The Apparatus of Disbelief: Miranda Fricker in the Age of Synthetic Media

On what happens when it’s not just people who get discredited, but credibility itself.

There’s a specific kind of unease that’s become familiar lately that is not quite paranoia, not quite cynicism, but something in between. You watch a video clip and notice a faint wrongness about the mouth. You read a statement attributed to someone and realize you have no clean way to verify it. You see a wave of comments all expressing the same view in subtly different words and wonder, briefly, if any of those people exist. Then you move on, because what else can you do.

This is epistemic anxiety in its contemporary form. Most of the frameworks we have for thinking about it are either too technical (detect the deepfake, verify the source, check the metadata) or too individualist (be a more careful thinker, cultivate healthy skepticism, don’t share before you fact-check). Both of these miss something important, something structural and social about what’s actually happening to our collective capacity to know things.

Miranda Fricker, a philosopher currently at NYU, published a book in 2007 called Epistemic Injustice: Power and the Ethics of Knowing that I believe applies to what many are experiencing in their relationship with AI. Fricker’s book wasn’t written about AI, rather it was written about something that seems much older and more intimate: what happens when a person is wronged not by being hurt or defrauded, but by being not believed. Her framework was built around human prejudice and social hierarchy, but the concepts she developed have a strange and uncomfortable resonance with where we are now, and applying them to the current moment reveals things that purely technical or individualist framings don’t.

What Fricker Argued

The core claim of Epistemic Injustice is that there is a distinctive kind of wrong that can be done to someone specifically in their capacity as a knower. Not injured, not stolen from but wronged as someone who perceives, reasons, and communicates. Fricker identifies two main forms of this.

The first is testimonial injustice. This is what happens when a hearer gives a speaker less credibility than they deserve because of prejudice, specifically, identity prejudice. Socially privileged groups are given an “excess of credibility,” treated as the authority on their own experiences and others’, while oppressed groups experience a “credibility deficit,” regarded as unqualified to describe even their own experiences. The mechanism is what Fricker calls “identity power”, the way existing social hierarchies get smuggled into our assessments of who is worth listening to. These are not merely individual harms: testimonial and hermeneutical injustices come from and refer back to a social fabric within which the biases and prejudices that animate them are tightly entangled.

The canonical examples tend to involve race and gender. For example, a Black witness whose account of a crime is discounted, a woman in a medical consultation whose reported symptoms aren’t taken seriously. Psychiatric patients are particularly susceptible to this credibility deficit. The structure is always the same: the person’s testimony is accurate; they’re not believed; the reason they’re not believed is social, not epistemic.

The second form is hermeneutical injustice, and it operates at a deeper level. Hermeneutical injustice occurs “when a gap in collective interpretive resources puts someone at an unfair disadvantage when it comes to making sense of their social experiences.” This isn’t about being disbelieved. It’s about lacking the very concepts needed to articulate what’s happening to you.

Fricker’s paradigm case is Carmita Wood. While working in the Physics Department at Cornell University in the early 1970s, Wood had experienced persistent and unwanted sexual advances from her boss. When she quit and applied for unemployment insurance, she was required to explain why she’d left, but found herself at a loss for words. What she experienced is what we would now classify as a textbook case of sexual harassment, but at the time, this concept was not yet operative within her social milieu. She reported that she quit for “personal reasons” and was denied unemployment insurance.

The concept simply didn’t exist yet. Which meant the experience couldn’t be named, couldn’t be shared, couldn’t be validated. Before the concepts of “workplace bullying,” or “sexual harassment,” had been developed, the victims of hermeneutical injustice were unable to make these phenomena fully intelligible to themselves and to others. This failure was due to a lack of hermeneutical resources, a deficit traceable to structural injustice: the society the victims lived in had no interest in generating or using novel concepts that captured these particular experiences.

What generates hermeneutical injustice, on Fricker’s account, is unequal participation in what she calls the collective hermeneutical project. She defines this as the ongoing social process by which a society develops its shared vocabulary for understanding experience. Those who are involved in concept production have an “unfair advantage in structuring collective social understanding” that leaves a gap in collective understanding where a useful concept should be. The marginalized don’t just suffer individually. They’re locked out of the room where meaning gets made.

