Epistemic humility

Working Definition

Epistemic humility is the disposition to notice the limits of one's own knowledge and competence, take those limits seriously, and let them govern practical epistemic decisions: how strongly to believe, whom to listen to, and what to count as proof of one's own error. In the technical sense established by virtue epistemology, it consists in proper attentiveness to one's intellectual limitations and “owning” them—acknowledging their existence, being willing to state them aloud, and adjusting one's behavior accordingly—while empirical psychology measures it as the metacognitive ability to recognize the fallibility of one's beliefs. In everyday talk it is routinely confused with diffidence, polite self-deprecation, or the relativist slogan that “everyone has their own truth,” none of which are cognitive virtues. Epistemic humility does not demand believing weakly; it demands believing in proportion to the evidence and knowing what one does not know.

Genesis and Global Lineage

An older lineage runs from Socratic awareness of ignorance in the Apology, through Nicholas of Cusa's “learned ignorance” (De docta ignorantia, 1440), to John Locke's discipline of proportional assent in the Essay Concerning Human Understanding (1689): humility as a rule for governing claims rather than a feeling of lowliness. The modern concept grew out of responsibilist virtue epistemology, inaugurated by Linda Zagzebski in Virtues of the Mind (1996) and developed by Robert C. Roberts and W. Jay Wood in Intellectual Virtues (2007), which devoted a full chapter to humility. Jason Baehr's The Inquiring Mind (2011) gave this movement a systematic theory of intellectual virtues as traits of character, and his joint paper with Dennis Whitcomb, Heather Battaly, and Daniel Howard-Snyder, “Intellectual Humility: Owning Our Limitations” (online 2015, print 2017), formulated the benchmark account: humility is proper attentiveness to, and owning of, one's intellectual limitations. In parallel, personality psychologists constructed validated measurement scales (Mark Leary's team, 2017), and the John Templeton Foundation made the science of intellectual humility a major funding priority. While psychologists and virtue epistemologists usually speak of intellectual humility, the broader term epistemic humility circulates across bioethics, political philosophy, and artificial intelligence safety, where it denotes a design norm for institutions and socio-technical systems rather than a personal trait alone.

Central European Lineage

Two thinkers formed in interwar Vienna articulated core elements of the concept decades before contemporary virtue epistemology, though without documented influence on the post-2011 literature. Karl R. Popper, who studied at the University of Vienna from 1918 and earned his doctorate under Karl Bühler in 1928, forged his fallibilism in the Vienna of Logik der Forschung (1934/1935) and, while in exile in New Zealand, defined the rationalist attitude in Chapter 24 of The Open Society and Its Enemies (1945) as admitting that “I may be wrong and you may be right, and by an effort, we may get nearer to the truth.” For Popper, humility is a readiness to invite and learn from public criticism rather than timidly holding back from bold claims. Friedrich A. von Hayek, who earned doctorates in law and political science at the University of Vienna and directed the Austrian Institute for Business Cycle Research (Österreichisches Institut für Konjunkturforschung, founded in 1927) before moving to the London School of Economics in 1931, closed his 1974 Nobel Memorial Lecture “The Pretence of Knowledge” with the warning that recognizing the insuperable limits to one's knowledge ought to teach the student of society “a lesson of humility.” In Hayek's hands, humility is an institutional and political principle: a safeguard against scientistic central planning that mistakes what is statistically measurable for the dispersed, local knowledge held by millions of individuals.

Why Now

The concept moved into the public spotlight when several crises simultaneously exposed the cost of inflated certainty. The replication crisis—symbolized by the Open Science Collaboration report in Science (2015)—forced the social and biomedical sciences to admit that a large share of published findings fail to replicate, embedding humility into research procedures via preregistration, open data, and registered reports. The COVID-19 pandemic (from March 2020) turned the communication of evolving uncertainty into a political flashpoint: public-health authorities who revised guidance as data arrived were accused of incompetence, while feigned certainty early on eroded long-term trust. Online polarization and outrage culture led philosophers (Michael P. Lynch, Know-It-All Society, 2019) and psychologists to investigate intellectual humility as a civic antidote; a landmark review by Tenelle Porter, Igor Grossmann, and colleagues in Nature Reviews Psychology (2022) synthesized a decade of empirical findings on its predictors and benefits. Finally, the deployment of large language models since 2023 gave the problem a sharp engineering form: the GPT-4 technical report demonstrated that reinforcement learning from human feedback degraded the model's probabilistic calibration, turning hallucinations into a statistical lack of epistemic humility.

Explanatory Power

The concept separates three things everyday language lumps together: how much someone knows, how confident they feel, and how they react when confronted with evidence of error. First, it explains Philip Tetlock's findings in Expert Political Judgment (2005): across 284 experts and more than 80,000 predictions, the best forecasters were not those with the deepest specialist knowledge, but the “foxes” who treated their own models as provisional hypotheses and revised their confidence after every outcome; raw expertise without humility actually worsened predictions because “hedgehogs” used their erudition to defend failing theories. Second, it shows why institutions, not just individuals, can be epistemically arrogant: Hayek's “pretence of knowledge” diagnoses any bureaucracy that optimizes only what fits its dashboard while ignoring tacit, dispersed realities on the ground. Third, it distinguishes verbal hedging from calibrated humility: a language model programmed to append “I might be mistaken” to every paragraph is not humble if its internal probabilities diverge from its error rates, whereas an expert speaking decisively can still be humble if they state in advance what evidence would change their mind.

Axes of Debate

Underappreciated Angle

The cheapest form of humility—saying “I could be wrong”—costs nothing and changes nothing. The costly form is an advance commitment: stating, before the data arrive, what specific outcome you will count as proof that you were mistaken. Study preregistration, Tetlock's Brier-scored forecasting tournaments, and Popperian falsification criteria are all epistemic humility written into institutional rules that cannot be quietly rewritten after the fact. Seen this way, humility is a property of systems rather than temperaments: an organization composed entirely of modest, polite individuals will still act with hubris if it lacks mechanisms that make error visible and costly to its author. The flip side is equally important: humility is not the minimization of confidence. A language model or analyst that becomes systematically underconfident after fine-tuning is just as poorly calibrated as an arrogant one; the virtue lies in matching confidence to evidence, not in self-deprecation.