Or how our silent predictions make discrimination feel normal
I find myself in a situation where I have older colleagues who often make those jokes – precisely because they feel they’re in an all-male environment – those blatantly sexual jokes about female colleagues which, to be honest, sometimes… are told with such naturalness and light-heartedness, as if we were joking about, I don’t know, any other topic, but without realising that, in fact, they’re really being disrespectful. […] And the problem is that I sometimes, or rather very often, find myself in a bit of a minority; so, I also struggle to put a stop to these incidents, in the sense that I have to endure them as a bystander, I see others going along with it and laughing at the joke or perhaps adding to it, and I am perhaps in the minority, or perhaps—I don’t know if the others are also concerned about the issue but don’t say anything—so the incident, well, goes unpunished, so it’s as if we all approve of the joke, even though I might not be the one adding to it or laughing at it or… Evidently, it’s either ignored or it’s perceived that, in any case, “I’m fine with it too”.
Italian participant
What this participant names is not an isolated incident. It is a structural phenomenon: the way in which a person’s silence becomes legible to the group as consent. No policy was violated and no explicit coercion occurred; yet something was built: a feeling of unanimity that permits the harm to go, in the speaker’s own words, unpunished. Understanding how such feelings are generated, and why they are so difficult to interrupt, requires looking not only at social dynamics but at the cognitive architecture that underlies them.
The predictive brain: a brief theoretical account
Contemporary neuroscience and cognitive psychology converge on a model of the mind that differs significantly from the intuitive picture most of us have of it. We imagine perception as a reactive process: the world provides stimuli, the brain receives and processes them, and conscious experience is the result. This picture is, in important respects, inverted. The brain, as Andy Clark has argued with sustained analytical force, is fundamentally a prediction machine. It does not passively await sensory stimuli, but continuously generates predictive models of what is about to happen; it does so on the basis of models derived from past experience, cultural learning and acquired habits. Incoming data are used primarily to confirm, adjust or correct these predictions.
This theoretical framework (i.e., predictive processing) posits that perception is the brain’s ‘best guess’ about the causes of sensory signals, shaped by everything it has previously learnt about the structure of the world. When reality does not match expectations, prediction errors arise that come at a high cognitive cost. The system is therefore strongly motivated to minimise surprises, which it achieves not only by updating its models when predictions fail, but also by actively selecting environments, attention and behaviours in ways that confirm what it already expects.
The brain does not perceive the world as it is; it perceives the world as it has learned to expect it will be, and those expectations carry the weight of cultural history.
The implications extend well beyond sensory neuroscience. If the brain’s representations of social reality — of who speaks freely, who is heard, who is credible, who poses a threat — are themselves predictive constructs, then those representations inherit the biases and normative assumptions embedded in the cultural contexts from which they were learned. The predictive brain is not a neutral processor. It is a socially constituted organ, trained on datasets that are anything but impartial.
Anticipation as a mechanism of social reproduction
To understand how predictive processing becomes a mechanism of inequality, it is useful to distinguish between two registers in which anticipation operates. The first is explicit and deliberate: the conscious stereotypes and prejudices that individuals can, in principle, identify, interrogate, and attempt to correct. The second is implicit, automatic, and largely inaccessible to introspection — the predictive priors that shape perception before conscious awareness intervenes. It is in this second register that discrimination becomes, in the precise sense, invisible: not merely to observers, but often to perpetrators and to those who bear its costs alike.
Return to the testimony above. The speaker identifies a precise cognitive bind: they are in a minority, they do not laugh, they do not add to the joke — and yet their silence is read by the group as approval. This is not a misreading. It is the output of a collective predictive system that has already resolved the question of group consensus before any deliberate evaluation occurs. The joke does not feel like a transgression to those who make it because the predictive architecture of the room has pre-assigned its legitimacy. The silence of the bystander, correspondingly, is not registered as dissent; it is processed as ambient agreement, absorbed into the background of what the situation is expected to look like.
