MACHINE GUESSING, PIGEON SUPERSTITION AND TRUST IN AI
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
CEP USM
Abstract
In this paper I argue that the predictive prowess of artificial intelligence methods, such as machine learning, is not sufficient as an epistemic warrant for us to allocate trust in them. This is because of two reasons: 1) they are opaque in ways that other methodology is not and 2) they fail in ways that other methodology does not. In order to motivate my analysis, I compare the prowess of some machine learning methodology in image recognition to the ability for pattern identification by some animals, such as pigeons, and suggest that while useful in certain practical circumstances, their success does not constitute the kinds of epistemic warrants that we should aim for in more epistemically demanding contexts such as science, medicine, and/or policy-making.
Description
ALVARADO, Ramón. Machine guessing, pigeon superstition and trust in AI. In: Şcoala internaţională de metodologie în ştiinţele socio-umane. Dezvoltarea personală și educația pentru societate: temeiuri epistemologice actuale: conferinţa știinţifică internaţională, ediţia a 4-a, 19-20 noiembrie, 2020. Chișinău: CEP USM, 2020, pp. 97-98.
Keywords
artificial intelligence methods, machine learning, epistemic warrant, pattern identification
Citation
ALVARADO, Ramón. Machine guessing, pigeon superstition and trust in AI. In: Şcoala internaţională de metodologie în ştiinţele socio-umane. Dezvoltarea personală și educația pentru societate: temeiuri epistemologice actuale: conferinţa știinţifică internaţională, ediţia a 4-a, 19-20 noiembrie, 2020. Chișinău: CEP USM, 2020, pp. 97-98.