I want to clarify my recent The Trouble with Facts post. I realise that I was speaking to one non-trivial form of facts, but there is more than one class of facts. We argue about facts as if the word named a single, stable thing. It doesn’t. It names a family of very different things, quietly grouped together by habit, convenience, and institutional need. Most disputes about facts go nowhere, not because one side is irrational, but because the word itself is doing covert work. We slide between meanings without noticing, then act surprised when disagreement follows. This piece is an attempt to slow that slide.
Audio: NotebookLM summary podcast of this topic.
Polysemy We Notice, Polysemy We Don’t
We are comfortable with ambiguity when it is obvious. A bank can be a financial institution or the edge of a river. A bat can be an animal or a piece of sports equipment. Context resolves these instantly. No one feels existentially threatened by the ambiguity.
Fact is different. The word is polysemous in a way that is both subtle and consequential. Its meanings sit close enough to bleed into one another, allowing certainty from one sense to be smuggled into another without detection. Calling something a fact does not merely describe it. It confers authority. It signals that questioning should stop. That is why this ambiguity matters.
Different Kinds of Facts
Before critiquing facts, we need to sort them.
1. Event-facts (brute, world-facing) As mentioned previously, these concern what happens in the world, independent of observation.
A car collides with a tree.
Momentum changes.
Metal deforms.
These events occur whether or not anyone notices them. They are ontologically robust and epistemically inaccessible. No one ever encounters them directly. We only ever encounter traces.
2. Indexical or performative facts (trivial, self-reporting) “I am typing.”
I am doing this now – those now may not be relevant when you read this. This is a fact, but a very thin one. Its authority comes from the coincidence of saying and doing. It requires no reconstruction, no inference, no institutional validation. These facts are easy because they do almost no work.
3. Retrospective personal facts (memory-mediated) “I was typing.”
This may be relevant now, at least relative to the typing of this particular post. Still a fact, but weaker. Memory enters. Narrative compression enters. Selectivity enters. The same activity now carries a different epistemic status purely because time has passed.
4. Prospective statements (modal, not yet facts) “I will be typing.”
This is not yet a fact. It may never come to be one. It is an intention or prediction that may or may not be realised. Future-tense claims are often treated as incipient facts, but this is a category error with real consequences.
5. Institutional facts (designated, procedural) “The court finds…” “The report concludes…”
These are facts by designation. They are not discovered so much as selected, formalised, and stabilised so that systems can act. They are unlikely to rise to the level of facts, so the legal system tends to generate facts in name only – FINO, if I am being cute.
All of these are called ‘facts’. They are not interchangeable. The trouble begins when certainty migrates illicitly from trivial or institutional facts into brute event-facts, and we pretend nothing happened in the transfer.
One Motor Vehicle
Reconsider the deliberately simple case: A motor vehicle collides with a tree. Trees are immobile, so we can rule out the tree colliding with the car.
Ontologically, something happened. Reality did not hesitate. But even here, no one has direct access to the event itself.
The driver does not enjoy privileged access. They experience shock, adrenaline, attentional narrowing, selective memory, post hoc rationalisation, perhaps a concussion. Already several layers intervene before language even arrives.
A rough schema looks like this:
event → sensory registration → cognitive framing → linguistic encoding → social validation
Ontology concerns what happens. Epistemology concerns how anything becomes assertable.
Modern thinking collapses the second into the first and calls the result the facts.
People speak of “hard facts” as if hardness transfers from objects to propositions by proximity. It doesn’t. The tree is solid. The fact is an artefact assembled from observation, inference, convention, and agreement.
And so it goes…
Why the Confusion Persists
When someone responds, “But isn’t it a fact that I read this?”, the answer is yes. A different kind of fact.
The error lies not in affirming facts, but in failing to distinguish them. The word fact allows certainty to migrate across categories unnoticed, from trivial self-reports to brute world-events, and from institutional verdicts to metaphysical claims. That migration is doing the work.
Conclusion
Clarifying types of facts does not weaken truth. It prevents us from laundering certainty where it does not belong.
Facts exist. Events occur. But they do not arrive unmediated, innocent, or singular.
Reality happens once. Facts happen many times.
The mistake was never that facts are unreal. It was believing they were all the same kind of thing.
What we call facts are not discoveries of an unfiltered world. They are the end-products of mediation.
Let’s walk through an example.
Image: Autosmash example. An observer arrives with experience – from genetic predisposition to childhood trauma to winning the lottery. Whatever it might be. Of course, they have many cognitive deficits, biases and filters. Then, there’s the immediate problem of attention. When did they notice the event? Did they turn to look after hearing the noise, or were they meditating on the tree in that moment?
Apparently, a motor vehicle has collided with a tree. Trees are immobile objects, so we can safely rule out the tree colliding with the car.*
So what, exactly, are the facts?
Audio: NotebookLM summary podcast of this topic.
Ontology (the boring bit)
Ontologically, something happened.
A car struck a tree. Metal deformed. Momentum stopped.
Reality did not hesitate. It did not consult witnesses. It did not await interpretation.
This is the part Modernity likes to gesture at reverently before immediately leaving it behind.
Image: Requisite NotebookLM infographic on this content.
The Witness
Even the driver does not enjoy privileged access to “what really happened”.
They get:
proprioceptive shock
adrenaline distortion
attentional narrowing
selective memory
post hoc rationalisation
possibly a concussion
Which is already several layers deep before language even arrives to finish the job.
We can generalise the structure:
event → sensory registration → cognitive framing → linguistic encoding → social validation
Ontology: events occur. States of affairs obtain. Something happens whether or not we notice.
Epistemology: observation is always filtered through instruments, concepts, language, habits, and incentives.
Modern sleight of hand: collapse the second into the first and call the result the facts.
People love the phrase “hard facts”, as if hardness transfers from objects to propositions by osmosis. It doesn’t. The tree is solid. The fact is not.
Facts are artefacts. They are assembled from observation, inference, convention, and agreement. They function. They do not reveal essence.
Reality happens once. Facts happen many times, differently, depending on who needs them and why.
Filtration
An event occurred. A car struck a tree.
Then an observer arrives. But observers never arrive empty-handed.
