejected the notion of mental representations, instead arguing for direct perception of environmental opportunities. Worryingly, though, his approach suffers from several critical limitations.

First, and ironically enough, Gibson's framework tends toward environmental determinism - not quite equating our neuronal structures to Turing, but the logic remains just as circular. By suggesting that perception is primarily about detecting invariant features of the environment, Gibson inadvertently diminishes the active, creative role organisms play in constructing their experiential worlds. The ecological approach struggles to account for how the same environment can afford radically different possibilities to different organisms or even to the same organism in different states.

Perhaps more problematically, ecological psychology lacks statistical rigour in explaining how organisms handle uncertainty. In a world of noisy sensory data and ambiguous environmental cues, organisms must somehow navigate probabilities—a fact that Bayesian models handle elegantly, but Gibson's approach handwaves away. The mathematical elegance of predictive processing models, with their hierarchical precision-weighted prediction errors, simply has no counterpart in ecological psychology.

Active inference offers a more nuanced framework that addresses these shortcomings. Rather than being mere computational devices or passive detectors of affordances, organisms actively generate and test hypotheses about their environment through embodied action. This isn't merely prediction in the statistical sense—it's an ongoing, messy dialogue between organism and environment, where perception and action blur together in ways that defy computational analogies.

The critical insight is that brains don't simply compute probabilities about static inputs; they continuously generate possibilities and test them through embodied exploration. A computer executes algorithms on data; an organism participates in dynamic systems that include bodies, objects, and shared environments. When I reach for a cup, I'm not calculating joint angles and force vectors—I'm engaging in a fluid, context-sensitive activity that unfolds through time.

This perspective exposes the absurdity of comparing brains to language models or neural networks. LLMs may generate impressively coherent text through statistical pattern recognition, but they lack the fundamental property that defines cognition: the capacity to actively reshape an environment through purposeful action. They process information but don't genuinely perceive or act in any meaningful sense.

The circularity in comparing brains to AI systems becomes painfully apparent when we recognise that AI architectures were originally inspired by simplified neural models—models that had already stripped away the embodied, exploratory nature of biological cognition. We designed systems based on our limited understanding of brains, then mistakenly concluded that brains must work like those systems. In essence, it's common for neuroscientists and data scientists alike to view the brain as a system of artificial neural networks because that's precisely how we designed those networks in the first place.

A more fruitful approach recognises that cognition emerges from the dynamic interplay between brain, body, and world—not from computational processes occurring solely within neural tissue. This perspective doesn't diminish the brain's complexity but situates it within the broader context of embodied existence, where meaning arises through active engagement rather than passive computation or direct perception of affordances.

Perception, The Brain, And AI

First, consciousness and perception are often bridged with the processes and even output of various machine learning models. We as sentient, thinking beings wouldn’t be able to perceive, learn, or memorise without the skilful interaction of our sensorimotor capabilities. Better yet, our brain would be remiss of any cognitive function without us conducting action in our environments. For approaches such as Alva Noe’s, perception is an emerging phenomenon, borne from organism-environment interactions. In the context of AI, whilst models or similar agents are far from owning a physical body or mind, we can argue that they have a sort of ‘virtual body’ that exists by virtue of learning about their environment through reinforcement. This ‘virtual body’ acts as a proxy (without the need for a ‘real life’ brain) that permits AIs to learn skills not so dissimilar to methods us living organisms engage in.

Before theories like Noe’s Enactism, simple, one-way neural networks consisting of perception - then action would be dubbed the process of perceiving things around us.

Perception -> Action

This is because, traditionally, people have the habit of separating different process to specific regions in the brain, wrapping up perception in a neat ‘package’. In other words, a process like seeing requires one area of the brain, whilst higher-cognition tasks like thinking, planning, and executing actions belong to another. This argument is often referred to as the domain-specific concept of the brain. Yet most contemporary thinkers agree that designating actions to specific brain regions is redundant. Unlike an AI system, the human brain is not an inorganic or artificial construct; indeed, its areas are densely interconnected, and the lines get really blurry when it comes to understanding decision and precision. In mathematical theories, we create complex Neural Networks, which try to explore the relationships between decision and precision. These networks are trained to be sensitive to features of the visual environment which correspond to a relevant task, in which the brain is engaged.

The Brain Is A Predictive Organ

While these models generally try and account for embodied consciousness, in my opinion, there are alternative (and complementary) theories which better explore broader cognitive mechanisms. Take for example predictive processing. This model, stemming originally from contemporary neuroscientist Karl Friston, seeks to address these processes separate from active perception. Predictive processing proposes that the brain constantly generates predictions about sensory input, and updates internal models (e.g. the representations we form of the world, data collated from our daily tasks and interactions with the environment around us) according to the incoming information in order to minimise prediction errors.

Mathematical Frameworks in Perception and Learning

The intersection of cognitive science and machine learning reveals profound mathematical structures underlying both biological and artificial intelligence. At the foundation lies Bayesian inference, formalized as P(causes|sensory data) ∝ P(sensory data|causes) × P(causes), which describes how systems update beliefs based on new evidence. In predictive processing, this manifests through hierarchical probabilistic models where prediction errors drive learning according to Error_weighted = Precision × (Actual - Predicted). This precision-weighting mechanism allows both brains and artificial systems to adaptively adjust learning based on uncertainty and context.

Neural Architectures and Information Flow

Modern neural networks mirror cortical organization through hierarchical processing layers. In both biological and artificial systems, information flows through successive transformations, each extracting increasingly abstract features. The mathematics of attention mechanisms, expressed through self-attention matrices A = softmax(QK^T)V, provides a formal framework for understanding how both systems selectively process information. These mechanisms help explain phenomena from selective attention to conscious awareness.

Active Inference and Environmental Interaction

The mathematics of active inference unifies perception and action through the principle of free energy minimization: F = E_q[log q(s) - log p(s,o)], where systems act to minimize the difference between predicted and actual sensory states. This framework, sharing mathematical foundations with reinforcement learning, explains how both organisms and artificial agents learn optimal policies through environmental interaction. The temporal dynamics of these interactions can be captured through recurrent processing, formalized in RNNs through ht = σ(Wxxt + Whht-1 + b). Where emergent approaches often struggle to explain the mechanistic details of how consciousness and perception arise, PP provides a clear computational account that bridges multiple levels of analysis - from individual neurons to global brain dynamics. The framework's ability to explain phenomena from perceptual illusions to psychiatric disorders, while simultaneously informing the development of more capable AI systems, demonstrates its exceptional explanatory power and practical utility.

Consciousness, Information Integration, Predictive Processing, and Learning Dynamics

Information integration theory provides mathematical tools for quantifying consciousness through Φ, measuring the irreducibility of information in a system. This connects to modern transformer architectures in ML, where multi-head attention mechanisms create similar patterns of integrated information processing. These mathematical bridges suggest consciousness might emerge from specific patterns of information flow and integration that can be formally characterized and potentially replicated in artificial systems.

The mathematics of learning in both domains involves gradient descent on prediction errors, expressed as Δw ∝ -∂E/∂w, where E represents prediction error and w represents system parameters. This suggests deep similarities in how biological and artificial systems adapt to their environments. The hierarchical organization of these learning processes can be formalized through nested prediction errors at multiple scales, providing a mathematical framework for understanding how complex cognitive abilities emerge from simpler predictive mechanisms.

This mathematical grounding helps explain how both biological and artificial systems bootstrap complex cognitive abilities from simpler mechanisms. The ongoing dialogue between these fields continues to deepen our understanding of intelligence while raising new questions about the nature of consciousness and cognition. Through these formal frameworks, we move beyond philosophical speculation to testable models of how intelligence emerges in both natural and artificial systems.