Fricker’s virtue-ethical response to both forms involves cultivating what she calls testimonial justice and hermeneutical justice, sensitivity in hearers to their own prejudicial deflations of credibility, and alertness to the possibility that when someone struggles to articulate an experience, the problem may be in the conceptual resources available, not the person. The form that hermeneutical virtue must take is an alertness or sensitivity to the possibility that the difficulty one’s interlocutor is having as she tries to render something communicatively intelligible is due not to its being nonsense or her being a fool, but rather to some sort of gap in collective hermeneutical resources.

The Problem Fricker Didn’t Have to Face

What makes Fricker’s framework so useful right now is that it names something individualist epistemologies tend to miss: credibility is a social resource, distributed unevenly across a social fabric, and the conditions for its just distribution can themselves be unjust. The problem isn’t just that you or I might be bad at assessing evidence. The problem is that the apparatus through which we collectively assess credibility can be compromised at the structural level.

Fricker built this insight to describe what human prejudice does to individual knowers, but there’s an extension available, one she didn’t need to make in 2007, but that feels pressing now. The synthetic media environment doesn’t just wrong specific individuals in their capacity as knowers. It threatens the apparatus of credibility-assessment itself.

Consider what’s actually happening. Coordinated deployment of AI-generated content manufactures synthetic consensus, exploiting platform architecture to manipulate perceptions of public opinion. As users become aware of AI-generated content, skepticism extends to authentic content, a phenomenon where rational actors discount all digital evidence. This is sometimes called the liar’s dividend in discussions of deepfakes: you don’t have to successfully deceive people with a fake video of someone. You just have to make people uncertain enough about all video that authentic footage can be dismissed as potentially fabricated. Deepfakes are no longer designed to fool everyone, but to make everyone doubt everything. Even genuine experts and journalists increasingly face reverse credibility attacks, where authentic voices are dismissed as synthetic because they sound or look too polished.

This is a new and specifically structural form of what Fricker described. It’s not that a particular person is being wronged in their credibility, it’s that the conditions for credibility are being systematically degraded for everyone. Rather than focusing on universal persuasion, the evidence indicates that synthetic media’s primary harm lies in destabilizing epistemic certainty, imposing a skepticism tax of sorts that shifts confidence away from institutional sources toward closed, peer-based networks.

The skepticism tax is exactly the right framing. Fricker’s testimonial injustice imposes an unfair burden on particular speakers. They have to work harder than others to be believed, to establish credibility that would be assumed for someone of a different identity. The synthetic media environment generalizes this burden. Now everyone has to work harder to be believed. The tax is more universally distributed, but the incidence isn’t equal in that those who already carried a credibility deficit carry more of it, while those with excess credibility have buffer to lose.

Hermeneutical Injustice Without a Culprit

Fricker’s hermeneutical injustice is, in her framing, purely structural, there’s no individual agent to hold responsible for the fact that a concept doesn’t exist. This is one of the more philosophically interesting features of her account, and it creates real puzzles for how to assign responsibility and develop remedies. The current epistemic situation has a structurally similar quality, but with an added wrinkle: the gaps in our collective hermeneutical resources are being actively engineered. We don’t yet have good shared concepts – legal, social, epistemic, moral – for many of the things that are happening to our information environment. What do we call it when a political actor floods a discourse with AI-generated consensus until real consensus becomes undetectable? What’s the right word for the experience of watching something and being unable to determine its ontological status? What concept captures the specific form of self-doubt that comes from knowing your own perception might be manipulated at industrial scale? Our response to this challenge remains largely technical: detect fakes, verify sources, spot manipulation. This approach is insufficient.

The insufficiency is partly hermeneutical. We’re in the position of Carmita Wood, struggling to articulate something that’s genuinely happening, lacking the vocabulary because the concepts haven’t been developed yet. And the people with the most interest in developing those concepts, those most exposed to coordinated epistemic manipulation, those most harmed by synthetic consensus, are often not the people with access to the rooms where concepts get developed.

There’s a specific cruelty here that Fricker’s framework helps to identify. The people whose testimony is already subject to credibility deficits, whose accounts are already liable to be dismissed because of who they are, are also disproportionately the targets of coordinated disinformation. The assignment of credibility or the lack thereof is often a result of existing systemic hierarchies and normative standards, both of which are often so deeply ingrained within societies that even members of marginalized groups may become convinced of its truth. When the credibility apparatus is further degraded by synthetic media, the people with the least credibility buffer lose the most.

