Miranda Fricker made epistemology moral, and then moved on

On 30 September, at ten past four in the afternoon, Miranda Fricker will stand up in Deutsches Haus on Columbia's Morningside campus and begin the first of three John Dewey Lectures. The series runs to 2 October under the title "Ethical Pressures: Shaping Wills and Bending Time in Moral Address". Nothing in that title mentions knowledge, testimony, credibility or injustice. For the philosopher who gave the English-speaking world the phrase "epistemic injustice", the omission is the most interesting thing about the event.

The standard account of Fricker runs like this: a feminist epistemologist coined a serviceable term in 2007, the term escaped the seminar room, and a literature of several thousand papers grew up around it. All of that is true and all of it undersells her. The achievement of Epistemic Injustice: Power and the Ethics of Knowing (Oxford University Press, 2007) was not lexical. It was to show that the passing of knowledge between persons has a moral structure — that one act can go wrong epistemically and ethically at once, for the same reason, and that a theory of knowledge blind to this is missing something about its own subject. Analytic epistemology had spent decades on justification and the Gettier problem. Fricker pointed out that most of what any of us knows we were told, by somebody, in a room, and that rooms have power in them.

Fricker has been Julius Silver Professor of Philosophy at New York University and co-director of the New York Institute of Philosophy since 2022. Before that she taught at Birkbeck, at Sheffield, where she remains an honorary professor, and at the CUNY Graduate Center. She is a Fellow of the British Academy and of the American Academy of Arts and Sciences, and gave the 2022 presidential address to the Eastern Division of the American Philosophical Association, on forgiveness.

She has since gone somewhere else. Her current projects, as she lists them herself, are blame, apology and forgiveness, and the idea of "origins" in the ethical philosophy of Bernard Williams, who supervised her Oxford doctorate with Sabina Lovibond and died in 2003. The Uehiro Lectures she gave at Oxford in June were on blame; the Dewey Lectures are on moral address; the Tanner Lectures she is due to give at Berkeley in 2027 are on Williams. A philosopher whose name is welded to one concept has spent roughly a decade writing about something else.

Her paradigm cases are a novel and a film, and that was deliberate#

Testimonial injustice, in Fricker's definition, occurs when "prejudice causes a hearer to give a deflated level of credibility to a speaker's word". The term for the damage is an identity-prejudicial credibility deficit: the hearer discounts what is said because of who is saying it, and the discount tracks a prejudice about the speaker's social group rather than anything about her reliability.

Her central case is Tom Robinson, the black defendant in Harper Lee's To Kill a Mockingbird, whose testimony an all-white Alabama jury cannot hear as testimony at all. Her second is quieter and more useful for it. In The Talented Mr Ripley, Marge Sherwood works out what has happened to her fiancé, and Herbert Greenleaf, the missing man's father, sets her conclusion aside as female intuition rather than a claim about facts. She is right. He does not hear her as someone in a position to be right.

Two fictional examples to anchor a work of analytic epistemology is an unusual choice, and Fricker made it with her eyes open. The point of the Greenleaf case is that the prejudice at work is neither malicious nor stupid. Greenleaf is courteous; he is also operating a rule about who gets to hold facts. Fricker's argument is that this is an epistemic failure with an ethical core rather than a lapse of manners with an epistemic side effect. The prejudice is why the belief is not formed, and the speaker is wronged, in her words, in a capacity essential to human value. That is the structural claim, and it is what makes the book more than a taxonomy: the moral wrong and the epistemic failure are one event described twice.

The remedy she proposed is where the account is weakest#

Fricker's cure is a virtue. Hearers assess credibility through what she calls testimonial sensibility: a trained, largely perceptual disposition, socially acquired, that runs ahead of conscious judgement. Since that training is what carries the prejudice, the corrective has to work at the same level. The virtue of testimonial justice, on her account, requires a hearer to reliably neutralise prejudice in her judgements of credibility (2007, p. 92) — to notice the downward pull and correct upward for it.

It is an elegant proposal and it asks an enormous amount of individuals. Elizabeth Anderson pressed the obvious objection in Social Epistemology in 2012, in "Epistemic Justice as a Virtue of Social Institutions": where credibility deficits are produced by segregation, by who sits on which committee and who never meets whom, a well-trained hearer inside a badly designed institution can do very little. Virtue talk can become a way of assigning homework to individuals in place of reform. That is less a refutation of Fricker than a statement of scale, and scale is the whole question when the subject is injustice.

Carmita Wood had no word for what was happening to her#

The book's second half turns to a harder case. Hermeneutical injustice is, in Fricker's formulation, "the injustice of having some significant area of one's social experience obscured from collective understanding owing to a structural identity prejudice in the collective hermeneutical resource" (2007, p. 155). The wrong is not disbelief. The concept required to make the experience intelligible, to others and to oneself, does not exist, because the people who needed it were kept out of the business of making meanings.