Bridging Minds and Machines: Why Predictive Processing Prevails

While emergent approaches like Noë's enactivism offer valuable insights into perception and consciousness through organism-environment interactions, predictive processing provides a more comprehensive and mathematically grounded framework for understanding cognition. The strength of PP lies in its ability to unify perception, action, and learning under a single mathematical framework that aligns remarkably well with both neural computation and modern machine learning architectures.

The mathematical formalization of PP through hierarchical Bayesian inference and precision-weighted prediction errors provides testable hypotheses about brain function that emergent approaches often lack. This mathematical rigor has allowed PP to make precise predictions about neural activity patterns, perceptual phenomena, and behavioral outcomes. The success of PP-inspired machine learning models in replicating human-like learning and behavior provides strong empirical support for its core principles.

Moreover, PP's integration with machine learning has created a virtuous cycle of understanding. As we develop more sophisticated AI systems based on predictive processing principles, we gain new insights into how biological brains might implement similar computations. The striking parallels between deep learning architectures and cortical organization suggest that PP captures fundamental principles of intelligent information processing, whether in silicon or neural tissue.

This synergy between PP and machine learning continues to advance our understanding of human cognition while pushing the boundaries of artificial intelligence. By providing a formal mathematical framework that applies equally well to biological and artificial systems, PP offers a unique bridge between cognitive science and AI, promising deeper insights into the nature of intelligence itself.

Parkinson’s Disease (PD) is an example of a neurodegenerative condition in which individuals’ executive function and motor performance are compromised. Parkinson’s is a widespread, neurological condition, in which one’s motor capabilities and executive functions are impacted (Caso & Cooper, 2020). Yet there is little work discerning whether these abnormalities arise as a result of motor, sensorimotor, or perceptual integration deficits. This project proposes a fusion of computational neurophenomenological techniques with multimodal neuroimaging methods to comprehensively investigate the perceptual processing for individuals with Parkinson’s Disease.

At present, there are some computational models which have calculated specific input weights to mirror or emulate “damaged” cortical pathways (Caso & Cooper, 2020). But a novel approach to studying perception, hybrid predictive coding (HPC) neural network models, have not been sufficiently explored in relation to perceptual processes in Parkinson’s Disease. This model endeavours to capture the nuanced interaction between top-down and bottom-up signalling theories regarding perception (Rauss & Pourtois, 2013; Tscshantz et al., 2021), creating new research opportunities to explore context-free perception in neurodegenerative conditions. HPC proposes that the brain uses a hierarchical structure, with higher levels generating predictions, and lower levels processing sensory input and computing prediction errors. In short, predictive coding provides a holistic, biologically plausible account for top-down and bottom-up signalling processes, suggesting the existence of a cortical architecture that is constantly updating predictions based on sensory inputs. It is plausible to assume this model, in conjunction with experimental paradigms, can better understand how PD impacts integration of sensory input data.

There are key theories this project will seek to investigate:

1. This project proposes that perceptual refinement of individuals with PD may be distorted, thus leading to improper processing of perceptual stimuli, which in turn contributes to motor degradation and an intensification of overall symptomatology. To investigate this, experimental paradigms (e.g. fMRI, visual testing protocols) will be conducted in conjunction with a hybrid predictive coding model.

2. This project will aim to extend beyond the HPC architecture to train hidden states which can then be used to train the generative model. In Parkinson's disease, the discriminative model's ability to generate accurate and robust hidden states is impaired due to the neurological changes associated with the disease, leading to deficits in perceptual learning and prediction. This project proposes that this would also result in a reduced ability for people with PD to update predictions based on new sensory input data, specifically in context-free scenarios, such as gist perception.

Research pinpoints to these predictive mechanisms being pejoratively affected by PD (Diederech et al., 2014), but with so few computational models specifically investigating this, it is hard to understand the full impact of neurodegenerative conditions on perception. This project proposes a method to address this gap. Subsequently, this PhD will present a biological plausible architecture for predictive coding, in conjunction with experimental techniques, to relate context-free perception processes in PD brains.

Current Perception Models & Approach To Parkinson’s Disease

There have been a number of theories advocating for a primarily probabilistic or bottom-up approach in evaluating the process of perception (Cleeremans, 2005; Friston, 2005; Marr, 1982; Van Essen & Mausen, Seth, 2015). There are biological accounts which show specific neuronal connectivity pinpointing to bottom-up sensory processing with respect to spatial mapping (Basso, Bickford & Cang, 2021). At the same time, a broader understanding of predictive principles in the brain has since emerged. As such, some influential theories consider perception to be a process of top-down, Bayesian inference (Friston, 2005; Lee & Mumford, 2003), in which prior predictions are incorporated with incoming data to form perceptual models. In contrast to bottom-up models, the Bayesian perspective asserts that perception develops mostly through a top-down predictive framework (Friston, 2005; Seth, 2015). According to this perspective, perception is an active process, in which individual agents construct the best possible representation of a scene, based upon sensory data as well as prior knowledge, as opposed to a bottom-up signal feed.

Figure 1: scheme of learning by prediction error. The red represents a prediction error occurring when the reward differs from its prediction (in this case, expecting to bite into a sweet, crisp apple, finding out it is sour and soft). The green shows no error existing when the outcome matches the prediction; the behaviour remains unchanged.

The work of Hinton, Friston and colleagues built on this further by demonstrating ways of learning “priors” which can shape top-down connection weights in an effort to reduce “prediction errors”, or otherwise stated, the difference between the expected and the observed. In models such as these, top-down flows of information leave prediction errors to flow upwards in a hierarchy of interconnected layers, also known as “predictive coding”(Rao & Ballard, 1999).  Neurobiological theories that propose a deterioration of subcortical visual pathways may be partially responsible for their movement deficiencies (Diederech et al., 2014) can be further supplemented by neurophenomenological accounts to explore perception processes of individuals with Parkinson’s Disease (PD). Experimental findings have confirmed these theoretical approaches, suggesting that late-stage PD patients are more likely to have visual, perceptual impairments independent of movement dysfunction when distinguishing erratic displacements in vision test trails (Bernadine’s et al., 2021). Thus if prediction and perceptual processes are considered to be a sum of inference problems resolved by the brain (Friston & Kiebel, 2009), it is vital to investigate how Parkinsonian brains model the dynamic systems around them.

There are still strong accounts linking bottom-up signalling to specific perception processes, such as gist perception (Kveraga, Ghuman & Bar, 2007), inferring defective top-down visual processing in Parkinson’s individuals (Diederich et al., 2014; Nemcova et al., 2017; Uchiyama st al., 2012; Zhu et al., 2016; Joyce et al., 2018). Gist perception involves a number of processes to quickly form an understanding of a scene or event whilst scanning for visual irregularities (Evans & Tresiman, 2005). It is therefore an essential process from which we may orientate ourselves when in complex, unpredictable scenarios, or by mapping the space around us. It is reasonable to assume that individuals with PD will face a greater difficulty experiencing spatial, object-mapping context compared to healthy groups. Little work - computational or experimental - exists in studying perception in Parkinson’s Disease, acting as a significant motivator for is proposed work.

Hybrid Predictive Coding, Context-Free Perception, & Parkinson’s Disease

Figure 2: (Left) In the HPC model, sensory input X processes perceptual context at an amortised (rapid) rate; perceptual models at Z predict Z at an iterative rate. (Right): Approximate plot of how well a given system’s predictions match the sensory output, computing the HPC equation. All three Prediction Lines show prediction errors over time. Prediction errors are then used to update the brain’s predictions for the future. Compellingly, the bottom-up amortised feature allows for speedier inference in data environments (Tscshantz et al., 2021), whilst the top-down aspect of the model permits for more effective, iterative inference within smaller data sets. The equation has been adapted to mimic possible perceptual prediction errors in neurodegenerative conditions.

