Epistemic injustice: when a voice is prevented from becoming testimony

Epigraph

“in their capacity as a knower”.

Miranda Fricker, Epistemic Injustice (2007), Introduction, p. 1. This verified fragment follows the English original; RU uses our own translation. It concerns a wrong done to someone specifically as a participant in knowledge. Introduction on the author’s website.

What this article is about

Someone reports a malfunction. People listen, nod, and continue working. An hour later, a more influential colleague repeats the information, and the problem finally becomes a fact requiring action. In this hypothetical scene, the first participant was not prevented from speaking words. They were prevented from conveying knowledge.

Epistemic injustice concerns failures of recognition: who receives trust, whose descriptions become intelligible, and who can participate in shared understanding of an event. The concept does not turn every disagreement into discrimination or require belief in every account. It helps distinguish substantive doubt from doubt attached to a person beforehand. This article considers Fricker’s two initial distinctions, conditions of listening, and possibilities of repair. The practical aim is to make testimony available for honest examination while retaining the right to doubt. A respectful “I heard you” matters when we understand what was heard and what possibility it gives the speaker next.

I. A voice can sound and remain excluded

The right to speak differs from the possibility of recognition as an information source. A meeting may appear open: everyone received two minutes and no question was prohibited. Yet one account is considered information, another emotion, and a third noise. Equal time conceals unequal status.

Fricker calls reduced credibility owing to prejudice testimonial injustice. The reduction’s grounds matter in her initial model: doubt does not become unjust merely because a speaker wanted trust. Hermeneutical injustice, the second concept, concerns unequal conditions for understanding social experience. Fricker, Introduction, pp. 1 and 4–7.

The opening scene reveals the first mechanism. Content remained unchanged while the response changed with the speaker. This grounds investigation of differing trust. Its cause cannot yet be named confidently: the second colleague may have confirmed competence and the first a history of error. But systematic differentiation of equally reliable messages by gender, age, or origin raises a question about recognition’s procedure.

The harm is double. Useful information is missed, and someone learns that their ability to report the world does not count. Even when information is accepted later, the initial exclusion does not automatically disappear. Authorship, participation in correction, and the right to explain conditions may remain with someone else.

Precise repair therefore begins beyond general goodwill. Content and origin must be returned: who reported what, when, and on which grounds. This is no cult of the first speaker. Error should be recorded too. The route from report to decision should depend on verifiable substance, rather than habitual allocation of the right to count as knowledgeable.

II. Doubt has a right to exist

Trust is necessary for shared life, but unconditional trust is incompatible with examination. People can err, repeat rumor, or misunderstand. Requiring equal belief in everyone removes distinctions of grounds rather than injustice.

In a hypothetical example, an employee reports a device is dangerous. He heard it from an acquaintance but lacks the model and circumstances. Another participant provides measurements and describes a particular failure. Different trust is appropriate, explained by access and information quality. Respect for the first remains: his signal can be clarified without becoming a ready-made conclusion.

Another situation arises when the same detailed report is rejected because its author is “too young” or “does not look like a specialist.” Why that feature concerns competence on this particular question must be shown. Age can relate to experience, but is not itself a check of the report. Appearance connects even less securely with truth.

Personal trust and trust in a claim should be distinguished. Someone can be loved while their causal hypothesis is rejected. A reliable observation can be recognized without liking its speaker. Merging these levels makes every clarification feel like rejection of closeness and every accepted fact like moral approval of the speaker.

Honest doubt can be explained: event time is missing, the source unclear, or a contrary observation exists. Prejudiced doubt often changes requirements after each answer. Details are demanded, then an independent witness, then proof that will still be disbelieved. Listeners should ask what could change their assessment. Without an answer, examination may merely imitate openness while the negative verdict is already settled.

