Someone tells you something you could not know alone.
A doctor explains a scan. A colleague reports what happened in a meeting you missed. A friend describes another person's behavior. A journalist quotes a source from a place you have never visited. A teacher says that a formula works even before you understand why.
Most human knowledge enters life this way.
Not through direct observation.
Through other minds.
The problem is obvious: if we depend on testimony, we can also be misled by it.
Epistemic trust names one part of the human solution to that problem.
Definition
For this Knowledge Library:
epistemic trust is a person's context-sensitive readiness to treat communicated information from another person or institution as sufficiently reliable, relevant and generalizable to be learned from, while remaining capable of revising that trust when evidence, source quality or context changes.
That definition deliberately includes both openness and selectivity.
Epistemic trust is not the absence of vigilance.
It is what makes social learning possible when vigilance has not produced a reason to close the channel.
The construct lives in more than one research tradition
There is no single disciplinary owner of epistemic trust.
In developmental and social-cognitive research, scholars study how people decide whom to learn from, how testimony is weighed and how cues such as competence, honesty and consensus shape selective trust.
Dan Sperber, Hugo Mercier and colleagues use the related concept of epistemic vigilance to describe mechanisms that protect communication from misinformation while allowing social learning to remain advantageous.
In the mentalizing and psychotherapy literature, Peter Fonagy, Chloe Campbell, Patrick Luyten and colleagues use epistemic trust more specifically to describe openness to socially transmitted knowledge that feels authentic and personally relevant.
These traditions overlap.
They should not be collapsed as though they were identical.
Trust is not the opposite of vigilance
The most useful correction to everyday language is this:
Trust and scrutiny are not simple opposites.
A person can begin open to a source and still inspect the claim.
They can trust a physician about radiology and not about tax law.
They can trust a friend about what she saw while remaining uncertain about her explanation of why it happened.
Epistemic vigilance, as Sperber and colleagues describe it, concerns the cognitive and social work of deciding when communication deserves acceptance.
Trust becomes dangerous when it bypasses that work entirely.
Source credibility is narrower
Source credibility usually refers to perceived or evidenced qualities of a communicator, such as expertise, trustworthiness or relevant access.
Epistemic trust is broader because it describes the learner's stance toward socially transmitted knowledge.
The same source can be credible in one domain and not another.
The same person can also differ in their readiness to learn from others across contexts.
This distinction matters for measurement.
A source-credibility scale does not automatically measure epistemic trust, and a general trust questionnaire does not establish whether someone calibrates learning to source quality.
The clinical tradition adds mistrust and credulity
One of the strongest recent developments has been the explicit separation of epistemic trust, epistemic mistrust and epistemic credulity.
In this framework, mistrust is not simply low trust. It is a tendency to treat communicated knowledge as unreliable or personally irrelevant even when learning might be warranted.
Credulity is not simply high trust. It is openness without sufficient discrimination, including difficulty checking whether information should be believed or generalized.
That three-part distinction is conceptually valuable because a person can oscillate between shutting information out and taking it in too easily.
A single trust continuum would miss that pattern.
Measurement is still developing
Campbell and colleagues introduced the Epistemic Trust, Mistrust and Credulity Questionnaire, or ETMCQ, in 2021.
The original validation included two studies with 500 and 705 participants and reported a three-factor structure.
Subsequent translations and validations have supported parts of that structure while also exposing instability, especially in some mistrust subscales and cross-language models.
A revised ETMCQ-R was published in 2025 with improved psychometric properties.
This is precisely why the Institute should not treat "epistemic trust" as a mature scalar trait with one settled score.
The measurement programme is active.
A questionnaire cannot capture the whole social-learning process
Self-report asks people to describe their typical stance toward information from others.
Actual epistemic trust unfolds in a relationship among person, source, claim and context.
Someone may report general mistrust while behaving selectively and sensibly in a domain they know well.
Another person may endorse statements of openness while deferring automatically to status in practice.
Behavioral tasks, longitudinal designs, source-learning experiments and context-specific measures can therefore answer questions that a global questionnaire cannot.
Construct validity depends on matching the measure to the claim.
Developmental evidence shows that trust can be selective early
Children do not simply absorb everything adults tell them.
Developmental research on testimony shows sensitivity to informants' past accuracy, knowledge access and other cues to reliability.
This does not mean children possess adult-level source criticism.
It means the architecture of social learning is selective from early in development.
That finding matters because epistemic trust should not be represented as a late, purely reflective choice.
Some of its components are relational, developmental and often fast.
The mentalizing account makes personal relevance central
In the Fonagy and Allison tradition, information becomes learnable not only because a speaker appears competent but because communication is experienced as directed to this person in a way that recognizes their mind.