HPC provides a deeper insight in understanding the different functions of visual perception, specifically, computing the differences between fast iterations of context-free perception (e.g. gist perception), and idiographic, detail-specific processes, such as item recognition. It acts as a powerful framework to understand visual perception and its contributing role to neurodegenerative conditions like PD. The HPC model is strongly appropriate for understanding perception and rapid gist perception, particularly with respect to Parkinson’s Disease. Firstly, it retains the generative capacities of the predictive coding algorithm whilst incorporating bottom-up predictions, making it a compelling architecture to study potential degradation in sensory data input. Second, there is evidence that whilst Bayesian frameworks are capable of predicting medical outcomes in neurodegenerative patients (Halperin et al., 2020; Tussah et al., 2023), there is limited research in bridging the computational-experimental gap with respect to understanding subcortical visual pathways, prediction, and Parkinson’s Disease, due to conflicting debate between top-down and bottom-up signalling prioritisation. HPC’s architecture aims to examine both.

System and Methods to be Used

This project will focus on tackling its subject matter through two main approaches:

1. Undertaking gist perception tests in both normal controls and PD groups, using a variety of stimuli that can target or bypass the subcortical visual pathway in an effort to study activation. Subjects’ gist perception can be captured through functional neuroimaging (fMRI) to evaluate low and high levels of activations during gist perception tasks. Testing within the PD groups must account for the different stages of the condition, starting with initial mild cognitive impairment stage and covering initial diagnosis plus subsequent progression. This will help track progression of behavioural and neuronal changes in gist perception. Ethics procedures will be adhered to by maintaining confidentiality, consent, and privacy of all participants.

2. Building on the hybrid predictive coding model’s equation (Tscshantz et al.2023), in an effort to better represent and investigate the perceptual processes in PD brains, notable context-free perception processes. We can apply secondary experimental data, or large-scale datatsets, to train these models in an effort to capture the associations between data and participant variables in a probabilistic manner. To adapt this equation to mimic prediction uptake in PD brains, this project proposes the following key features:

2.1 Reduced Learning Rate

In a modified equation, we replace the learning rate α with a PD specific learning rate, α, which is weighted to represent non-control groups (See examples of this in Caso & Cooper’s model, 2020).

2.2 Increased Noise in Sensory Input

We introduce an additional term, ε, to represent an increase of noise or variability to sensory input experienced within PD.

2.3 Altered Prior Beliefs

Lastly, we can introduce a bias term to the equation, β, to represent a priori knowledge. The bias term is added to reflect the possibility of skewing in prior beliefs, impacted by PD. By incorporating this bias term, we can investigate how PD brains integrate prior knowledge with sensory input data, and how this may differ from neurologically adaptive brains.

Research Questions & Closing Statement

In short, this project seeks to investigate Parkinson’s Disease by combining experimental and computational paradigms. The aim will be to extend the existing work on hybrid predictive coding by manipulating the model’s hidden states, adjusting specific equation parameters, and generating predictions about impaired sensory input. The models will be continually compared and contrasted from the fMRI work we will undertake, and from peers or other experimental results emerging from the literature. The outputs from these models will then inform experimental paradigms towards Parkinson’s Disease, as well as differentiate between motor and sensory symptomatology.

As a result, this model will help move towards a more complete understanding of how individuals with Parkinson’s experience an overall context (e.g. detailed focal perception versus spatial mapping), and whether future research should discern between these distinct forms perceptual prediction. By investigating computational models which specify in understanding context-specific perception, we may capture a better picture behind the sensory input and movement dysfunction of Parkinson’s Disease, in turn leading to more successful medical outcomes, and better understanding of how humans ‘perceive’ on an idiographic scale.

References

Basso, M. A., Bickford, M. E., & Cang, J. (2021). Unraveling circuits of visual perception and cognition through the Superior Colliculus. Neuron, 109(6), 918–937. doi:10.1016/j.neuron.2021.01.013

Bernardinis, M., Atashzar, S. F., Patel, R. V., & Jog, M. S. (2021). Abnormal vision-based displacement perception in parkinson’s disease. Frontiers in Neuroscience, 15. doi:10.3389/fnins.2021.676469

Caso, Andrea and Cooper, Richard P. (2022) Executive functions in aging: an experimental and computational study of the Wisconsin Card Sorting and Brixton Spatial Anticipation tests. Experimental Aging Research 48 (2), pp. 99-135. ISSN 1096-4657.

Cleeremans, A. (2005). Computational correlates of consciousness. Progress in Brain Research, 81–98. doi:10.1016/s0079-6123(05)50007-4

Diederich, N. J., Stebbins, G., Schiltz, C., & Goetz, C. G. (2014). Are patients with parkinson’s disease blind to blindsight? Brain, 137(6), 1838–1849. doi:10.1093/brain/awu094

Evans, K. K., & Treisman, A. (2005). Crossmodal binding of audio-visual correspondent features. Journal of Vision, 5(8), 874–874. doi:10.1167/5.8.874

Friston K. A theory of cortical responses. Philosophical transactions of the Royal Society B: Biological sciences. 2005;360(1456):815–836. pmid:15937014

Friston, K., & Kiebel, S. (2009). Predictive coding under the free-energy principle. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1521), 1211–1221. doi:10.1098/rstb.2008.0300

Halperin, O., Karni, R., Israeli-Korn, S., Hassin-Baer, S., & Zaidel, A. (2020). Overconfidence in Visual Perception in Parkinson’s Disease. doi:10.1101/2020.09.09.289082

Kveraga K, Ghuman AS, Bar M. Top-down predictions in the cognitive brain. Brain and cognition. 2007;65(2):145–168. pmid:17923222

Lee TS, Mumford D. Hierarchical Bayesian inference in the visual cortex. JOSA A. 2003;20(7):1434–1448. pmid:12868647

Marr D. Vision: A computational investigation into the human representation and processing of visual information. CUMINCAD. 1982;.

Nemcova Elfmarkova, N., Gajdos, M., Rektorova, I., Marecek, R., & Rapcsak, S. Z.(2017).Neural evidence for defective top-down control of visual processing in parkinson’s an alzheimer’s disease. Neuropsychologia, 106, 236–244. doi:10.1016/j.neuropsychologia.2017.09.034

Rao, R. P., & Ballard, D. H. (1999). Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2(1), 79–87. doi:10.1038/4580

Rauss K, Pourtois G. What is bottom-up and what is top-down in predictive coding? Frontiers in psychology. 2013;4:276. pmid:23730295

Seth, A. K. (2015). Presence, objecthood, and the phenomenology of Predictive Perception. Cognitive Neuroscience, 6(2–3), 111–117. doi:10.1080/17588928.2015.1026888

Tscshantz, A., Millidge, B., Seth, A. K., & Buckley, C. L. (2023). Hybrid predictive coding: Inferring, fast and slow. PLOS Computational Biology, 19(8). doi:10.1371/journal.pcbi.1011280

Tushar, R. K. Patel, E. Aggarwal, K. Solanki, O. Dahiya and S. A. Yadav, "A Logistic Regression and Decision Tree Based Hybrid Approach to Predict Alzheimer's Disease," 2023 International Conference on Computational Intelligence and Sustainable Engineering Solutions (CISES), Greater Noida, India, 2023, pp. 722-726, doi: 10.1109/CISES58720.2023.10183466.

Uchiyama, M., Nishio, Y., Yokoi, K., Hosokai, Y., Takeda, A., & Mori, E. (2015). Pareidolia in parkinson’s disease without dementia: A positron emission tomography study. Parkinsonism &amp; Related Disorders, 21(6), 603–609. doi:10.1016/j.parkreldis.2015.03.020

Van Essen DC, Maunsell JH. Hierarchical organization and functional streams in the visual cortex. Trends in neurosciences. 1983;6:370–375.