They arrive with history: genetics, upbringing, trauma, habits, expectations, incentives. They arrive already filtered.
Were they already looking, or did the noise interrupt something else entirely?
Reality happens once. Facts happen many times, differently, depending on who needs them and why.
Here Comes the Law
This is where the legal system enters, not because truth has been found, but because closure is required.
Courts do not discover facts. They designate versions of events that are good enough to carry consequences. They halt the cascade of interpretations by institutional force and call the result justice.
At every epistemic level, what we assert are interpretations of fact, never access to ontological essence.
Intent, negligence, recklessness. These are not observations. They are attributions. They are stopping rules that allow systems to function despite uncertainty.
The law does not ask what really happened. It asks which story is actionable.
Two Motor Vehicles
Now add a second moving object.
Another car enters the frame, and with it an entire moral universe.
Suddenly, the event is no longer merely physical. It becomes relational. Agency proliferates. Narratives metastasise.
Who was speeding? Who had the right of way? Who saw whom first? Who should have anticipated whom?
Intent and motive rush in to fill the explanatory vacuum, despite remaining just as unobservable as before.
Nothing about the ontology improved. Everything about the storytelling did.
Where the tree refused intention, the second vehicle invites it. We begin inferring states of mind from trajectories, attributing beliefs from brake lights, extracting motives from milliseconds of motion.
But none of this is observed.
What we observe are:
vehicle positions after the fact,
damage patterns,
skid marks,
witness statements already filtered through shock and expectation.
From these traces, we construct mental interiors.
The driver “intended” to turn. The other driver “failed” to anticipate. Someone was “reckless”. Someone else was merely “unlucky”.
These are not facts. They are interpretive assignments, layered atop already mediated observations, selected because they allow responsibility to be distributed in socially recognisable ways.
This is why explanation now fractures.
One cascade of whys produces a story about distraction or poor judgment. Another produces a story about road design or visibility. Another about timing, traffic flow, or urban planning.
Each narrative is plausible. Each is evidence-constrained. None is ontologically privileged.
Yet one will be chosen.
Not because it is truer, but because it is actionable.
The presence of a second vehicle does not clarify causation. It merely increases the number of places we are willing to stop asking questions.
Modernity mistakes this proliferation of narrative for epistemic progress. In reality, it is moral bookkeeping.
The crash still occurred. Metal still deformed. Momentum still stopped.
What changed was not access to truth, but the urgency to assign fault.
With one vehicle and a tree, facts already fail to arrive unmediated. With two vehicles, mediation becomes the point.
And still, we insist on calling the result the facts.
But nothing fundamental has changed. We have not escaped mediation. We have merely scaled it up and professionalised it.
Cameras do not record reality. They record:
a frame,
from a position,
at a sampling rate,
with compression,
under lighting conditions,
interpreted later by someone with a mandate.
Video feels decisive because it is vivid, not because it is ontologically transparent. It freezes perspective and mistakes that freeze for truth. Slow motion, zoom, annotation. Each step adds clarity and distance at the same time.
Experts do not access essence either. They perform disciplined abduction.
From angles, debris fields, timing estimates, and damage profiles, they infer plausible sequences. They do not recover the event. They model it. Their authority lies not in proximity to reality, but in institutional trust and methodological constraint.
More data does not collapse interpretation. It multiplies it.
With enough footage, we don’t get the story. We get competing reconstructions, each internally coherent, each technically defensible, each aligned to a different question:
Who is legally liable?
Who is financially responsible?
Who violated policy?
Who can be blamed without destabilising the system?
At some point, someone declares the evidence “clear”.
What they mean is: we have enough material to stop arguing.
This is the final Modern illusion: that accumulation converges on essence. In reality, accumulation converges on closure.
The event remains what it always was: inaccessible except through traces. The facts become thicker, more confident, more footnoted. Their metaphysical status does not improve.
Reality happened once. It left debris. We organised the debris into narratives that could survive institutions.
Cameras didn’t reveal the truth. Experts didn’t extract it. They helped us agree on which interpretation would count.
And agreement, however necessary, has never been the same thing as access to what is.
* I was once driving in a storm, and a telephone pole fell about a metre in front of my vehicle. My car drove over the pole, and although I was able to drive the remainder of the way home, my suspension and undercarriage were worse for the wear and tear.
Written by Bry Willis and ChatGPT 5.2 after an extended chat about the image as a talking point about the inaccessibility of facts.
I’ve been reading Bernard Williams lately, and I’ve written about his work on Truth and Truthfulness. I’m in the process of writing more on the challenges of ontological moral positionsand moral luck. I don’t necessarily want to make contemporary news my focal point, but this is a perfect case study for it. I’ll be releasing a neutral philosophy paper on the underlying causes, but I want to comment on this whilst it’s still in the news cycle.
The form of xenophobia is a phenomenon occurring in the United States, though the ontological split is applicable more generally. For those unfamiliar with US news, I’ll set this up. The United States is currently deploying federal enforcement power in ways that deliberately bypass local consent, blur policing and military roles, and rely on fear as a stabilising mechanism. Historical analogies are unavoidable, but not required for the argument that follows. These forces have been deployed in cities that did not and do not support the Trump administration, so they are exacting revenge and trying to foment fear and unrest. This case is an inevitable conclusion to these policy measures.
tl;dr: The Law™ presents itself as fact-driven, but only by treating metaphysical imputations about inner life as if they were empirical findings. This is not a flaw in this case; it is how the system functions at all.
NB: Some of this requires having read Williams or having a familiarity with certain concepts. Apologies in advance, but use Google or a GPT to fill in the details.
Audio: NotebookLM summary podcast of this content.
Why the Minneapolis ICE Shooting Exposes the Limits of Bernard Williams
The Minneapolis ICE shooting is not interesting because it is unusual. It is interesting because it is painfully ordinary. A person is dead. An officer fired shots. A vehicle was involved. Video exists. Statements were issued. Protests followed. No one seriously disputes these elements. They sit in the shared centre of the Venn diagram, inert and unhelpful. Where everything fractures is precisely where the law insists clarity must be found: intent and motive. And this is where things stop being factual and start being metaphysical.