The theory argues that ostensive cues, mentalizing and experiences of being understood can open a channel for social learning.
This is a richer claim than ordinary credibility.
It is also more difficult to test cleanly.
The empirical programme remains younger than the theoretical ambition, and clinical associations should not be mistaken for complete causal confirmation.
Mistrust can be adaptive
A low-trust stance is not automatically dysfunctional.
In an exploitative environment, skepticism may protect the person.
Members of groups with a history of institutional mistreatment may have good reasons to demand stronger evidence from authorities.
A system that labels such caution "epistemic mistrust" without examining context risks turning an institutional trust problem into an individual deficit.
This is one of the most important ethical boundaries for the construct.
The question is not simply how trusting a person is.
It is whether openness and caution are calibrated to the information ecology they inhabit.
Credulity can coexist with mistrust
At first this seems contradictory.
How can someone distrust and overtrust at the same time?
Because trust can be fragmented.
A person may reject mainstream institutions while accepting a charismatic outsider with almost no verification. They may assume most people are deceptive while granting complete authority to one favored source.
The clinical literature's separation of mistrust and credulity is useful here.
It reminds us that generalized suspicion does not guarantee discernment.
Sometimes it simply changes who receives unquestioned trust.
Institutions create the conditions under which trust becomes rational
Individual training can improve source evaluation.
But institutions also shape whether trust is deserved.
Transparent methods, visible corrections, accessible evidence, clear uncertainty and accountable conflicts of interest make epistemic trust easier to calibrate.
Opaque authority, hidden incentives and repeated communication failures do the opposite.
This changes the civilizational framing.
A healthy information system should not demand that people "trust experts."
It should build practices that make warranted trust inspectable.
AI systems complicate the source problem
A generative system can sound confident without possessing the kind of personal accountability people associate with a human source.
It can summarize evidence accurately, hallucinate a citation or blend high- and low-quality material into the same fluent tone.
Users may anthropomorphize it, distrust it reflexively or rely on it because its language feels coherent.
The epistemic-trust question therefore becomes unusual:
What exactly is the source?
The model? The developer? The retrieval corpus? The cited paper? The human who configured the workflow?
Trust needs to be decomposed across that chain.
Trust should be claim-bounded
A useful discipline is to make trust specific.
Not "I trust this person."
Instead:
I trust this source to have relevant access.
I trust this institution's measurement process for this type of data.
I trust this paper to support this narrow claim, subject to its design.
I trust this expert within this domain, not outside it.
Claim-bounded trust reduces the risk that prestige in one area spills into authority everywhere.
Calibration is more useful than maximum trust
The goal of epistemic development is not to increase trust as high as possible.
It is to improve calibration.
Reliable sources should receive more weight than unreliable ones.
Strong evidence should update confidence more than weak evidence.
New contradictory evidence should reopen a conclusion that had previously felt settled.
A person who trusts nobody cannot benefit from collective knowledge.
A person who trusts indiscriminately cannot protect themselves from manipulation.
Discernment lives between those failures.
Nordic and European relevance lies in institutional design
European traditions of public-service media, research transparency, ombuds functions, open science and administrative accountability provide useful settings for studying warranted trust.
They should not be romanticized as automatically high-trust systems.
Their value for the Institute is methodological: they make it possible to ask how institutional procedures create, lose and repair epistemic trust.
The UCL programme on epistemic trust also gives this construct a substantial European research base.
Regional relevance here is substantive, not decorative.
Trust is always trust about something
Epistemic trust becomes vague when it is treated as a global quantity attached to a person or institution.
A physician may be highly trustworthy about infectious disease and poorly positioned to interpret macroeconomics. A close friend may be honest and caring while having no expertise on the claim at hand. A public institution may be reliable about statistical definitions yet unable to know what happened in a particular household.
This means epistemic trust has at least four coordinates:
- source: who or what is providing the information;
- domain: what kind of knowledge is being claimed;
- claim: how specific and checkable the proposition is;
- stakes: what follows if the information is wrong.
The same source can rationally receive different levels of trust across these coordinates.
A mature trust architecture is therefore granular. It does not ask, "Do I trust this institution?" as though trust were one switch. It asks, "What does this source have reason and capacity to know here?"
Mistrust can be accurate without being sufficient
Research traditions focused on epistemic trust sometimes describe mistrust as a barrier to social learning.
That can be true.
It can also be adaptive.
People who have experienced unreliable caregiving, discrimination, institutional neglect or deception may have learned that openness carries cost. The fact that generalized mistrust later constrains learning does not make its history irrational.