Whittington JC, Bogacz R. An approximation of the error backpropagation algorithm in a predictive coding network with local Hebbian synaptic plasticity. Neural computation. 2017;29(5):1229–1262. pmid:28333583

Zhu Z., Zhang F., Hu H., Bakshi A., Robinson M. R., Powell J. E., et al. (2016). Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat. Genet. 48 481–487. 10.1038/ng.3538

Verbal parity, in its essence, represents a machine's capacity to engage in textual discourse with the finesse and adaptability of human interpretation. It's a narrower aim than AGI, unburdened by the need for visual acuity, motor dexterity, or real-time physical interaction. Yet in its specificity lies power and proximity.

To fully grasp the implications of verbal parity, we must first divest ourselves of the shackles imposed by the Turing test paradigm. While Turing's thought experiment has long served as a touchstone for AI capabilities, verbal parity transcends this framework. It's not about deception or mimicry, but about genuine linguistic competence across a broad spectrum of domains.

The trajectory towards verbal parity is not merely theoretical; it's empirically observable. The past half-decade has witnessed a veritable explosion in natural language processing capabilities. Models that once stumbled over basic semantic tasks now navigate complex linguistic landscapes with apparent ease. This rapid progress begs the question: how close are we to the precipice of verbal parity?

If current trends persist - and there's little reason to suspect they won't - we may find ourselves face-to-face with verbally proficient AI within a startlingly brief timeframe. Conservative estimates place this milestone within the next decade or two, but more bullish projections suggest it could materialise in as little as five years.

The ramifications of such an achievement are difficult to overstate. Entire sectors of the economy, particularly those dominated by white-collar professions, stand on the brink of radical transformation. Any job that primarily involves textual communication - from legal analysis to creative writing - could potentially be automated. The socioeconomic ripples of such a shift would be profound and far-reaching.

Moreover, verbal parity could serve as a catalyst for a form of limited singularity. While not as all-encompassing as an AGI-driven intelligence explosion, a verbally superintelligent AI could potentially accelerate scientific research, technological development, and even AI advancement itself, creating a recursive loop of rapid progress.

This prospect, while exhilarating, also casts the AI control problem in a new light. The timeline for developing robust AI safety measures may be significantly shorter than many have assumed. We find ourselves in a race against time, needing to solve complex ethical and technical challenges before verbal parity becomes a reality.

It's worth noting that current language models, impressive as they are, still fall short of true verbal parity. They lack the ability to learn and adapt in real-time, and their context windows, while expanding, remain limited. However, these hurdles seem surmountable in the near term, especially when compared to the Herculean challenges that AGI presents.

Intriguingly, the pursuit of verbal parity forces us to confront fundamental questions about the nature of intelligence and consciousness. Can a system achieve human-level linguistic competence without possessing genuine understanding or self-awareness? This query echoes long-standing debates in philosophy of mind and cognitive science, from the Chinese Room argument to more recent discussions on the hard problem of consciousness.

As we stand at this critical juncture, interdisciplinary collaboration becomes not just beneficial, but essential. Fellow computer and data scientists must engage with linguists, cognitive scientists with philosophers; AI researchers with ethicists. Only through such cross-pollination of ideas can we hope to navigate the complex landscape that verbal parity presents.

In the spirit of discussion, I propose  the following predictions, fully cognisant of the inherent risks in such optimistic predictions:

1. Verbal parity will be achieved independently of and prior to artificial general intelligence. (85% confidence)

2. We will witness the emergence of verbal parity within the next 15 years. (60% confidence)

3. The advent of verbal parity will trigger significant disruptions in white-collar employment within five years of its achievement. (75% confidence)

These predictions are not mere academic exercises. They serve as a clarion call to policymakers, industry leaders, and researchers alike. The potential advent of verbal parity demands proactive consideration and preparation across multiple domains of society.

As we venture into this uncharted territory, we must remain vigilant yet open-minded. The development of verbally proficient AI will undoubtedly challenge our preconceptions about intelligence, consciousness, and the unique qualities of human cognition. It may force us to reevaluate long-held beliefs about the nature of understanding and the boundaries between biological and artificial intelligence.

In this context, the philosophical insights of Saul Kripke offer a valuable lens through which to examine the AGI issue, and by extension, verbal parity. Kripke's work on naming and necessity provides a framework for understanding the fundamental nature of intelligence and consciousness that transcends mere behavioural or functional equivalence. His notion of rigid designators - terms that refer to the same object in all possible worlds where that object exists - challenges simplistic equations of intelligence with behaviour. Applied to AGI and verbal parity, this suggests that even if we create machines that perfectly mimic human linguistic behaviour, we cannot assume they possess the same kind of intelligence or consciousness that humans do. The "intelligence" or "understanding" we attribute to AGI might be as different from human intelligence as water is from H2O in Kripke's famous example - functionally equivalent, perhaps, but metaphysically distinct. This Kripkean perspective urges us to look beyond mere performance metrics in our pursuit of AGI and verbal parity, and to grapple with deeper questions about the essential nature of intelligence and understanding.

In conclusion, while the quest for AGI continues to capture headlines and fuel speculative fervour, we would do well to turn our gaze to the nearer horizon. Verbal parity, with its profound implications and imminent potential, represents not just a technological milestone, but a pivotal moment in human history. It is a threshold that, once crossed, may irreversibly alter the landscape of human-machine interaction, cognitive science, and our very understanding of intelligence itself.

I propose the following definition:

(M) Something matters when it possesses significance sufficient to merit concern.

A Commonsense Worldview

This simple formulation invites deeper reflection. Two key ideas emerge:

Firstly, the belief that some things matter is fundamental to our common understanding of the world. It's how we intuitively perceive our environment before engaging in philosophical inquiry. Some philosophers dub this as the “common sense” view of the world - in other words, the assumption we carry prior to any logical or rigorous investigation beyond what we perceive.

Secondly, this pre-theoretical notion of mattering implies the existence of real, discoverable value in the world. This contrasts with more contemporary views suggesting that value is merely projected, constructed, or imagined.

From the perspective of commonsense, a diverse array of entities matter. Most prominently, we consider individuals to matter - our family members, friends, companion animals, and associates. Our major life goals and commitments also rank highly, as they orient us and imbue our existence with subjective meaning.

More abstractly, many believe that certain classes of beings matter intrinsically, such as humans, sentient creatures, or rational agents. Even inanimate objects can matter: revered artworks, monuments, historic buildings, cherished books, and family heirlooms often garner significant concern. Natural features like ancient trees, ecosystems, and geological formations are also frequently regarded as mattering.

This brief survey highlights an important point: there exists considerable disagreement about what matters, to what degree things matter, and which things matter intrinsically. Yet the very existence of such disagreements underscores our shared belief that some things do indeed matter.

The Metaphysics of Mattering

This belief in mattering manifests in our actions and moral judgments. We help our loved ones because they matter to us. We consider harming others wrong because we believe people matter. We would find it unconscionable to destroy historically significant sites for trivial reasons, as we deem such places to matter.

Of course, the belief that something matters does not necessarily mean it actually does matter. Perhaps nothing matters, as Nihilism posits. However, even if this were true, understanding our belief in mattering remains crucial to explaining human behaviour and motivation. If we believed “nothing mattered”, then while we might still act as if the objects of our own desires (whatever those might be) mattered, we would be deprived of any idea of having ends that ought to be pursued; of obligations that ought to be met. Additionally, nor would we have any sense of there being objectively “better” and “worse” desires to manage. There would be no difference between pursuing a worthwhile, satisfying career versus rotting in bed all day if this was the case.  Of course, there will be people who do not hold such distinctions, but these individuals are few and far between, with possible mitigating, neuropsychological factors influencing their perspectives. In other words, Nihilism is not the view of commonsense.

