Епістемологічні межі довіри та обґрунтування знання в неінтерпретованих системах екстрації даних

Abstract

This paper examines non-interpretable «black box» data extraction systems from an epistemological perspective, arguing that in a digital society such systems, though fundamental to knowledge production, challenge the classical definition of knowledge as «justified true belief» (JTB). While a «black box» may provide outputs that are true (accurate) and generate belief (the user trusts the data), it fails to supply the third crucial component: justification (logos). Without access to the system’s reasoning, the user lacks a rational basis for belief and must rely on faith in the algorithm. We conclude that interpretability is not merely a technical feature but an epistemological necessity if AI outputs are to be legitimately considered knowledge.