This distinction matters ethically. Interventions should not begin by demanding that a mistrustful person become more trusting. They should ask whether the information environment has become more trustworthy and whether the person has usable ways to discriminate among sources.
In institutional design, rebuilding epistemic trust may therefore require behavioral evidence from the institution: transparency, correction, competence, responsiveness and accountability.
Trust cannot be repaired entirely inside the person who was asked to trust.
Measurement should distinguish relational state from general tendency
Questionnaires such as the ETMCQ and later revisions provide useful ways to study trust, mistrust and credulity as reported tendencies.
But the same participant may respond very differently to a teacher, partner, unknown website, physician or AI system.
Researchers should decide whether they are measuring:
- a relatively general orientation toward communicated knowledge;
- trust in a particular relationship;
- judgment of a specific source;
- willingness to act on a specific claim;
- openness to socially transmitted information under uncertainty.
Those are related, not interchangeable.
A good study makes the unit of trust explicit before selecting the measure.
Artificial agents make source competence harder to locate
When a generative AI answers a question, the visible source is not necessarily the epistemic source.
The response may depend on training data, retrieved documents, system instructions, model architecture and the user's prompt. The interface can sound singular while the knowledge chain is distributed.
This creates an unusual trust problem. A user can calibrate trust to the brand or conversational persona while having little access to the provenance of the particular claim.
One response is not permanent distrust. It is claim-bounded verification: ask for sources, inspect them, distinguish generation from retrieval, and lower confidence where the evidence chain is opaque.
Epistemic trust in AI may therefore need to be partly procedural rather than relational.
What the evidence does not show
The evidence does not show that epistemic trust is one universally agreed psychological variable.
It does not show that higher trust is always healthier or more accurate.
It does not justify diagnosing skepticism toward institutions as individual pathology without examining the history and quality of those institutions.
Current self-report measures do not capture every form of real-world source evaluation.
Clinical associations do not establish that disrupted epistemic trust is the sole cause of psychopathology or that restoring trust alone produces recovery.
And no trust measure can substitute for checking whether a source is actually correct.
What this changes for Inner Technology
Epistemic trust belongs in Inner Technology because no person can verify the world alone.
Human capability therefore includes knowing how to receive knowledge from others without surrendering judgment.
A practical reflection is not "Who do I trust?"
It is more precise:
What does this source actually know?
How do they know it?
Why is this information relevant to me?
What would make me reduce or increase confidence?
Am I rejecting the claim because the source is weak, or because I dislike what the claim would require me to reconsider?
Those questions turn trust from a feeling into an epistemic relationship.
The borrowed knowledge
The doctor finishes explaining the scan.
The patient cannot independently reproduce the imaging physics, the pathology training or the statistical evidence behind every sentence.
Some reliance is unavoidable.
But reliance does not have to be blind.
The source can show the image, name uncertainty, explain alternatives and invite a second opinion.
The patient can ask, compare and return with better questions.
That is the deeper architecture of epistemic trust.
Not belief without scrutiny.
The ability to learn from another mind while keeping one's own judgment alive.
Research references
- Sperber D, Clément F, Heintz C, Mascaro O, Mercier H, Origgi G, Wilson D. Epistemic vigilance. Mind & Language. 2010;25(4):359-393. https://doi.org/10.1111/j.1468-0017.2010.01394.x
- Fonagy P, Luyten P, Allison E. Epistemic petrification and the restoration of epistemic trust: A new conceptualization of borderline personality disorder and its psychosocial treatment. Journal of Personality Disorders. 2015;29(5):575-609. https://doi.org/10.1521/pedi.2015.29.5.575
- Fonagy P, Campbell C, Bateman A. Mentalizing, attachment, and epistemic trust in group therapy. International Journal of Group Psychotherapy. 2017;67(2):176-201. https://doi.org/10.1080/00207284.2016.1263156
- Campbell C, Tanzer M, Saunders R, et al. Development and validation of a self-report measure of epistemic trust. PLOS ONE. 2021;16(4):e0250264. https://doi.org/10.1371/journal.pone.0250264
- Fonagy P, Allison E. Beyond mentalizing: Epistemic trust and the transmission of culture. The Psychoanalytic Quarterly. 2023;92(4):599-640. https://doi.org/10.1080/00332828.2023.2290023
- Campbell C, Delamain H, Saunders R, Tanzer M, Milesi A, Nolte T, Allison E, Luyten P, Fonagy P. Development and validation of the Revised Epistemic Trust, Mistrust and Credulity Questionnaire (ETMCQ-R). BJPsych Open. 2025;11(5):e191. https://doi.org/10.1192/bjo.2025.10813