We need to be able to understand the bedrock to what it means to something mattering. When we say that something matters, then it matters either intrinsically or instrumentally.  If something matters intrinsically, this means that it matters in itself, rather than in virtue of acting in some kind of relation with something else.  On the other hand, if something matters instrumentally, then the “something” doesn’t carry this instrumentality, rather, we evaluate it only in virtue of its connection to something that does matter in itself (or that is instrumental). As this definition outlines instrumental mattering to be essentially linked to intrinsic mattering, I will not make much of the distinction between the two. So, delving deeper beyond these definitions, we can distinguish between two concepts:

Stance-Independent Mattering (SIM): The facts concerning mattering are fixed by reality, independent of anyone's beliefs, preferences, or attitudes.

Stance-Dependent Mattering (SDM): The facts concerning mattering are determined entirely by the beliefs, preferences, or attitudes of agents.

In the final analysis, only Stance-Independent Mattering (SIM) truly captures the essence of what it means for something to matter. This concept, while implicit in our earlier definition, is not necessarily presupposed by it. It's crucial to distinguish between what matters to someone and what objectively matters. We can, after all, be mistaken about what matters - consider extreme cases like Hitler’s idea of what was important to him, or more mundane examples, like someone who believes life's ultimate purpose is collecting Pokémon cards. These examples illustrate that mattering isn't merely a subjective mental state.

Take suffering, for instance. If suffering matters in a stance-independent sense, it's because something inherent to the experience of suffering itself - its unpleasantness or intrinsic negativity - provides a reason for our concern. Thus, SIM doesn't reduce mattering to mere psychological states.

Moreover, SIM doesn't suggest that we create or manufacture what matters. Rather, it allows for the possibility that some things matter regardless of anyone's beliefs or desires. If this aligns with our common-sense view of the world, then we're implicitly committing to SIM. Consequently, rejecting SIM means departing from this intuitive perspective, at least to some degree.

While this doesn't conclusively prove SIM's validity, it does offer some support. This follows from the philosophical principle that our common-sense commitments warrant prima facie justification.

Endorsing SIM as the true nature of mattering leads us to recognize its essential connection with Value Realism - the existence of stance-independent value. Without Value Realism, there would be no objective value in the world to determine what matters. In essence, SIM logically entails Value Realism.

To conclude, anyone adhering to a common-sense worldview must seriously consider the concept of mattering. We all operate as if some things matter, and more specifically, as if they matter in the way described by Stance-Independent Mattering (SIM): where facts about what matters are determined by reality, independent of anyone's beliefs, preferences, emotions, or attitudes.

This commitment to SIM, in turn, binds us to Value Realism: the existence of stance-independent value. Ultimately, this perspective allows for the possibility of being mistaken about what matters, as it posits that some things genuinely do matter - possessing intrinsic positive or negative value - while others do not.

It makes sense that we rest on our laurels a bit too much, at the best of times. At our core, we live out our days (and by extension, our lives) as a result of the actions we take. These actions are derived through probabilistic phenomena that you’re likely unaware of. How should we get to work? What do I want for breakfast? Should I wear trousers or jeans today? Whether you’re cognisant of this fact or not, you probably make thousands of these kinds of decisions a day. On a daily basis, we can see how our decision-making processes are predicated upon gut feelings and pre-curated heuristics. Say you’re walking home at night, and see a young person, dressed in a hoodie and casual attire. The first thought may (or may not - you’ll be a far better person than I) be to cross the road, or try to cut your walk home a little short. These heuristics can be mental minimaps, or ‘shortcuts’, designed to essentially make our cognitive processes a little less exhausting in the sea of stimuli we expose ourselves to on a regular basis. But as you can see from this example, they can also create biases. They’re the LLM of our decision making; capable of extracting swathes of information and training our brain, whilst risking larger, more blatant errors or ‘hallucinations’.

In short, if we live guided by heuristics (or as some may articulate, “vibes”), society would plummet into chaos. If we are indeed, in essence, irrational agents, then we have the schools of philosophy and science to thank for any semblance of societal structure. It’s heralded as a methodology so useful in understanding our present, limited reality, as it places our own biases and limitations in the centre of things, and emphasises a framework of externalising and testing these blindspots.

Scientific Method vs. Science

 It’s important we differentiate between the term scientific method, and the holistic system derived from other branches of the term science. The latter consists of a body of individuals who participate in - and disseminate - the ‘system’ that is science, through journals, research facilities, industry, and universities. The scientific method is a distinct, a priori method which is considered separate from science. We can have the scientific method without science, but we cannot have science without the scientific method.

Something that is inherently core to this method is a concept called falsifiability. In other words, a theory must be testable, and consequently, open to rebuttal when opposing evidence comes into the fray. Philosopher Karl Popper concerned himself with this notion of falsifiability in a good number of ways. There’s a Popperian classic which continues to be applied in understanding the core values of the scientific method even today. Suppose one theory posits that all swans are white. How can we verify that this theory is valid? The obvious method would be to check that every swan is indeed white. Yet herein lies the problem. No matter how many white swans you count, you can never truly be 100% certain that there isn’t a black swan lurking around somewhere. Menacingly.

By this logic, you can never prove that the theory is true. Strikingly, the mere sight of one black swan tears this theory into tatters. Falsification is a brutal, honest beast: a single example can decimate a universal statement. Popper considered this as a well-developed branch of deductive logic.

Falsification is a useful tool in taming our basest decision-making. Offering facts and evidence to intuitive, knee-jerk assumptions can cull even the most crepuscular of biases, which is why they can be so informative in this present conversation landscape. As discussed before, reason and logic aren’t always what guides our decision-making - particularly when emotions are compromised. Instead, we behave in ways which reaffirm preconceptions, core beliefs, and even moral frameworks which have been ‘intuited’ (as opposed to argued). These moral assumptions can lead back to nearly all of the major political, social, and cultural discourse. Morality-justified reasoning, such as the just-world hypothesis (amongst other fallacies), can lead to people evaluating arguments in a biased approach to reach a conclusion that they prefer.

(This is possibly one of the reasons why I resent this notion of “building an online brand” using reels and Instagram fodder - you have to appear deathly certain of a position, or people consider you wishy-washy, as opposed to reasonable and measured).

We can see that this capacity to apply moralistic fallacies extends not only to what we believe ourselves, but towards what we think others ought to believe. A series of experiments conducted in the 2021 paper by Cusimano & Lombrozo exposed the participants to fictional scenarios and personas, within which contained one factually accurate write-up of the scenario, a belief-driven justification, and a moral justification for the belief-based account. The confederates of these experiments asked the participants which account that the fictional character should be more inclined to believe. The qualitative reports concluded that participants unconsciously applied non-evidential norms - essentially, more emotive arguments - as they proposed that the fictional protagonist ought to buy into the belief-driven account more than the factual report.

Sense of Hume-r

Interestingly, these participants fell into the assumption that a moral justification of the events outweighed the importance of a factual account. They decided what these fictional characters ought to believe in a given paradigm, even when a moral justification contradicted the historical accuracy of events.

According to David Hume’s famed aphorism, you cannot derive an ought from an is. In essence, projecting a specific value system from an “ought” towards another, purportedly reality-based “is” creates a problem in our discourse. We can apply morality-driven “oughts” across an array of societal discussion; gun laws, AI, technological advancements, political conflict, just to name a few, and make arguments which can impede evidence-driven cases. At face value, this is a sound idea: elucidating belief-centric arguments would help dilute the lack of scientific rigour when making bold claims, especially in the school of psychology and human behaviour.

There are a couple of ways in which we can challenge Hume’s claim, however. For one, some philosophers contend that the is-ought issue goes beyond a falsification issue, and more to do with language and semantics.

The is-ought problem, as articulated by Hume, is when a philosophical treatise moves from making "is" claims to making "ought" claims, without explaining how the shift is made. Claiming that same-sex marriage is “bad” because of theological reasons, for example, doesn’t sufficiently (or necessarily) explain the shift which warrants an “ought” which we need to abide by, for example. But what if we do explain the shift? Do our moral “oughts” become automatically valid?