III. Speech form does not determine information quality

Confidence makes a message persuasive but guarantees no reliability. Anxiety, pauses, and unfamiliar language may hinder a listener while remaining compatible with accurate observation. Acknowledging this particularly matters in conversations about events costly to their speakers.

A hypothetical participant reports recurring interruption and loses her thread. Another gives a coherent account in familiar business language. The listener tends to trust the second. But coherence may reflect presentation practice, and a pause the search for words for previously undiscussed experience. Assessing the message requires returning attention to events.

Form still matters. An incoherent account can obstruct verification; terminology can conceal contradiction. Difficulty understanding and falsity, however, are different conclusions. Episodes can be separated, sequence clarified, and the account restated in the listener’s words with an opportunity for correction.

A particular danger occurs when calmness is required only of the less influential participant. One raised voice seems decisive; another unreliable. The difference needs establishing through cases, rather than guessing from theory. But the evaluation rule should withstand comparison: does speech affect consideration of substance consistently?

The listener’s inner work involves delaying an immediate impression. “I struggle to follow” does not mean “this person knows nothing.” “Their language is familiar” does not mean “they are right.” Brief clarification can recover useful information otherwise lost through form. This changes the relationship: someone becomes an interlocutor with whom meaning can be recovered, rather than an object assessed by how conveniently they deliver it.

IV. Sometimes a shared language is missing

Hermeneutics is the work of understanding and interpretation. Hermeneutical injustice arises through unequal participation in explaining experience, rather than every missing word. Significant events are harder to understand and communicate when available concepts were developed without the people concerned.

Fricker’s introduction gives the example of sexual harassment in an environment lacking the relevant critical concept. This is a philosophical example of the problem’s structure; the entire history of the term is not recounted here. Fricker, pp. 1 and 6–7.

A hypothetical workplace shows a broader question. Someone tries to explain being required to remain available in the evenings despite the official workday’s end. The organization has words for a shift, overtime task, and voluntary help. None accurately describes the recurring expectation. Choosing the wrong word makes dismissal easy by definition: “That is not overtime.” Discussion ends before the event is examined.

A new name does not solve everything either. It may be too broad and join different situations. Conditions need description: who expects an answer, what happens upon refusal, and how often? A term should help scrutiny rather than replace it.

An important distinction lies between someone finding expression difficult and someone supposedly lacking meaningful experience. The first leaves work for conversational participants. The second closes it and assigns every deficit to the speaker. Sometimes allowing several descriptive sentences instead of a predefined category is enough. Meaning becomes available before a final name. This begins shared understanding without confirming every later theory.

V. A concept helps and can take over experience

A name connects cases, reveals repetition, and enables discussion of change. But it can also dictate what supposedly happened to someone. When a prepared category comes first, the account is fitted to it and loses details.

In a hypothetical conversation, learning “epistemic injustice” first explains someone’s irritation at meetings. This may be a useful distinction. But if every disagreement now confirms the term, substantive objections vanish. A concept intended to improve scrutiny becomes protection against it.

Apply the term after describing the mechanism. Is trust reduced? Why? Who has information access? Which conditions might otherwise explain the difference? Separate answers provide a more precise picture than disputing whether an event deserves a prominent name. Even with classification open, a concrete communication rule can sometimes improve already.

Not every injustice is epistemic. An unpaid bonus may concern employment, an unsuitable deadline organization, and a rude remark interpersonal relations. An epistemic aspect appears when participation in knowing is affected: a message is unrecognized, explanation prohibited, or question excluded beforehand. Several aspects may coexist without merging into one word.

In “Evolving Concepts of Epistemic Injustice,” Fricker discusses the concept’s development and relations to direct and indirect discrimination. This supports openness to new cases without making the term mean every dissatisfaction. Fricker, 2017, opening pages. Conceptual precision protects speakers too: a particular mechanism makes the needed procedural change easier to explain.

VI. Silence has different causes

No report does not prove no knowledge. Someone may be silent through ignorance, exhaustion, unwillingness, or expected futility. One outward sign conceals different decisions and conditions. Respect for silence means assigning no single meaning to it.