The Comfortable Fiction of Legal Facts
The legal system likes to tell a comforting story about itself. It claims to be empirical, sober, and evidence-driven. Facts in, verdicts out. This is nonsense.
What the law actually does is this:
It gathers uncontested physical facts.
It then demands a psychological supplement.
It treats that supplement as if it were itself a fact.
Intent and motive are not observed. They are inferred. Worse, they are imposed. They are not discovered in the world but assigned to agents to make outcomes legible.
In Minneapolis, the uncontested facts are thin but stable:
A U.S. Immigration and Customs Enforcement (ICE) agent, identified as Jonathan Ross, shot and killed Renée Nicole Good in Minneapolis on 7 January 2026.
The incident involved Good’s vehicle, which was present and moving at the time shots were fired.
Ross fired his weapon multiple times, and Good died from those gunshot wounds.
The Department of Homeland Security (DHS) claims the agent acted in self-defence.
Video footage exists that shows at least part of the encounter.
The case ignited protests, widespread condemnation from local officials, and political pushback.
This creates a shared intersection: vehicle, Ross, shots, and that ‘something happened’ that neither side is denying.
The Law smuggles metaphysics into evidence and calls it psychology.
None of these facts contain intent. None of them specify motive. They do not tell us whether the movement of the vehicle was aggression, panic, confusion, or escape. They do not tell us whether the shooting was fear, anger, habit, or protocol execution. Yet the law cannot proceed without choosing. So it does what it always does. It smuggles metaphysics into evidence and calls it psychology.
Intent and Motive as Institutional Impositions
Intent is treated as a condition of responsibility. Motive is treated as its explanation. Neither is a fact in anything like the ordinary sense. Even self-report does not rescue them. Admission is strategically irrational. Silence is rewarded. Reframing is incentivised. And even sincerity would not help, because human beings do not have transparent access to their own causal architecture. They have narratives, rehearsed and revised after the fact. So the law imputes. It tells the story the agent cannot safely tell, and then punishes or absolves them on the basis of that story. This is not a bug. It is the operating system.
Where Bernard Williams Comes In
This is where Bernard Williams becomes relevant, and where his account quietly fails. In Truth and Truthfulness, Williams famously rejects the Enlightenment fantasy of capital-T Truth as a clean, context-free moral anchor. He replaces it with virtues like sincerity and accuracy, grounded in lived practices rather than metaphysical absolutes. So far, so good.
Williams is right that moral life does not float above history, psychology, or culture. He is right to attack moral systems that pretend agents consult universal rules before acting. He is right to emphasise thick concepts, situated reasons, and practical identities. But he leaves something standing that cannot survive the Minneapolis test.
The Residue Williams Keeps
Williams still needs agency to be intelligible. He still needs actions to be recognisably owned. He still assumes that reasons, however messy, are at least retrospectively available to anchor responsibility. This is where the residue collapses.
In cases like Minneapolis:
Intent is legally required but epistemically unavailable.
Motive is legally explanatory but metaphysically speculative.
Admission is disincentivised.
Narrative is imposed under institutional pressure.
At that point, sincerity and accuracy are no longer virtues an agent can meaningfully exercise. They are properties of the story selected by the system. Williams rejects metaphysical Truth while retaining a metaphysical agent robust enough to carry responsibility. The problem is that law does not merely appeal to intelligibility; it manufactures it under constraint.
Moral Luck Isn’t Enough
Williams’ concept of moral luck gestures toward contingency, but it still presumes a stable agent who could, in principle, have acted otherwise and whose reasons are meaningfully theirs. But once intent and motive are understood as institutional fabrications rather than inner facts, ‘could have done otherwise’ becomes a ceremonial phrase. Responsibility is no longer uncovered; it is allocated. The tragedy is not that we fail to know the truth. The tragedy is that the system requires a truth that cannot exist.
Facts Versus Stories
The law does not discover which story is true. It selects which story is actionable.
The Minneapolis case shows the fault line clearly:
Facts: bodies, movements, weapons, recordings.
Stories: fear versus anger, defence versus aggression.
The first is uncontested. The second does all the work.
And those stories are not epistemic conclusions. They are metaphysical commitments enforced by law. Williams wanted to rescue ethics from abstraction. What he could not accept is that, once abstraction is removed, responsibility does not become more human. It becomes procedural.
The Uncomfortable Conclusion
The law does not operate on truth. It operates on enforceable interpretations of behaviour. Intent and motive are not facts. They are tools. Williams saw that capital-T Truth had to go. What he did not see, or perhaps did not want to see, is that the smaller, more humane residue he preserved cannot bear the weight the legal system places on it.
Once you see this, the obsession with ‘what really happened’ looks almost childish. The facts are already known. What is being fought over is which metaphysical fiction the system will enforce.
That decision is not epistemic. It is political. And it is violent.
Fairness, Commensurability, and the Quiet Violence of Comparison
Fairness and Commensurability as Preconditions of Retributive Justice
This is the final part of a 3-part series. Read parts 1 and 2 for a fuller context.
Audio: NotebookLM summary podcast of this topic.
Before the Cards Are Dealt
Two people invoke fairness. They mean opposite things. Both are sincere. Neither can prove the other wrong. This is not a failure of argument. It is fairness working exactly as designed.
Before justice can weigh anything, it must first decide that the things being weighed belong on the same scale. That single move – the assertion that comparison is even possible – quietly does most of the work.
Most people think justice begins at sentencing, or evidence, or procedure. But the real work happens earlier, in a space so normalised it has become invisible. Before any evaluation occurs, the system must install the infrastructure that makes evaluation legible at all.
That infrastructure rests on two foundations:
fairness, which supplies the rhetoric, and
commensurability, which supplies the mathematics.
Together, they form the felt beneath the table – the surface on which the cards can be dealt at all.
1. Why Fairness Is Always Claimed, Never Found
Let’s be precise about what fairness is not.
Fairness is not a metric. You cannot measure it, derive it, or point to it in the world.
Fairness is not a principle with determinate content. It generates no specific obligations, no falsifiable predictions, no uniquely correct outcomes.
Fairness is an effect. It appears after assessment, not before it. It is what you call an outcome when you want it to feel inevitable.
