AI as an Epistemic Authority: do We Trust an Algorithm More than a Human?

Introduction. The article is devoted to the study of the phenomenon of AI as a new type of epistemic authority. The purpose of the work is to conduct a systematic epistemological analysis of trust in the algorithm as a source of knowledge and to identify its consequences for the cognizing subject and society. The relevance of the research is determined by the rapid spread of large language models (LLMs) as a universal adviser on issues that previously required consulting a human expert. Methodology and sources. The research is based on the conceptual framework of social epistemology, cognitive psychology and philosophy of technology. The study uses the concepts of rational trust in authority (Hardwig: competence and well-intentionedness as conditions of trust; Zagzebski: intellectual virtues), the phenomenon of automation bias from the field of cognitive psychology, as well as a critical analysis of the structural biases of AI systems. The research is theoretical in nature and is conducted through a consistent analysis of five theses on the epistemic status of AI. Results and discussion. It has been established that AI occupies the position of epistemic authority based on utilitarian effectiveness rather than subjective virtues. The illusion of objectivity and the automation b ias l ead t o a t endency t o t rust t he a lgorithm to t he detriment of human expertise. The fundamental difficulty of interpreting LLMs (the black box phenomenon) significantly limits the classical condition of rational trust – the possibility of independent verification of how a conclusion was reached. The systematic delegation of thinking to an algorithm leads to the atrophy of critical abilities and the deskilling of the subject. The mediation conception of AI as an aggregator of collective knowledge is considered promising, but limited due to the selectivity of data and dispersed responsibility. Conclusion. AI forms a fundamentally new and more fragile type of epistemic relationship. The answer to this challenge is “critical algorithmic literacy”, a systemic competence that becomes a prerequisite for maintaining cognitive autonomy in a society that delegates knowledge to an algorithm.

Authors: Alexander M. Zharov

Direction: Philosophy

Keywords: epistemic authority, artificial intelligence, social epistemology, automation bias, black box, algorithmic literacy, large language models


View full article