In “Silence and Institutional Prejudice,” Fricker distinguishes imposed inability to be heard and a listener’s attentive silence. She also shows that not every credibility deficit is unjust: an innocent mistake about competence is possible. Fricker, author’s PDF, pp. 1–5.

A hypothetical team therefore needs two checks. Who is not asked despite having information? What happens after they are asked? Invitation helps little if the answer never enters a decision. Repeated episodes may reduce someone’s reports; that reduction is then mistaken for lack of initiative.

Recognition cannot be extracted by force. The right to be heard includes the right not to speak. Someone may decline to share even in a kind environment. A useful invitation explains its subject and leaves refusal unpunished. It requires no proof of hidden experience and makes silence no further duty of the speaker.

A listener’s silence can be active: let an account finish, avoid rushing to a category, and avoid filling a pause with advice. It can also be evasion. When someone reports a concrete problem and expects an intelligible response, endless “space for expression” cannot replace examination. Listening should connect with a next step: what is established, what needs clarification, who responds, and when the result will return to discussion.

VII. A system can exclude without a rude remark

Prejudice may operate through procedure as well as personal attack. A form accepts certain reports, a schedule excludes participants, or a complaint rule demands information held only by the other side. People follow instructions outwardly, while important testimony cannot pass.

Consider hypothetical support. Reporting a missing document requires its number, which the user never received. Their message returns as incomplete. Neither discrimination nor epistemic injustice in a strict sense is proven yet. But a mechanism is found: information about a failure is blocked by a requirement incompatible with that failure.

Such analysis helps more than early accusations of malice. Conditions can be checked and an alternative route created. Then examine who the earlier rule particularly affected and why. If it systematically excluded a group’s reports from recognition, injustice gains firmer grounds.

Fricker’s institutional-prejudice work concerns both individual listeners and public institutions’ justice. Here the argument is applied to access arrangements as a philosophical comparison; no particular service is declared investigated. “Silence and Institutional Prejudice”.

Responsibility for procedure cannot be transferred entirely to speakers. Teaching form completion helps if the form fits the event. Otherwise training merely conceals exclusion. A system should recognize its own boundary. Technical “not accepted” otherwise becomes human “does not exist,” and no record proves no problem when the recording procedure itself prevented reporting it.

VIII. Repair needs a trace

Recognizing a missed message begins correction but does not complete it. Which decision lacked the information, what needs revision, and how can the person regain participation in examination? An apology without procedural change may leave the mechanism intact.

The second colleague repeated the malfunction report in the opening scene. The team could simply repair the failure. But the first participant again becomes merely the occasion for an event documented by others. More precise action records the initial report, clarifies its grounds, and includes its author if they wish.

No public ceremony is required. Correcting minutes, restoring the original observation’s reference, and naming a processing rule may suffice. Repair’s scale follows harm’s scale. A decision affecting many needs consequential checks; one conversational problem may need only a clear correction of its retelling.

Impossible promises should be avoided. A system can acknowledge unjust exclusion without fully restoring lost time or opportunity. An honest response distinguishes what was repaired, what remains, and what is offered next. Recognition must not become a demand for forgiveness or immediate renewed cooperation.

A trace permits examining the result. Who receives reports now? Did grounds of trust change? Can inconvenient information still be conveyed? One successful episode proves no permanent repair. A small concrete error needs no endless general audit either. Proportionate examination of the initial mechanism is needed. Repair has content when the deprived action is available again: offering information, receiving an explanation of a decision, and objecting if needed.

IX. The listener may also not know

Good listening includes acknowledging limits. Sometimes a listener lacks context or terms. This does not automatically make them guilty. Once a boundary appears, however, they can ask, postpone judgment, or declare the unintelligible meaningless.