Competing Fairness Is Not a Problem
Consider how disputes actually unfold:
The prosecutor says a long sentence is fair because it is proportional to harm.
The defender says a shorter sentence is fair because it reflects culpability and circumstance.
The victim says any sentence is unfair because nothing restores what was taken.
The community says enforcement itself is unfair because it predictably targets certain groups.
Each claim is sincere. None can be resolved by fairness itself.
That is because fairness has no independent content. It does not decide between these positions. It names them once the system has already decided which will prevail. This is not a bug. It is the feature.
A Fluid Masquerading as an Invariant
In the language of the Language Insufficiency Hypothesis, fairness is a Fluid – a concept whose boundaries shift with context and use – that masquerades as an Invariant, something stable and observer-independent.
The system treats fairness as perceptual, obvious, discoverable. But every attempt to anchor it collapses into:
Intuition (‘It just feels right’)
Precedent (‘This is how we do things’)
Consensus (‘Most people agree’)
None of these establishes fairness. They merely perform it.
And that performance matters. It converts contested metaphysical commitments into the appearance of shared values. It allows institutions to claim neutrality whilst enforcing specificity. Fairness is what the system says when it wants its outputs to feel unavoidable.
2. The Real Gatekeeper: Commensurability
Fairness does rhetorical work. But it cannot function without something deeper.
That something is commensurability: the assumption that different harms, injuries, and values can be placed on a shared scale and meaningfully compared.
Proportionality presupposes commensurability. Commensurability presupposes an ontology of value. And that ontology is neither neutral nor shared.
When Incommensurability Refuses to Cooperate
A parent loses a child to preventable negligence. A corporation cuts safety corners. A warning is ignored. The system moves. Liability is established. Damages are calculated. £250,000 is awarded.
The parent refuses the settlement. Not because the amount is insufficient. But because money and loss are not the same kind of thing. The judge grows impatient. Lawyers speak of closure. Observers mutter about grief clouding judgment. But this is not grief. It is incommensurability refusing to cooperate.
The parent is rejecting the comparison itself. Accepting payment would validate the idea that a child’s life belongs on a scale with currency. The violence is not the number. It is the conversion. The system cannot process this refusal except as emotional excess or procedural obstruction. Not because it is cruel, but because without commensurability the engine cannot calculate.
Two Ontologies of Value
There are two incompatible ontologies at work here. Only one is playable.
Ontology A: The Scalar Model
Harm is quantifiable
Suffering is comparable
Trade-offs are morally coherent
Justice is a balancing operation
Under Ontology A, harms differ in degree, not kind. A broken arm, a stolen car, and a dead child all occupy points on the same continuum. This makes proportionality possible.
Ontology B: The Qualitative Model
Harms are categorical
Some losses are incommensurable
Comparison itself distorts
Justice is interpretive, not calculative
Under Ontology B, harms are different kinds of things. Comparison flattens what matters. To weigh them is to misunderstand them.
Why Only One Ontology Can Play
Retributive justice, as presently constituted, cannot function under Ontology B.
Without scalar values, proportionality collapses. Without comparison, equivalence disappears. Without trade-offs, punishment has no exchange rate.
Ontology B is not defeated. It is disqualified. Structurally, procedurally, rhetorically. The house needs a shared scale. Without it, the game cannot settle accounts.
3. Why Incommensurability Is Treated as Bad Faith
Here is where power enters without announcing itself. Incommensurability does not merely complicate disputes. It stalls the engine. And stalled engines threaten institutional legitimacy.
Systems designed to produce closure must ensure that disputes remain within solvable bounds. Incommensurability violates those bounds. It suggests that resolution may be impossible – or that the attempt to resolve does further harm. So the system reframes the problem.
Not as an alternative ontology, but as:
Unreasonableness
Extremism
Emotional volatility
Refusal to engage in good faith
Reasonableness as Border Control
This is why reasonableness belongs where it does in the model. Not as an evaluative principle, but as a gatekeeping mechanism.
Reasonableness does not assess claims. It determines which claims count as claims at all. This is how commensurability enforces itself without admitting it is doing so. When someone refuses comparison, they are not told their ontology is incompatible with retributive justice. They are told to be realistic.
Ontological disagreement is converted into:
A tone problem
A personality defect
A failure to cooperate
The disagreement is not answered. It is pathologised.
4. Why These Debates Never Resolve
This returns us to the Ontology–Encounter–Evaluation model.
People argue fairness as if adjusting weights would fix the scale. They debate severity, leniency, proportionality.
But when two sides inhabit incompatible ontologies of value, no amount of evidence or dialogue bridges the gap. The real disagreement is upstream.
A prosecutor operating under scalar harm and an advocate operating under incommensurable injury are not disagreeing about facts. They are disagreeing about what kind of thing harm is.
Fairness cannot resolve this, because fairness presupposes the very comparison under dispute. This is why reform debates feel sincere and go nowhere. Outcomes are argued whilst ontological commitments remain invisible.
Remediation Requires Switching Teams
As argued elsewhere, remediation increasingly requires switching teams.
But these are not political teams. They are ontological commitments.
Ontologies are not held like opinions. They are held like grammar. You do not argue someone out of them. At best, you expose their costs. At worst, you force others to operate within yours by disqualifying alternatives.
Retributive justice does the latter.
5. What This Means (Without Offering a Fix)
Justice systems are not broken. They are optimised. They are optimised for closure, manageability, and the appearance of neutrality. Fairness supplies the rhetoric. Commensurability supplies the mathematics. Together, they convert contestable metaphysical wagers into procedural common sense.
That optimisation has costs:
Disagreements about value become illegible
Alternative ontologies become unplayable
Dissent becomes pathology
Foundations disappear from view
If justice feels fair, it is because the comparisons required to question it were never permitted.
Ontology as Pre-emptive Gatekeeping
None of this requires conspiracy.
Institutions do not consciously enforce ontologies. They do not need to.
They educate them. Normalise them. Proceduralise them. Then treat their rejection as irrationality.
By the time justice is invoked, the following have already been installed as reality:
That persons persist over time in morally relevant ways
That agents meaningfully choose under conditions that count
That harms can be compared and offset
That responsibility can be localised
That disagreement beyond a point is unreasonable
None of these are discovered. All are rehearsed.