This is what makes the concept of epistemic violence — developed in feminist philosophy and science studies — so structurally important, and so difficult to operationalise without recourse to cognitive theory. Epistemic violence does not require violent intent. It operates through the ordinary mechanics of recognition and credibility: through whose testimony is assumed to require corroboration, whose silence is read as consent rather than foreclosure. When these dynamics are grounded in predictive priors shaped by gendered or racialised hierarchies, the violence becomes genuinely systemic — distributed across the nervous systems of all participants, including those who are most harmed by it.
When the prior absorbs the harm
Perhaps the most analytically striking feature of the predictive processing account is its explanation of why bystanders — including those who are, like our speaker, privately troubled — so often fail to act. This is not simply a matter of social cowardice, though social pressure is real. It is, at a deeper level, a consequence of how the predictive system resolves the tension between a prior expectation and discrepant sensory data.
Clark and others have noted that when incoming data conflicts with strong predictive priors, the system has two options: update the prior, or attenuate the prediction error signal — effectively discounting the discrepant data as noise. In social environments where sexist humour has been sufficiently normalised, the prior for what a workplace interaction looks like already incorporates such moments as baseline expectation rather than anomaly. The joke does not register as a rupture requiring response; it registers as confirmation of what was already anticipated. The bystander’s discomfort, meanwhile, is an internal state without external correlate — a prediction error that the social field provides no framework for expressing.
The speaker captures this with precision: “I don’t know if the others are also concerned about the issue but don’t say anything.” This epistemic uncertainty — the impossibility of knowing whether one’s private dissent is shared — is itself a structural effect. It is produced by an environment in which the dominant predictive prior forecloses the expression of counter-norms, leaving each dissenting individual isolated inside their own unvoiced objection, unable to coordinate with others who may share it.
Structural origins, individual substrates
A critical caveat is necessary here, lest this framework be read as relocating structural problems inside personal psychology. The predictive priors that encode social hierarchy are not generated by individuals in isolation; they are products of structured social environments — of institutions, media ecosystems, legal regimes, and cultural practices that systematically distribute recognition and authority along lines of gender, race, and class. The brain learns from the world it inhabits. When that world is organised by inequality, the brain’s predictive models become repositories of that inequality — accurate maps of an unjust terrain that, precisely by accurately representing it, help to reproduce it.
This is the deepest sense in which the predictive brain is a political organ. It does not merely passively record social structure; through the anticipatory behaviours it generates and the perceptions it constructs, it actively participates in the maintenance of that structure. Challenging systemic discrimination therefore requires not only changing policies and institutions — though it requires that — but also addressing the conditions under which the brain forms its predictive models: the cultural environments and relational practices through which social hierarchy is encoded as expectation.
Toward a critical cognitive science of inequality
The convergence of predictive processing theory with feminist epistemology and critical race theory opens a rich and underexplored research agenda. Several questions seem particularly urgent. First, what are the neurobiological mechanisms by which culturally specific hierarchies come to be encoded as predictive priors, and over what timescales? Second, are there conditions — pedagogical, institutional, relational — under which predictive models of social authority become more plastic, more susceptible to revision in the light of counter-stereotypical experience? Third, and most practically: what forms of collective action are capable of disrupting the shared predictive field that transforms individual dissent into apparent consent?
The testimony we began with suggests one answer, tentatively. The speaker’s difficulty lies precisely in their isolation — in the impossibility of coordinating with other silent dissenters whose inner states remain invisible. Interventions that make dissent legible, that create social technologies for expressing objection without requiring heroic individual courage, may work partly by introducing prediction errors into an otherwise self-confirming system: by making visible what the predictive field had rendered invisible, and thereby opening space for the prior to be revised.
The meeting room, the lunchbreak, the workplace group chat where the joke circulates: these are not merely sites of interpersonal dynamics. They are places where structure becomes sensation — where the weight of gendered hierarchy is felt as a silence that cannot coordinate with other silences, as a discomfort that finds no form. Understanding how that process works is not a detour from the study of justice. It is, increasingly, one of its most necessary paths.