A hypothetical manager first hears about work between official tasks. “We have no such process” may follow because the diagram omits it. Alternatively: “Which actions do you mean?” The latter confirms no complaint. It allows examination of whether the diagram matches actual work.

Admitting ignorance is hard when a role requires confidence. But procedural confidence may conceal missing information. Authority is then defended at knowledge’s expense. More durable authority distinguishes: I am responsible for deciding but lack some circumstances; I need to hear who observed them.

In Chapter II of On Liberty, John Stuart Mill connects understanding a position with serious consideration of objections, rather than one’s own retelling of an opponent alone. This is a liberal philosophical argument about discussion, rather than Fricker’s completed theory. Mill, Chapter II.

Expectations have a limit. One listener need not specialize in every possible experience. They can name a restriction and find another channel. Transfer must nevertheless not endlessly send someone onward. An intelligible next participant and preserved subject are needed. Otherwise ignorance becomes the speaker’s recurring burden: beginning again each time without ever conveying testimony through to a decision.

X. Justice to an account does not choose a winner

Two people may sincerely describe one event differently. Their right to participate in knowledge does not require accepting both explanations. A fair procedure retains disagreement and distinguishes contradiction, perspective, and uncertainty.

In a hypothetical dispute, one employee says a proposal was rejected without consideration. Another recalls a substantive objection. A record, sequence, and decision criterion can be checked. Without a record, two testimonies remain. No winner should be selected through vocal confidence or preassigned victim status. Uncertainty can be retained while rules for recording future decisions change.

Fairness applies to whoever is criticized too. Prejudice cannot be used to prohibit their explanation. They may know circumstances the first speaker does not. Examining that information does not cancel possible injustice; it establishes it more accurately.

Symmetry does not require equal risk. One participant controls documents; another depends on their decision. A formally equal demand for evidence may imply different possibilities. Equal rules therefore need assessment of access. A missing document cannot be demanded from someone never given it.

The result may be bounded: content suffices to examine a failure, while the cause of earlier distrust remains unknown. This permits action without invented motive and leaves a next question open if information appears. A speaker’s dignity is retained through actual substantive consideration, rather than promised victory. For everyone, this is firmer than goodwill dependent on how conveniently their account sounds.

XI. The right to know connects with changing the shared explanation

People participate in knowledge as more than fact providers. They can propose a distinction, ask a new question, and challenge an accepted description. Participation is limited if only supplying material is allowed while interpretation belongs to others beforehand.

A hypothetical project collects users’ stories, then declares all difficulties insufficient familiarity. One participant objects that the instructions require an impossible step. If they may report inconvenience but not discuss explanation, knowledge loses an important test. The participant can be wrong, but the question requires a substantive answer.

This gives no interlocutor equal competence in every theory. Questions differ by subject. But competence gives no right to declare a message insignificant before considering its connection with the conclusion. Even a short objection can expose a hidden assumption.

Fricker’s question about the person as knower therefore has practical depth. Can someone change both a record’s row and the way the group understands events? Recognition of this possibility guarantees no acceptance of the change. It retains a duty to examine its grounds.

Return to the malfunction report. A corrected mechanism allows the first participant to say both “the device stopped” and “you overlook the conditions in which this recurs.” Voice becomes testimony and question. A new distinction enters the picture. Justice and knowledge meet here: someone gains a place in understanding the world, while the group gains sight of what its earlier rules could not reveal.

XII. Trust can be restored without demanding forgetting

After procedural repair, someone may still doubt the next account will be considered. This is not necessarily stubbornness or unwillingness. Earlier experience supplied information about the environment; new promises have yet to provide comparable grounds for revision.

A hypothetical manager announces proposals will now retain authorship. Someone repeatedly ignored does not hurry to speak. Demanding instant trust as proof of goodwill would be unfair. A small verifiable step is more reliable: discuss one question, retain the result, and allow inspection of the record. New procedural information appears through action.