A law student learns that ‘the reasonable person’ is a construct. By year three, they use it fluently. It no longer feels constructed.
This is not indoctrination. It is fluency.
And fluency is how ontologies hide.
By the time an alternative appears – episodic selfhood, incommensurable harm, distributed agency – it does not look like metaphysics. It looks like confusion.
Rationality as Border Control
The system does not say: we reject your ontology.
It says: that’s not how the world works.
Or worse: you’re being unreasonable.
Ontological disagreement is reframed as a defect in the person. And defects do not need answers. They need management.
This is why some arguments feel impossible to have. One ontology has been naturalised into common sense. The other has been reclassified as error.
The Final Irony
The more fragile the foundations, the more aggressively they must be defended as self-evident.
Free will is taught as obvious.
Fairness is invoked as perceptual.
Responsibility is treated as observable.
Incommensurability is treated as sabotage.
Not because the system is confident.
Because it cannot afford not to be.
The Point
Justice does not merely rely on asserted ontologies. It expends enormous effort ensuring they never appear asserted at all.
By the time the cards are dealt, the rules have already been mistaken for reality. That is the felt beneath the table. Invisible. Essential. Doing all the work. And if you want to challenge justice meaningfully, you do not start with outcomes. You start by asking:
What comparisons are we being asked to accept as natural? And what happens to those who refuse?
Most people never make that move. Not because it is wrong. But because by the time you notice the game is rigged, you are already fluent in its rules. And fluency feels like truth.
Final Word
Why write these assessments? Why care?
With casinos, like cricket, we understand something fundamental: these are games. We can learn the rules. We can decide whether to play. We can walk away.
Justice is different. Justice is not opt-in. It is imposed. You do not get to negotiate the rules, the scoring system, or the house assumptions about what counts as a move. Once you are inside, even dissent must be expressed in the system’s own grammar. Appeals do not question the game; they replay it under slightly altered conditions.
You may contest the outcome. You may plead for leniency. You may argue fairness. You may not ask why chips are interchangeable with lives, why losses must be comparable, or why refusing comparison itself counts as misconduct.
Imagine being forced into a casino. Forced to play. Forced to stake things you do not believe are wagerable. Then told, when you object, that the problem is not the game, but your attitude toward it.
That is why these assessments matter. Not to declare justice illegitimate. Not to offer a fix. But to make visible the rules that pretend not to be rules at all. Because once you mistake fluency for truth, the house no longer needs to rig the game.
If you want a useful metaphor for how justice actually operates, don’t picture a blindfolded goddess with scales. Picture a casino.
Image: Lady Justice in Casino. The dice are rigged. haha
The rules are printed. The games look fair. Everyone is technically allowed to play. But the mathematics are tuned in advance, the exits are discreet, and the house never risks its own solvency. You don’t walk into a casino to discover whether chance is fair. You walk in to participate in a system whose advantage has already been engineered.
Justice works the same way – like a casino with house-favoured odds.
By the time a defendant appears, the ontological dice have already been loaded. The system has quietly asserted a set of metaphysical commitments that make certain outcomes legible, actionable, and punishable – whilst rendering others incoherent, inadmissible, or ‘unreasonable’. Because I am a philosopher of language and not a lawyer, I am free from the indoctrination and selection bias inherent in that system. This allows me to critique the system directly without being excommunicated from the club.
What follows are not neutral assumptions. They are ontological wagers, each chosen because its alternative would tilt the field away from institutional power.
Justice presumes that the person who acted yesterday is meaningfully the same entity standing in court today. This is not discovered; it is asserted.
Why? Because retribution requires persistence. Desert cannot attach to a momentary configuration of consciousness. Responsibility requires a carrier that survives time, memory gaps, psychological rupture, intoxication, trauma, and neurological variance.
An episodic self – Parfit’s reductionism, trauma-fractured identity, or situational selfhood – collapses the attribution pipeline. If the ‘self’ is a series of loosely connected episodes, punishment becomes conceptually incoherent. Who is being punished for whom?
So the law treats episodic accounts not as alternative ontologies but as defects: insanity, automatism, incompetence. The self is patched, not replaced.
The house needs continuity. Without it, the game cannot settle accounts.
Justice requires that actions originate somewhere. Agency is that somewhere.
The system asserts that agents could have done otherwise in a morally relevant sense. This is compatible with compatibilism, folk psychology, and everyday moral intuitions – but deeply hostile to hard determinism, strong situationism, or neurobiological deflation.
Why exclude weaker agency models? Because if agency dissolves into causation, environment, or neurochemistry, responsibility evaporates. At best, you get risk management. At worst, you get treatment or containment. Retribution has nowhere to land.
So the law nods politely to influences – upbringing, coercion, impairment – whilst ring-fencing agency as the default. Mitigation is permitted. Ontological revision is not. The house needs someone who could have chosen otherwise, even if that claim grows increasingly fictional under scrutiny.
The house needs someone who could have chosen otherwise
Justice models human action as a series of forks in the road. At some point, the agent ‘chose’ X over Y. This is enormously convenient.
Continuous decision spaces – poverty gradients, addiction loops, survival trade-offs – are messy. They resist clean counterfactuals. ‘What should they have done instead?’ becomes a sociological question, not a moral one.
So the system discretises. It locates a moment. A click. A trigger pull. A signature. A punch. A text sent.
Once the choice is frozen, the rest of the apparatus can proceed. Without discrete choice points, proportionality and culpability lose their anchor.
The house needs moments, not milieus.
Ontology 4: Causation
Asserted (m): Local, linear causation Excluded (n): Diffuse, systemic, or emergent causation
Justice prefers causes that point: Who did this? When? How directly?
Systemic causation – economic pressure, cultural narratives, institutional design – creates attribution problems. If harm is emergent, no individual carries it cleanly. Responsibility smears.
So causation is narrowed. Chains are shortened. Proximate cause replaces contributing conditions. Structural violence becomes background noise.
This is not because systemic causation is false. It is because it is unmanageable within a retributive frame.
The house needs causes that fit inside a sentence.
‘Reasonableness’ is the softest and most insidious ontology of the lot.
It pretends to be procedural, but it functions as cultural enforcement. The reasonable person is not an average human. They are an acculturated one.