Attempts to refute Hume's law take issue with this specific, linguistic articulation of the problem. For example, in The Collapse of the Fact/Value Distinction, Hillary Putnam writes:

One clue that the claim presupposes a substantial metaphysics (as opposed to being a simple logical point) is that no one, including Hume himself, ever takes it as merely a claim about the validity of certain forms of inference, analogous to the claim “you cannot infer ‘p8Cq' from ‘p or q.’” Indeed, if the claim were simply one about the form of certain inferences, it would prohibit one from ever inferring “you ought to do x in such-and-such circumstances” from “for you to do x in such-and-such circumstances is good, and for you to refrain from doing x in those circumstances is bad.

Of course, many philosophers would reply to this example by saying that it does not run afoul of Hume’s dictum because it is a case of inferring an “ought” from an “ought.” But that is my point. Their ability to recognize statements such as “for you to do x in such-and such circumstances is good, and for you to refrain from doing x in those circumstances is bad” as a case of an “ought” turns not on any feature of the form of the statement but rather on an understanding of its content.”

“Nor did Hume himself (or any of his readers) understand the claim as one about the canons of formal inference. Rather, Hume assumed a metaphysical dichotomy between “matters of fact” and “relations of ideas” (the dichotomy that constituted his early anticipation of “the analytic-synthetic distinction”). What Hume meant was that when an “is” judgment describes a “matter of fact,” then no “ought” judgment can be derived from it. Hume’s metaphysics of "matters of fact” constitutes the whole ground of the alleged underivability of “oughts” from “ises.”

Another important theoretical development since Hume’s argument can be found in the work of logician Arthur Prior. Prior demonstrates that, by starting with a set of “is” claims as well as a set of “ought” statements, we can then create a logically coherent claim which actually allows us to derive an ought from an is.

Consider the following, which I’ll label SC: “either it’s sunny or cats are cute.” 

Whether we consider SC an is or ought claim, we’re still going to be able to derive an “is” from an “ought.” We can see this in this example if SC is an “ought” claim. It’s predicated by “it’s sunny”. And even if people consider it an “is” claim, we can still derive an “ought”. We can derive “cats are cute” from SC plus “it’s not sunny”. Therefore, whether SC is an “is” or whether it’s an “ought”, we are still able to derive this “ought” from a given “is”.

This is just one logical example of challenging Hume’s concerns towards ought-is statements. With this, we can start to see some of the potential flaws within deductive reasoning. If a premise is flawed or flimsy, it doesn’t matter how well you present your arguments; the foundations are shaky. For another article, I’ll be delving into more arguments countering the purely empiricist and sceptical account held by some fine thinkers such as David Hume.

The reason is this: we cannot only address the passioned, illogical claims by poor deductive reasoning across debates without acknowledging the gap in moral reasoning. “Subjective morality” is a fragile, ill-conceived concept; we’ll be exploring this in more detail in future articles.

An interdisciplinary approach to provide insight into willful behaviours is becoming more and more prominent. The combination of game theory and computational reinforcement suggests robust pathways in understanding our boundaries for what we choose and why, and by extension, when things go “wrong” in specific neurological and psychological disorders. These fields operate within highly specific, quantitative landscapes, but can be integrated successfully into experimental paradigms (such as fMRI and electrophysiologial measures) to study the experiences and behaviour of patients. This approach does a great job at better understanding neurobiological mechanisms, but then what? And so what? Can it understand choice in a way that eradicates the argument of free will entirely? Can we truly describe the brain in a purely computational fashion?

There is a tangible, impalpable sense of secularism, inculcated in cultural discussion. If one fails to scrutinise internet dialectic for long enough, the view that any and all things in our knowable universe can be resolved by the scientific method seems to be the default position. As a computational psychologist with a philosophy background, I can assure you: holding an argument defending either free will, or the notion of a deity, or something beyond a standardised determinism position, or hell, even stating something as daring as “I don’t really know”, is a feat which requires strong disposition and equally hearty will. In short, it is not always the popular position amongst neuroscientists or computer science proponents.

Despite this, prominent figureheads (with respectable arguments) put forth a non-deterministic view. David Chalmers is one noted philosopher, proposing that consciousness sets a very high bar for us to reach; namely, the “hard problem” of consciousness muddies the waters from which we try and understand our human minds. As a naturalistic dualist, Chalmers does not believe that consciousness is something attained from physical theories. Instead, Chalmers argues that the presence of wave-function collapse points consciousness being a fundamental, non-material aspect to our universe. This viewpoint is known as panpsychism, and argues against the materialist position through the following syllogisms:

1. In our world, there are conscious experiences

2. There is a logically possible world physically identical to ours, in which the positive facts about consciousness in our world do not hold.

3. Therefore, facts about consciousness are further facts about our world, over and above the physical facts.

4. So materialism is false.

(The Conscious Mind (1996), p.123)

Information, it seems, according to Chalmers, might be the non-physical property of dualism. This information, however, raises the ugly head of the hard problem in that every conscious experience from us all is considered subjective:

There is something it is like to see a vivid green, to feel a sharp pain, to visualize the Eiffel tower, to feel a deep regret, and to think that one is late. Each of these states has a phenomenal character, with phenomenal properties (or qualia) characterizing what it is like to be in the state.

The Blackwell Guide to Philosophy of Mind. chapter 5, "Consciousness and its Place in Nature"

Qualia, the subjective or experiential properties of our mental lives, has proven a contentious topic amongst philosophers and scientists alike. Having been catapulted into intelligentsia owing to a hotpot of technological advances, globalisation, a rise and fall of secularism and religiosity alike, and greater demands for understanding mental health, understanding what it means to “be'' is one of the 21st Century’s key existential questions. Interestingly, there was a time when the field of psychology has little interest in the function of the brain. Since the demand of an interdisciplinary approach for philosophers and scientists alike, psychologists are now advised to have some decent command of connectionist models or understanding of cognitive theories. Whether this is a good thing - that we understand the brain, and its insight into human behaviour overall - or if it detracts from the sociological emphasis once held by psychology, remains to be seen. At least for now.

This is where things get messy. Scientism - the position that philosophy is not needed to adequately answer these tough questions; that the scientific method can account for everything down the line - places greater onus on philosophers to hold down the fort on consciousness theories, either rehashing Chalmers’ sentiment that consciousness is thus far impenetrable, or trying to reconcile with materialist accounts from scientists. Even deterministic models aren’t immune to the critique of scientism. Take for example Karl Friston, a notable intellectual who has contributed heavily to this newer, integrated approach of computer science and cognition. Yet even still, experimental neuroscientists are quick to label his free energy principle  “the new Freud” of empirical science. It seems the back-and-forth between sciences, philosophy, and repackaging of approaches is yet to come to a peaceful middle ground.

Consequently, there comes a greater resentment for theories that resemble anything less than a determinist’s paradise. Philip Goff, panspychist and “practising agnostic”, acts as a surprising advocate for the fine-tuning argument. There are uncontroversial facts in physics which support this approach, and which Goff uses to support the idea of idealism in consciousness (and you don’t have to agree with this argument, to be clear): for example, the “nuclear force” which binds together the elements in an atom’s nucleus is represented in physics by the number 0.007. If the value had been 0.006, the universe would have contained nothing but hydrogen. If it had been 0.008, all hydrogen would have burned off in the Big Bang and water would never have existed. Even further, if the mass of the “down quark” – an elementary particle, necessary for matter to exist – altered from what it is, once again the universe would only have contained hydrogen, rather than the 60 million chemical compounds we know exist.