Everyone repeats the sexual harassment example; the history behind it is specific, and Fricker tells it properly, drawing on Susan Brownmiller's memoir of the women's movement, In Our Time. Carmita Wood was forty-four and had worked for eight years in Cornell's nuclear physics department when a distinguished professor there proved unable to keep his hands off her. She developed physical symptoms, asked for a transfer, was refused, and resigned. When she applied for unemployment benefit she had to state why she had left, and the reason she had was not a reason the state recognised, because it had no name. Her claim was denied.

In 1975 Wood went to Lin Farley, who taught a course on women and work at Cornell. Farley, with Susan Meyer and Karen Sauvigne, founded Working Women United and organised a campus speak-out on 4 May 1975 that drew 275 people. The naming happened while they were making posters for it. Sauvigne's account, which Fricker quotes:

"Eight of us were sitting in an office of Human Affairs, brainstorming about what we were going to write on the posters for our speak-out. We were referring to it as 'sexual intimidation,' 'sexual coercion,' 'sexual exploitation on the job.' None of those names seemed quite right. We wanted something that embraced a whole range of subtle and unsubtle persistent behaviors. Somebody came up with 'harassment.' Sexual harassment! Instantly we agreed. That's what it was."

By August, Farley was using the phrase in testimony before New York City's Commission on Human Rights: from a room in Ithaca to the municipal record in under four months, for a category of harm as old as paid employment. Fricker's point is that everything before that room was hermeneutical injustice, and the person who paid for it most directly was the woman whose benefit claim could not be written down.

The objections are serious and they have names#

José Medina's The Epistemology of Resistance (Oxford University Press, 2013) makes two charges. The first is that Fricker looks at the wrong end of the transaction. Credibility is not handed out one speaker at a time; it is an economy, and deficits are the shadow of excesses. Tom Robinson's problem, Medina argues, is inseparable from the credibility excess the white jury grants his accuser, Mayella Ewell (2013, pp. 65–6). Fricker treats excess as at most a minor epistemic misfortune. Medina thinks it is where the machinery is. The second charge is methodological: the unit of analysis should be a person's epistemic character across extended periods and a whole social context, not one exchange between hearer and speaker. Fricker's framing is individualist and virtue-theoretic; the phenomenon is neither.

Kristie Dotson's objection cuts closer. In "Tracking Epistemic Violence, Tracking Practices of Silencing" (Hypatia, 2011) she describes testimonial smothering: a speaker who judges her audience incompetent to receive what she has to say truncates her testimony to the part she thinks will land. Fricker's account has no place for this, because there the hearer does everything and the speaker is acted upon. Dotson's silenced speaker is calculating, often accurately. Her second contribution, contributory injustice, appears in "A Cautionary Tale: On Limiting Epistemic Oppression" (Frontiers: A Journal of Women Studies, 2012): the speaker has perfectly good interpretive resources from her own community, and the audience persists in using its inadequate ones. Fricker, Dotson argues, drew the line between injustice and bad epistemic luck in the wrong place.

Gaile Pohlhaus Jr made the same move on the hermeneutical side in Hypatia in 2012, under the name wilful hermeneutical ignorance: members of dominant groups declining to take up concepts the marginalised have already built. Which brings the argument to its sorest point. Fricker insists that hermeneutical injustice is purely structural (2007, p. 159) — that it has no perpetrator. If nobody does it, nobody can be asked to stop; and the remedy she offers, a hearer's alertness to the possibility that a speaker is wrestling with a vocabulary that does not fit, is thin against a harm of that size. An injustice without an agent sits badly in a book whose central insight is that epistemic life is a relationship between persons.

Notice where that objection leaves you: asking who can properly be blamed, for what, and to what end. Which is, more or less exactly, the subject Fricker has spent the last decade on.

Blame as an address to the person someone is not yet#

The turn is not a rumour; it is in the journals. "What's the Point of Blame? A Paradigm Based Explanation" appeared in Noûs (vol. 50, no. 1, pp. 165–183) and argues that the paradigm form of blame is communicative, and that its point is to increase the alignment between the blamer's moral understanding and the wrongdoer's. "Ambivalence About Forgiveness" (Royal Institute of Philosophy Supplement 84, 2018, pp. 161–185) treats forgiveness as a mechanism of moral-social influence which necessarily contains the seeds of its own corruption, which is why our attitude to it wobbles. "Forgiveness — An Ordered Pluralism" and her replies to critics ran in Australasian Philosophical Review 3(3) in 2019. In 2025, "Strange credibility: 'avowal' as functionally factive testimony" appeared in Jurisprudence 16(1): the old subject and the new one in one paper, confession treated as testimony whose credibility is conferred rather than earned.