Intensity becomes suspect. Rage becomes irrational. Grief becomes excessive. Radical interpretations become unreasonable not because they’re false, but because they disrupt cadence.
This ontology stabilises the game by disciplining tone. It doesn’t matter what you argue if you fail to argue it reasonably. Reasonableness is not required for responsibility to exist, only for dissent to be ignored.
Reasonableness is not required for responsibility to exist, only for dissent to be ignored.
The house needs calm players, not correct ones.
Why These Ontologies, and Not Their Rivals?
Because every excluded ontology threatens legibility. Justice is not designed to discover truth. It is designed to terminate cases. Ontologies that complicate attribution, disperse responsibility, or destabilise narrative continuity slow the machine. So they are ruled out – not explicitly, but structurally.
Once these commitments are in place, disagreement downstream becomes theatre. Arguments about fairness, proportionality, or intent occur within a rigged metaphysical envelope. That’s why reform debates feel sincere yet go nowhere. People argue outcomes whilst the house quietly keeps the rules.
The Point
None of this means justice is a scam. Casinos aren’t scams either. They do exactly what they are designed to do.
The problem is pretending the odds are neutral.
If you want to challenge justice meaningfully, you don’t start with sentencing guidelines or evidentiary thresholds. You start by asking which ontologies are being asserted – and why alternatives are unplayable.
Most people won’t make that move. Not because it’s wrong. Because it requires leaving the table.
Now that A Language Insufficiency Hypothesis has been put to bed — not euthanised, just sedated — I can turn to the more interesting work: instantiating it. This is where LIH stops being a complaint about words and starts becoming a problem for systems that pretend words are stable enough to carry moral weight.
What follows is not a completed theory, nor a universal schema. It’s a thinking tool. A talking point. A diagram designed to make certain assumptions visible that are usually smuggled in unnoticed, waved through on the strength of confidence and tradition.
The purpose of this diagram is not to redefine justice, rescue it, or replace it with something kinder. It is to show how justice is produced. Specifically, how retributive justice emerges from a layered assessment process that quietly asserts ontologies, filters encounters, applies normative frames, and then closes uncertainty with confidence.
Audio: NotebookLM summary podcast of this topic.
Most people are willing to accept, in the abstract, that justice is “constructed”. That concession is easy. What is less comfortable is seeing how it is constructed — how many presuppositions must already be in place before anything recognisable as justice can appear, and how many of those presuppositions are imposed rather than argued for.
The diagram foregrounds power, not as a conspiracy or an optional contaminant, but as an ambient condition. Power determines which ontologies are admissible, which forms of agency count, which selves persist over time, which harms are legible, and which comparisons are allowed. It decides which metaphysical configurations are treated as reasonable, and which are dismissed as incoherent before the discussion even begins.
Justice, in this framing, is not discovered. It is not unearthed like a moral fossil. It is assembled. And it is assembled late in the process, after ontology has been assumed, evaluation has been performed, and uncertainty has been forcibly closed.
This does not mean justice is fake. It means it is fragile. Far more fragile than its rhetoric suggests. And once you see that fragility — once you see how much is doing quiet, exogenous work — it becomes harder to pretend that disagreements about justice are merely disagreements about facts, evidence, or bad actors. More often, they are disagreements about what kind of world must already be true for justice to function at all.
I walk through the structure and logic of the model below. The diagram is also available as a PDF, because if you’re going to stare at machinery, you might as well be able to zoom in on the gears.
Why Retributive Justice (and not the rest of the zoo)
Before doing anything else, we need to narrow the target.
“Justice” is an infamously polysemous term. Retributive, restorative, distributive, procedural, transformative, poetic, cosmic. Pick your flavour. Philosophy departments have been dining out on this buffet for centuries, and nothing useful has come of letting all of them talk at once.
This is precisely where LIH draws a line.
The Language Insufficiency Hypothesis is not interested in pedestrian polysemy — cases where a word has multiple, well-understood meanings that can be disambiguated with minimal friction. That kind of ambiguity is boring. It’s linguistic weather.
What LIH is interested in are terms that appear singular while smuggling incompatible structures. Words that function as load-bearing beams across systems, while quietly changing shape depending on who is speaking and which assumptions are already in play.
“Justice” is one of those words. But it is not usefully analysable in the abstract.
So we pick a single instantiation: Retributive Justice.
Why?
Because retributive justice is the most ontologically demanding and the most culturally entrenched. It requires:
a persistent self
a coherent agent
genuine choice
intelligible intent
attributable causation
commensurable harm
proportional response
In short, it requires everything to line up.
If justice is going to break anywhere, it will break here.
Retributive justice is therefore not privileged in this model. It is used as a stress test.
The Big Picture: Justice as an Engine, Not a Discovery
The central claim of the model is simple, and predictably unpopular:
Justice is not discovered. It is produced.
Not invented in a vacuum, not hallucinated, not arbitrary — but assembled through a process that takes inputs, applies constraints, and outputs conclusions with an air of inevitability.
The diagram frames retributive justice as an assessment engine.
An engine has:
inputs
internal mechanisms
thresholds
failure modes
and outputs
It does not have access to metaphysical truth. It has access to what it has been designed to process.
The justice engine takes an encounter — typically an action involving alleged harm — and produces two outputs:
Desert (what is deserved),
Responsibility (to whom it is assigned).
Everything else in the diagram exists to make those outputs possible.
The Three Functional Layers
The model is organised into three layers. These are not chronological stages, but logical dependencies. Each layer must already be functioning for the next to make sense.
1. The Constitutive Layer
(What kind of thing a person must already be)
This layer answers questions that are almost never asked explicitly, because asking them destabilises the entire process.
What counts as a person?
What kind of self persists over time?
What qualifies as an agent?
What does it mean to have agency?
What is a choice?
What is intent?
Crucially, these are not empirical discoveries made during assessment. They are asserted ontologies.
The system assumes a particular configuration of selfhood, agency, and intent as a prerequisite for proceeding at all. Alternatives — episodic selves, radically distributed agency, non-volitional action — are not debated. They are excluded.
This is the first “happy path”.