Discussions of quarks and nuclear force can be dizzying, so let’s simplify it even more. If the electron mass differed even by a tidbit, the universe would contain zero atoms and zero chemical reactions. In essence: zero of the world we know. Goff tackles consciousness with an idealist approach; whilst materialist believes our mental processes - our subjective experiences - are driven purely by physical phenomena a la Friston, idealism asserts that consciousness is the fundamental substance of all nature. In other words, according to this model, it is a prerequisite to our behaviours, experiences, and physical responses. Whilst this doesn’t address the ethical or moral arguments for a designer or God (that’s for another Substack; this piece is already getting tired…), it does satisfy the irrationality of assuming these perfect environmental tweaks are “mere flukes”, thus putting materialism as an approach into question. So where does this fit into neurocomputational approaches? Will I be forever alone in my non-materialist beliefs?

One key “solution” I see put forth by many computationalists - notably in the neuropsychology fields - is that we can construct a mind through a sort of machine-based homunculism. In other words, if computers can learn billions of patterns, and brains can too, it would, in theory, make sense to assume that we can mimic neuronal structures through specific AI and machine learning algorithms. As such, there is nothing wrong with describing the brain as involved in information processing, as there is nothing wrong with describing the telephone system in the same vein. Still, we have to recognise that similes are entirely derivative upon the fact that these systems are connected to people who have real conscious informational states. And just to add - these states are dependent upon individual brain activity.

So, like him or loathe him, Chalmers’ notion of a “hard problem” of consciousness comes in clutch here. All information is information—to some conscious agent. Accordingly, neurons do not, considered in themselves, process information or send signs or receive messages. Assuming so is reminiscent of the homunculus fallacy. So far, taking a philosophical analysis, we can state that large language models (LLMs) such as ChatGPT can produce text, and even display instances of critical thinking and creativity, but they cannot replicate consciousness. This is because consciousness can’t be observed by looking inside your head, and ergo, cannot be practiced or taught to AI. In their attempts to find a connection between brain activity and experience, neuroscientists must rely on their subjects’ testimony, or on external markers of consciousness. But there are multiple ways of interpreting this data - fMRIs has a real limit to how much it can show regarding neural correlates, as every pixel corresponds to roughly 5.5 million neurons. Perhaps it is that consciousness must have evolved because it made a behavioural difference. Systems with consciousness must behave differently, and hence survive better, than systems without consciousness. This theory adequately satisfies a theistic (or gnostic) account to our existence.

From a philosophical standpoint, there are robust arguments from which we can argue the existence of God, idealism, or free will. It wasn’t until I studied more metaphysics and analytical philosophy that I began to unpick the follies in my own positions. Or rather, I’d call it a stubbornness; to be wholly frank, a brattish ignorance, fed by edgelord quotes from Nietzsche and Camus. Studying mathematics, neuroscience, history and philosophy as a whole has made me much more sceptical to the atheist movement, whilst cultivating a preference to idealism. It has been a valuable lesson in emphasising how and why I think the ways I do - not just picking conclusions from thin air.

Read more

If you’ve picked up a Final Fantasy game, you’ve picked up Berserk. If you’ve banged your head against a wall getting through Dark Souls, you’ve undergone similar frustrations and tests of fortitude as Miura’s protagonist, Guts. Hell, if you’ve watched any episode of Games of Thrones, Berserk’s influence stealths itself amongst certain scenes. In short: Berserk has become nothing short than a canonical hallmark of fictional prowess. It has indirectly impacted the mainstream ventures into flight and fantasy we see today - even if the original source remains unspoken as a comic book store gem; a cultural hallmark of whispering subreddits.

Despite such an untimely death, Miura’s legacy remains everlasting. No doubt one thing fans of the series will look back on fondly is the vying anticipating of chapter release month. Unlike other mangakas, Miura was notable for his long pauses between chapter releases. Being a Berserk fan meant playing Storyteller Roulette: one could wait a couple of months or a couple of years until seeing Guts’ tale unfold further. This was a double-edged sword. Obviously, nobody likes to be ghosted by their brooding, mercenary hero. So Berserk created what any other cultural zeitgeist would: a borderline maniacal community. The gaping voids in our hearts was filled by connection with other fans, haphazard reading interpretations, wild plot theories, and unsavoury meme reposting.

WARNING: SPOILERS AHEAD & CONTENT WARNING, S*XUAL ASSAULT DISCUSSION

Some discussions became somewhat wearied, but were still welcome by veteran and newbie fans alike. Will Guts finally manage to give Griffith what for? Will he venture out on his own, leaving the party behind? Is Femto Griffith or a separate entity? These are all pretty cool things to prattle on about.

Other subject matters require more sensitivity; fewer hypotheticals. Berserk is a story which dare not shy away from the abject horrors of its - and implicitly, our - world. Events such as the rape of Casca, and managing mental illnesses like PTSD and depression, are portrayed with such rawness and pain, one begins to question how Miura managed to mine out this richness, this honesty, from the recesses of his psyche.

What makes Berserk so beautiful, even when its pages are splattered with gore, grief and galvanising despair, is its unique take on masculinity through the depiction of Guts. Berserk is one of those rare gems of a story that centres a stereotypically “strong” man trying to overcome PTSD and trauma through various methods, some attempts more successful than others. In a culture where “red pill males” might worship a physically strong warrior typified by Guts’ visual design, it is both refreshing and compelling to see the same persona grapple with mental illness and his own perceptions of being “manly”. Guts is a fantastically subversive character; indeed, Miura purposefully gives him a background often attributed to fragility and femininity, ironically by the same types of men who’d eschew the notion of vulnerability or emotional validation for the “manly chad”.

Guts’ journey acts as a foil what we’re told to believe about these ultra masculine men - yet ironically because of this rawness, this bravery imparted in the face of his suffering, he is terrifyingly “manly” at the same time. Just not in the way those redpillers might first think.

Even if you haven’t read any of Miura’s story, it’s not hard to see that Guts’ visuals are designed to get us thinking in terms of “toughness”. A young man draped in black leather and shining armour, he is brash, pretty rude, and not afraid to get stabby-stabby on the battlefield. For a lot of first time readers, and even upon retrieval, his portrayal is jarringly “edgy” - easily mistaken as a two-dimensional trope, rather than masterful character building. It’s not until the Black Swordsman Arc that the readers are exposed to a simmering glimpse of Guts’s deep-seated pain and trauma.

A wordless pane evokes a thousand feelings. The reader sees a downtrodden swordsman, a quivering brow, and eyes stinging with tears. The contrast from the Guts we knew prior is devastating. At last, we have evidence that there is something more to this man than meets the eye. The transition or back-and-forth between angry edgelord to vulnerable, weeping man is what makes Black Swordsman the most intriguing arc of the series. Readers are now exposed to the truth of the matter: that Guts’ bravado is a fragile facade to help him survive his inner (and literal) demons.

This added character dimension if further explored within the Golden Age Arc. The reader sees his childhood traumas explained all whilst Guts makes his mark in the mercenary world. The face Guts puts forth for others is not the same as his internal, consistent strife, with which he must deal with in a manner that causes him more strain than dispatching hundreds of soldiers in one go. It’s almost hard to believe that this Herculean figure who slaughtered demons and played witness to the death of his many comrades was once a helpless child, abused and discarded. Based on what we know in psychology today, however, in many ways, it makes sense.

Through Guts’ character, Miura has not only created a foil to the ultra-masculine, “red-pill chad” archetype; he has challenged the simplistic take much of the media uses to describe the concept “masculinity”. For a warrior who hones such dominance and aggression on the battlefield, it appears contradictory that he is a survivor of rape and child abuse, a victim of betrayal from his close comrade, and a man haplessly in love and starved of companionship. Yet despite navigating the treacherous, blood-soaked world in which he lives, managing bouts of PTSD and depression alongside what it means to be human, Guts sticks a middle finger up towards the unrealistic and unhealthy depictions of “manly men” in mainstream media. His journey is unexpected, yet one we all root for: what was once a character who was apt at the ol’ slash-and-hack is now an individual who becomes aware of his own emotional pain.