At Oxford on 1, 3 and 8 June 2026, in the H B Allen Centre at Keble College, the Uehiro Lectures set the argument out in three moves under the title "Moral Pressures: Bending Time, Shaping Wills"; we reported them at the time. The first was functionalist: blaming is a diverse set of activities rather than one thing, and what the practice produces, when it works, is shared moral understanding. The second, on apology, drew on Foucault's treatment of avowal and made a claim purists will not like — that a wronged party may legitimately reach for non-moral leverage, up to and including threats, and that an apology can do its social work without being sincere. The third was on proleptic blame.

Prolepsis is Williams's device, from "Internal Reasons and the Obscurity of Blame" in Making Sense of Humanity: you address someone as though they already had a reason they do not have, in the hope that the addressing brings the reason into being. Fricker, following Williams and Victoria McGeer, proposes that even a wrongdoer with no relevant moral understanding has a baseline disposition to want the esteem of others, and that blame can hook onto it. So blame need not be a verdict on the past. It can be an attempt to call into existence the better agent who does not yet exist, and because the mechanism is a desire for esteem rather than moral argument, it does so without pretending the wrongdoer is already someone he is not.

The Dewey abstract states the general principle: our familiar responses to wrongdoing, blaming and demanding apology among them, "inevitably and quite properly involve the exertion of interpersonal pressures". Moral address is leverage applied by someone who has been wronged, and the view that it may consist only of reasons is a bad description of what people do to one another. Readers of Williams's campaign against moral systems will recognise the target.

Continuity is easy to miss here. This is the shape of the 2007 book: a two-person transaction with a normative structure, analysed from the receiving end, the interest lying in what the addressee is treated as. Only the transaction has changed, from telling to blaming.

The hardest question is whether a scoring system is a hearer at all#

Fricker's framework is the one people now reach for when algorithmic systems distribute credibility, and there is a great deal of that to explain. In the Netherlands, the tax administration ran a risk-classification system over childcare benefit claims; some 26,000 parents were wrongly accused of fraud between 2005 and 2019, dual nationality was among the features drawing scrutiny, and the Dutch data protection authority found the processing unlawful, discriminatory and improper. The parliamentary report of December 2020 was titled Unprecedented Injustice; the third Rutte cabinet resigned on 15 January 2021.

In England and Wales the mechanism was legal rather than statistical. Section 69 of the Police and Criminal Evidence Act 1984, which required a party relying on computer output to show the machine had been working properly, was repealed in 1999 on a Law Commission recommendation, leaving a common-law presumption that computers operate correctly unless the contrary is shown. Hundreds of subpostmasters were convicted on Horizon data they could not challenge. The Ministry of Justice opened a twelve-week call for evidence on the rule in 2025; the justice minister Sarah Sackman said a blanket "no questions asked" acceptance of digital evidence "can have a devastating impact on people's lives". Fricker's vocabulary fits that sentence so neatly it is tempting to stop there.

It should not be. Testimonial injustice is defined in terms of prejudice in a hearer: an attitude, held by an agent, working on a perception of the speaker as a member of a group. A gradient-boosted classifier has no attitudes and perceives nobody as anything; it computes associations over features. You can stretch "prejudice" to mean statistical dependence on a protected characteristic, as much of the applied literature quietly does, at the cost of the moral content that made the concept worth having. Or you can locate the injustice in the humans downstream, the caseworker reading the flag and the magistrate reading the printout, which is faithful to Fricker and probably right, and which treats as a bystander the thing that did the work.

Medina's objection is the more useful tool here. Credibility is relational, and what happened to the subpostmasters is best described as an extreme credibility excess granted to a machine, with the deficit falling on whoever stood opposite it. The Dutch case has the same shape. Hermeneutical injustice may travel further still: people subject to automated decisions routinely lack both the information and the concepts to describe what was done to them, while the operator holds the whole interpretive resource. That matches Fricker's definition closely, except that it has an identifiable holder and therefore, awkwardly for the 2007 account, a perpetrator. Where a system decides what a voter sees, what a worker is measured on or what a court treats as reasoning, the shortage of concepts is not general; it is on one side of the table, which is roughly what standpoint epistemology has been saying for some time.

None of which Fricker is currently writing about. The Oxford lectures are to become a book in Oxford University Press's Uehiro Series in Practical Ethics; they were live-streamed and, in a decision that will irritate anyone who missed them, not recorded. Whoever turns up at Deutsches Haus at the end of this month expecting credibility deficits will get apology, pressure and the uses of esteem instead. The concept she is famous for is doing perfectly well without her.