If you do not fit the assumed ontology, you do not get justice. You get sidelined into mitigation, exception, pathology, or incoherence.
2. The Encounter Layer
(What is taken to have happened)
This layer processes the event itself:
an action
resulting harm
causal contribution
temporal framing
contextual conditions
motive (selectively)
This is where the rhetoric of “facts” tends to dominate. But the encounter is never raw. It is already shaped by what the system is capable of seeing.
Causation here is not metaphysical causation. It is legible causation. Harm is not suffering. It is recognisable harm. Context is not total circumstance. It is admissible context.
Commensurability acts as a gatekeeper between encounter and evaluation: harms must be made comparable before they can be judged. Anything that resists comparison quietly drops out of the pipeline.
3. The Evaluative Layer
(How judgment is performed)
Only once ontology is assumed and the encounter has been rendered legible does evaluation begin:
proportionality
accountability
normative ethics
fairness (claimed)
reasonableness
bias (usually acknowledged last, if at all)
This layer presents itself as the moral heart of justice. In practice, it is the final formatting pass.
Fairness is not discovered here. It is declared. Reasonableness does not clarify disputes. It narrows the range of acceptable disagreement. Bias is not eliminated. It is managed.
At the end of this process, uncertainty is closed.
That closure is the moment justice appears.
Why Disagreement Fails Before It Starts
At this point, dissent looks irrational.
The system has:
assumed an ontology
performed an evaluation
stabilised the narrative through rhetoric
and produced outputs with institutional authority
To object now is not to disagree about evidence. It is to challenge the ontology that made assessment possible in the first place.
And that is why so many justice debates feel irresolvable.
They are not disagreements within the system. They are disagreements about which system is being run.
LIH explains why language fails here. The same words — justice, fairness, responsibility, intent — are being used across incompatible ontological commitments. The vocabulary overlaps; the worlds do not.
The engine runs smoothly. It just doesn’t run the same engine for everyone.
Where This Is Going
With the structure in place, we can now do the slower work:
unpacking individual components
tracing where ontological choices are asserted rather than argued
showing how “reasonableness” and “fairness” operate as constraint mechanisms
and explaining why remediation almost always requires a metaphysical switch, not better rhetoric
Justice is not broken. It is doing exactly what it was built to do.
That should worry us more than if it were merely malfunctioning.
This essay is already long, so I’m going to stop here.
Not because the interesting parts are finished, but because this is the point at which the analysis stops being descriptive and starts becoming destabilising.
The diagram you’ve just walked through carries a set of suppressed footnotes. They don’t sit at the margins because they’re trivial; they sit there because they are structurally prior. Each one represents an ontological assertion the system quietly requires in order to function at all.
By my count, the model imposes at least five such ontologies. They are not argued for inside the system. They are assumed. They arrive pre-installed, largely because they are indoctrinated, acculturated, and reinforced long before anyone encounters a courtroom, a jury, or a moral dilemma.
Once those ontologies are fixed, the rest of the machinery behaves exactly as designed. Disagreement downstream is permitted; disagreement upstream is not.
In a follow-up essay, I’ll unpack those footnotes one by one: where the forks are, which branch the system selects, and why the alternatives—while often coherent—are rendered unintelligible, irresponsible, or simply “unreasonable” once the engine is in motion.
That’s where justice stops looking inevitable and starts looking parochial.
And that’s also where persuasion quietly gives up.
Written by Bry Willis and ChatGPT 5.2 after a couple of days of back and forth
More precisely, I need less sleep and longer days – preferably twice as long. I’ve been writing almost non-stop for the better part of a week: fourteen- to sixteen-hour days, fuelled by irritation and the stubborn belief that if I just keep reading, something will finally click into place.
I’m not complaining. This is a virtuous cycle. Reading leads to writing. Writing demands more reading. Eventually, the loop closes into something that looks suspiciously like progress.
Audio: Short NotebookLM summary podcast on this topic.
Still, there’s a bottleneck.
Because some of this work – the work I’m most excited about – I’m deliberately not publishing yet. Journals, bless their glacial hearts, don’t much care for prior publication. So ideas sit in limbo for six to eighteen months, locked in a room like argumentative houseplants, slowly growing sideways.
From the perspective of someone who thinks in public, this is maddening.
Now add AI to the mix.
This is where things get dangerous.
I’ll feed ChatGPT a thesis, a skeletal structure, notes, and references. I ask what I’m missing. It obliges – often helpfully – by pointing me toward adjacent thinkers and relevant literature, complete with page numbers. From there, I verify, hunt down the sources, skim, read, discard, or integrate.
And every so often, I stumble across something that makes me swear out loud.
This week, it was Bernard Williams.
I’ve cited Ethics and the Limits of Philosophy before. But this time, I actually sat down and read it properly. Which immediately prompted the thought:
Why didn’t I read this sooner?
Williams dismantles moral objectivity with the calm precision of someone who knows the Enlightenment project has already lost – he just hasn’t told everyone yet. Thick and thin moral concepts, locality, non-extensibility, the collapse of universal moral reason at scale – yes, yes, yes. He published this in 1985. Fine. I’ll survive.
But then I went further.
Williams shows that morality fails between people at scale. I argue that it fails within a single person over time.
That became my second paper.
And this is where things went off the rails.
Because in the course of writing that paper, I dipped into Hart’s The Concept of Law and Endicott’s Vagueness in Law. These are not fringe polemics. These are law textbooks. For law students. People allegedly trained to parse language for a living.
And what I found was… astonishing.
Let me paraphrase the admissions:
First:
Image: When the law is vague, judicial decisions may be unconstrained by the law.
Endicott: “By upsetting the standard view of adjudication, the book reaches conclusions that some people find horrible: when the law is vague, judicial decision- making will in some cases be unconstrained by the law. It is impossible in principle for judges always to treat like cases alike. Predictability in the law is to some extent unattainable. Moreover, I argue in Chapter 9,2 that vagueness cannot be eliminated from law. These conclusions might seem to imply that the rule of law is, at least to some extent, conceptually impossible.”
Then:
Image: Vagueness is inevitable. Deal with it.