Neither side must endlessly prove moral worth. Repair aims at a working possibility, rather than a final personal verdict. A limited format, definite time, and intelligible way to stop can be agreed. If participation is too costly or unwanted, refusal remains a right, rather than admission of error.

Forgiveness should be distinguished from changed expectations. Someone may resume shared work without forgiving an episode, acknowledge repair but choose another environment, or trust one participant while doubting the system. These decisions concern their life; fair-listening theory does not assign the required result for them.

Substantive results appear in new conditions. A message has an intelligible route, authorship survives, doubt is explained, and the response permits objection. Repetition gives grounds for updating expectations. Trust is neither purchased by an attractive promise nor required as a duty. Experience gradually supports it when a voice can become knowledge and affect shared choice.

What this means in practice

Begin with one episode of a missed or rejected message. Record content, source, and grounds of response. No immediate diagnosis of a relationship or declaration of prejudice is needed. Establish what lacked consideration first.

Separate three questions: is the message true, are grounds for trust sufficient, and is procedure fair? Answers can differ. Someone may err under an unequal rule or report truth while the cause of earlier distrust stays unknown. This distinction prevents an excessively broad verdict.

As listener, ask for one example, allow correction of your account, and name information that would change assessment. Check whether the same threshold applies to others. If a description is unclear, specify what is missing rather than saying “it is subjective.”

As speaker, you may request procedure: “I want this observation checked,” “we need to clarify why identical reports received different answers,” or “the proposed category misses a condition.” You need not prove a general theory of your life for one question and may decline unsafe disclosure.

For shared work, preserve a trace: message, consideration, result. Repair both wording and the next report’s route. An accepted fact, confirmed injustice, and successful change are different results. Good listening distinguishes them and continues without requiring obligatory agreement or immediate forgiveness.

Key quotation

“He who knows only his own side of the case, knows little of that.”

John Stuart Mill, On Liberty (1859), Chapter II. The quotation follows the English original; RU uses our own translation. This argues for considering objections rather than accepting any information without scrutiny. Text.

Connections between traditions

Comparison	Similarity and boundary
Fricker: testimonial injustice ↔ epistemology of testimony	Grounds of trust matter. Not every mistaken doubt is prejudice.
Hermeneutical injustice ↔ shared description of experience	Shared language affects understanding. A missing word alone does not prove unequal participation.
Mill: a strong objection ↔ fair listening	Both require encountering another’s argument. Liberal defense of discussion does not exhaust social-prejudice theory.
Fricker: institution ↔ designing a reporting form	Procedure can restrict knowledge transfer. A hypothetical technical example proves no discrimination by a particular service.
Attentive silence ↔ a speaker’s voluntary refusal	Both preserve choice. Listening silence and imposed speaking silence perform opposite functions.

Primary sources and research

Miranda Fricker. Epistemic Injustice: Power and the Ethics of Knowing (2007), Introduction. Pages 1 and 4–7 were read: two initial concepts, prejudice, and epigraph context. Reading of the entire book is not claimed.
Miranda Fricker. “Evolving Concepts of Epistemic Injustice” (2017). Opening passages of the author’s PDF were read: conceptual development and discrimination.
Miranda Fricker. “Silence and Institutional Prejudice”. Pages 1–5 of the author’s manuscript were read: silence types, competence error, and institutional questions. The later historical criminal case is not used.
John Stuart Mill. On Liberty (1859), Chapter II. The passage on considering objections was read: key quotation and comparison of discussion traditions.

About the authors

Andrey Zarinsh is the author and founder of United Wisdom. He explores inner transformation, the unity of traditions, intercultural dialogue, and contemporary spiritual discipline. Instagram: @andrey.zarinsh.

Maria Evilkut is a coauthor of United Wisdom. Alongside her research, she seeks to test these ideas in practice and observe how principles of unity and mutual understanding work in real life between people and cultures. Instagram: @ilkutki.