Even better, Miura manages to cultivate a sensitive, delicate discussion on what it means to be “masculine”. He doesn’t condemn those who choose the stoic, reserved route. Instead, he simply leaves the door open for a man like Guts to express and feel. In other words, masculinity becomes a fluid concept; yielding according to what nourishes Guts and provides him with strength. In these instances, sometimes it’s owning the battlefield; in others, it’s reflecting on thoughts and feelings. In the world of Berserk, masculinity carries neither pressures nor stigma. It is a source of strength, but that strength can look like emotional intelligence, companionship, human connection, and self-awareness as much as it represents the sword.

We are influenced by the media we consume, it goes without saying. For readers everywhere, Guts is a positive influence, countering the unhealthy depictions of masculinity and mental illness we are bombarded with on a sociocultural level.

Miura created a masterpiece. One that has changed - and will continue to change - the way in which we lead our lives. Undoubtedly, the pang of grief we feel for his passing feels all the more painful, as a result.

I’d picked up a Final Fantasy game off the back of some recommendations after hearing ravings on its story, character writing, and its ample opportunities to swing a cartoonishly-long sword.

Now, I love me some DnD, I love me some Zelda, so at this point, I was pretty much sold into my foray into the Final Fantasy series.

Also, the chocobos. The chocobos got me real good.

In any case, there was another reason for me picking up Final Fantasy 7 in particular. As a story, it addressed one giant, sobbing-until-snot-drips-out-of-its-trunk elephant in the room. I was grieving, deeply. I was a young girl and had just recently experienced the unexpected loss of my oldest sister. And I needed something, someone, anything, to reach out to.

All to say, I was a husk. Days before the news of her death reached me, I was on autopilot. You know when they describe that nausea, or pit in your stomach, when you feel sick with worry or stress? Well, I’d say this was more of a wicker basket, which nestled itself into an insidious knot, festering away in the depths of my stomach. One car drive home that week, the beats of silence between myself and my grandmother were particularly serried.

I don’t remember how long it took to get back. But I distinctly remember that woven basket beginning to tauten across my abdomen. To asphyxiate.

Stepping into the door, the air ran cold with the morose vapour of endless sobbing. My sisters rushed over.

What the fuck was going on.

And that’s when I put two and two together. I bolted up to my mother’s room, where we both wailed, crushed by the weight of the unspoken horror set before us.

Emma was dead.

“AERITH IS GONE.”

Freeze-framing back into Midgar, our favourite emo protagonist, Cloud Strife, faces numerous deaths and losses in quick concession.

At the time, and even now, diving into stories (be it on paperback or through beautiful code) was the only way I managed to grapple with the cacophony of grief within. As I was playing the role of trauma-addled mercenary, I felt a deep recognition of Cloud’s feelings of helplessness and apathy.

Loss, and learning to face that loss, is a theme peppered across the span of the roughly-37-hours of playtime in Final Fantasy 7. At a time when the “sacrificial hero” trope was rife in fiction and media, FF7 challenged the interpretations of grief and loss head on. Death was no longer this sacrificial badge of honour for which one should yearn. It is cold, empty; the feeling of missing a step on the staircase in the darks, your stomach turning. One moment in the game encapsulates this senselessness in such flavour, I still find it hard to watch the cutscene unfold to this day. In an act of abject cruelty, life of Aerith Gainsborough, a beloved member of Cloud’s party, and the epitome of free-spirited hope and optimism, is snuffed out. There is no greater sacrifice, no greater reason, but a pained feeling of being wronged by fate, and a virulent helplessness in the fact of your loved one’s doomed mortality.

Perhaps I say this with bereaved-tinted glasses, but Aerith’s death hit me harder than I could have anticipated. I connected so many dots to Aerith and Emma via my play-through; both kind, compassionate, free-spirited young women who sought to help others no matter the obstacles faced, possessing dreams of their own. Dreams violently cut short by fate’s indiscriminate scythe.

The gut-punch of grief I experienced when my sister died felt so visceral, felt so similar to Cloud’s own sense of disarray in the wake of Aerith’s own doom.

“Aerith is gone. Aerith will no longer talk, no longer laugh, cry...... or get angry......What about us...... what are WE supposed to do? What about my pain? My fingers are tingling. My mouth is dry. My eyes are burning”.

The articulation of Cloud’s pain hammers home how primal grief can be. There’s no mourning Aerith’s intellectualisms, her abstract ways of thinking, or her contributions to the party’s goals. When someone dies, we mourn their humanity. We mourn the way they laugh (I am both grateful and horrified to say I have inherited Emma’s same goofy snort), or how loud their slippers are when they hit the kitchen floor, clinking mug against spoon in the early hours. We miss the funny expressions they make when they’re thinking really, really hard about something utterly bonkers. We miss the softness and depth of their hugs. Cloud missed Aerith as a human being, as her rightful self, not for what she could do for him or the game’s party. He missed her personhood, her brightness.

THE MEANING GRIEF GIVES US

Viktor Frankl was a famed psychotherapist who sought to explore positive existentialist views on suffering. It still astounds me to this day how someone can articulate such powerful ideals given the utter pervasive horrors one human (of many) had to face. During his period in the Auschwitz concentration camps, he began to contemplate how one carves out meaning in the face of relentless pain. He held the view that grief and trauma could be reframed as a crisis that affords us new meaning and purpose in life.

This isn’t to say that “whatever doesn’t kill us makes us stronger”. Sometimes, it leaves marks and bruises and wounds. It scorches our cognitive scripts, branding into us the blueprint that imply the world is a very scary place indeed. These etchings may take years of pain, medication, and therapy to scrub out.

Yet as Frankl poignantly reminds us:

“We must never forget that we may also find meaning in life even when confronted with a hopeless situation, when facing a fate that cannot be changed. For what then matters is to bear witness to the uniquely human potential at its best, which is to transform a personal tragedy into a triumph, to turn one’s predicament into human achievement. When we are no longer able to change a situation… we are challenged to change ourselves.” (Frankl, 1969, p135).

So when somebody we love dearly in our lives departs from us, be it via unexpected tragedy or even while they’re still alive, the role of meaning-making makes a massive difference in how we can continue with our lives again.

Cloud’s character arc perfectly encapsulates this need for personal transformation in the face of loss. He is forced to examine the way he views himself and his role in the world. He is forced to figure out how to carry on with his own life and carve out an identity without the help of a dear friend. He is forced to relinquish his survivor’s guilt, and live in a way that protects himself, his party members, and honours Aerith’s memory that makes him a very well-written protagonist indeed.

In my own world, Emma’s death marked the turning point of my own character arc.

I became responsible, not reckless; I saw wonder and excitement and gratitude and colour in a world that once seemed implausibly grey.

I recognised and respected my impermanence: an iridescent blur, flecked into the ebb and swell of our cosmos.

I saw a twinkle of adventure in the entropic maelstrom otherwise known as…well, life. And that’s the thing. Death - and grief - compels us to confront the reality of our short stay on this giant, political mess of a rock. Our own mortality gets us asking painful, important questions.

Who do I want to be? How do I want to treat people? What can I do to lead a life I am proud of? Why do I choose this path or way of living?

How can I leave the world a little bit brighter than when I last left it?

Aerith’s death helped me, a traumatised teenager going through a really bad emo phase, grapple with what it meant to experience loss.

And as both Frankl and Final Fantasy 7 taught me, grief can act as a harbinger of meaning, legacy, and most importantly, love.

In light of her upcoming anniversary, I thank Emma everyday for the light she has left behind for me; the sprinkling of her own stardust freckled amidst dark nooks.

I can still feel her. I can still hear her.

Very disorganised Spanish sessions asides, the person I have become today has been one of the best gifts she could ever have given me.

Thanks, big sis. I owe ya.

And I will be contemplatively playing Final Fantasy 7 Rebirth with these musings in mind.