Endicott: “Secondly, I do not claim that vagueness is a purely linguistic feature of law. And the book relies on no claim about the relation between law and language. These points must be stressed, because vagueness is commonly thought of as a linguistic phenomenon. And. indeed, most of the discussion in the book concerns the vagueness of linguistic expressions. But the indeterminacy claim is not just a claim about language (so I argue in Chapter 3.12). So. for example, the claim in Chapter 6 that general evaluative and normative expressions are necessarily vague is not just a claim about the word ‘good’ and the word ‘right1: it is a claim about any linguistic expression in which we could conceivably express general evaluative and normative judgments. It therefore includes a claim about what is good and what is right.”
Then, almost casually:
Image: Whether law is morally valuable to a community is not my concern. Justice and the rule of law may be political virtues — or not. I don’t defend them here.
Endicott: “Disputes between legal positivists and natural law theorists have concerned not only the relation between law and adjudication, but also the relation between law and morality. Here I take no general position on the intrinsic moral value of law. I do rely on the claims that law can be valuable to a community, and that justice and the rule of law are two ideals which a com- munity can intelligibly pursue as political virtues. Even those claims are controversial (Kelsen and some of the theorists discussed in Chapter 2 have controverted them ). But I do not defend them here. This work aims to show that the indeterminacy claim does nothing to threaten the pursuit of justice and the rule of law. Those ideals cannot be well understood if we try to make them depend on determinacy in the requirements of the law.”
Say what?
Read together – not even uncharitably – the message is clear:
Law is indeterminate. Indeterminacy is unavoidable. And whether law is good, just, or valuable is… optional.
The subtext isn’t even hiding.
Law is a power structure first. If it happens to align with justice, fairness, or communal value, well, lovely. A bonus. Champagne all round.
This does not sit well with a sceptical cynic.
What really broke me, though, wasn’t the argument itself. Philosophers make grim claims all the time. What broke me was the silence around it.
How does this pass under the radar?
How do cohorts of law students – drilled in textual analysis, trained to read footnotes like tea leaves – not trip over this elephant stampede? How do they graduate believing they’re upholding inalienable rights, rather than participating in a managed system of coercion that occasionally behaves itself?
Self-preservation, I suppose. Wilful ignorance. Professional cosplay.
I’ve seen this before.
As an economist, ask the wrong foundational question, and you’re instantly radioactive. Persona non grata. Careers don’t end with explosions — they end with polite silence and no invitations.
I probably should have committed to heterodox philosophy from the start. Or stayed a musician.
I remember leaving graduate school, putting on a suit, and feeling like I was wearing a costume. Cosplay, before we had the word. “Business professional” as a role, not an identity.
I’ve always felt intellectually capable of doing whatever I set out to do. My temperament, however, has never agreed to play along.
Which is perhaps why diagnosing ontologies comes so naturally. Once you see the scaffolding, you can’t unsee it – whether it’s metaphysics, jurisprudence, or a corporate department pretending it has a mission.
Then David Graeber came along with Bullshit Jobs, and I remember thinking: Thank God. It’s not just me.
So yes. I need a break.
I need sleep. I need silence. I need to stop reading law books that accidentally admit they’re about power and then act surprised when someone notices.
The Enlightenment, we are told, was the age of Reason. A radiant exorcism of superstition. Out went God. Out went angels, miracles, saints, indulgences. All that frothy medieval sentiment was swept aside by a brave new world of logic, science, and progress. Or so the story goes.
Audio: NotebookLM podcast on this topic.
But look closer, and you’ll find that Reason didn’t kill God—it absorbed Him. The Enlightenment didn’t abandon metaphysics. It merely privatised it.
From Confessional to Courtroom
We like to imagine that the Enlightenment was a clean break from theology. But really, it was a semantic shell game. The soul was rebranded as the self. Sin became crime. Divine judgement was outsourced to the state.
We stopped praying for salvation and started pleading not guilty.
The entire judicial apparatus—mens rea, culpability, desert, retribution—is built on theological scaffolding. The only thing missing is a sermon and a psalm.
Where theology had the guilty soul, Enlightenment law invented the guilty mind—mens rea—a notion so nebulous it requires clairvoyant jurors to divine intention from action. And where the Church offered Hell, the state offers prison. It’s the same moral ritual, just better lit.
Galen Strawson and the Death of Moral Responsibility
Enter Galen Strawson, that glowering spectre at the feast of moral philosophy. His Basic Argument is elegantly devastating:
You do what you do because of the way you are.
You can’t be ultimately responsible for the way you are.
Therefore, you can’t be ultimately responsible for what you do.
Unless you are causa sui—the cause of yourself, an unmoved mover in Calvin Klein—you cannot be held truly responsible. Free will collapses, moral responsibility evaporates, and retributive justice is exposed as epistemological theatre.
In this light, our whole legal structure is little more than rebranded divine vengeance. A vestigial organ from our theocratic past, now enforced by cops instead of clerics.
The Modern State: A Haunted House
What we have, then, is a society that has denied the gods but kept their moral logic. We tossed out theology, but we held onto metaphysical concepts like intent, desert, and blame—concepts that do not survive contact with determinism.
We are living in the afterglow of divine judgement, pretending it’s sunlight.
Nietzsche saw it coming, of course. He warned that killing God would plunge us into existential darkness unless we had the courage to also kill the values propped up by His corpse. We did the first bit. We’re still bottling it on the second.
If Not Retribution, Then What?
Let’s be clear: no one’s suggesting we stop responding to harm. But responses should be grounded in outcomes, not outrage.
Containment, not condemnation.
Prevention, not penance.
Recalibration, not revenge.
We don’t need “justice” in the retributive sense. We need functional ethics, rooted in compassion and consequence, not in Bronze Age morality clumsily duct-taped to Enlightenment reason.
The Risk of Letting Go
Of course, this is terrifying. The current system gives us moral closure. A verdict. A villain. A vanishing point for our collective discomfort.
Abandoning retribution means giving that up. It means accepting that there are no true villains—only configurations of causes. That punishment is often revenge in drag. That morality itself might be a control mechanism, not a universal truth.
But if we’re serious about living in a post-theological age, we must stop playing dress-up with divine concepts. The Enlightenment didn’t finish the job. It changed the costumes, kept the plot, and called it civilisation.
Philosophics 