What Fricker’s Vocabulary Gives Us

The particular contribution of reading Fricker in this context isn’t that she provides solutions. She doesn’t, her work is fundamentally diagnostic. What she gives us is a way of talking about the social and ethical dimensions of epistemic harm that technical frameworks don’t reach. When we frame AI-generated misinformation primarily as a detection problem, we’re treating it as a technical failure – the fake isn’t convincing enough yet, but we need better forensics to catch the ones that are. When we frame it as an individual reasoning problem – be a better critical thinker, maintain healthy skepticism – we’re treating it as a cognitive failure, something each person can in principle fix by upgrading their epistemic practice.

What Fricker’s framework adds is the recognition that epistemic harms are social injuries. They damage relationships of trust and recognition between people. They corrupt the conditions under which people can be heard, and under which they can make sense of their own experience. Testimonial and hermeneutical injustices prevent people from becoming who they are. That’s a striking formulation, and it applies here. When the apparatus of credibility is systematically compromised, people can’t fully become knowers, they can’t have their testimony properly received, they can’t develop shared concepts for what they’re experiencing, and they can’t participate as full members of the epistemic community.

There’s also something useful in Fricker’s insistence on the virtue-ethical dimension. She doesn’t just describe the structural conditions that generate epistemic injustice, she asks what epistemic virtue looks like in a world where those conditions exist. The answer isn’t simple detection or skepticism. It’s something more like a heightened attentiveness to the conditions that make credibility possible. Preserving truth in an environment saturated with synthetic media requires not only verifying claims but also establishing shared definitions of authenticity, accountability, and the epistemic trustworthiness of online content itself. That’s a collective and political project, not just an individual one.

The epistemic virtue appropriate to this moment isn’t, I think, generalized suspicion, though suspicion has its place. It’s something closer to what Fricker describes as hermeneutical sensitivity: an attentiveness to the possibility that when people around you are struggling to articulate an experience, the problem might be structural rather than personal. Applied to the current situation, it would mean staying alert to the ways the credibility environment itself might be distorted, and resisting the temptation to individualize failures of knowing that are actually failures of the shared infrastructure of trust.

The Limits of the Analogy

Fricker’s framework doesn’t map perfectly onto synthetic media, so I want to be precise about where it strains. Her account of testimonial injustice is fundamentally about identity prejudice; the specific mechanism that deflates a speaker’s credibility is a hearer’s prejudiced response to who that speaker is. The credibility disruption from synthetic media is partly different: it’s not necessarily targeted at identity in the same way. It corrupts the whole channel, not specifically the reception of particular speakers. This means some of her virtue-ethics, which are centered on individual hearers correcting their own biases, doesn’t translate directly.

Fricker’s critics, particularly José Medina, have also argued that she underestimates the resistant epistemic practices of marginalized communities, specifically their capacity to develop their own hermeneutical resources rather than waiting for hegemony to supply them. Fricker argues that collective hermeneutical resources are imbalanced, favouring powerful groups who possess appropriate interpretations, while marginalized voices often face significant interpretative challenges. Medina’s corrective matters here too: the response to hermeneutical gaps shouldn’t be understood only as waiting for better concepts to be handed down, but also as attending to the epistemic practices that communities under pressure develop from within. Communities facing coordinated disinformation often develop sophisticated internal practices of verification and trust that mainstream epistemology doesn’t notice or value.

These aren’t reasons to abandon Fricker’s framework. They’re reasons to use it carefully, as a conceptual resource rather than a template. What she provides is a way of holding onto the social and ethical stakes of what we’re describing when we talk about epistemic anxiety. The anxiety isn’t just about the difficulty of knowing things, though it is that. It’s also about what it means to live in a community where the conditions for recognizing each other as knowers are being compromised. About what gets lost when testimony can’t be trusted, when shared concepts don’t exist for shared experiences, when the apparatus through which people become legible to each other as rational agents is systematically degraded.

That’s not a technical problem, and it’s not a problem any individual can solve by thinking harder. It’s a form of injustice that is distributed, structural, and in urgent need of vocabulary.