Sie können Bookmarks mittels Listen verwalten, loggen Sie sich dafür bitte in Ihr SLUB Benutzerkonto ein.
Medientyp:
E-Artikel
Titel:
3Es for AI: Economics, Explanation, Epistemology
Beteiligte:
Kaul, Nitasha
Erschienen:
Frontiers Media SA, 2022
Erschienen in:
Frontiers in Artificial Intelligence, 5 (2022)
Sprache:
Ohne Angabe
DOI:
10.3389/frai.2022.833238
ISSN:
2624-8212
Entstehung:
Anmerkungen:
Beschreibung:
This article locates its roots/routes in multiple disciplinary formations and it seeks to advance critical thinking about an aspect of our contemporary socio-technical challenges by bracketing three knowledge formations—artificial intelligence (AI), economics, and epistemology—that have not often been considered together. In doing so, it responds to the growing calls for the necessity of further transdisciplinary engagements that have emanated from work in AI and also from other disciplines. The structure of the argument here is as follows. First, I begin by demonstrating how and why explanation is a problem in AI (“XAI problem”) and what directions are being taken by recent research that draws upon social sciences to address this, noting how there is a conspicuous lack of reference in this literature to economics. Second, I identify and analyze a problem of explanation that has long plagued economics too as a discipline. I show how only a few economists have ever attempted to grapple with this problem and provide their perspectives. Third, I provide an original genealogy of explanation in economics, demonstrating the changing nature of what was meant by an explanation. These systematic changes in consensual understanding of what occurs when something is said to have been “explained”, have reflected the methodological compromises that were rendered necessary to serve different epistemological tensions over time. Lastly, I identify the various relevant historical and conceptual overlaps between economics and AI. I conclude by suggesting that we must pay greater attention to the epistemologies underpinning socio-technical knowledges about the human. The problem of explanation in AI, like the problem of explanation in economics, is perhaps not only, or really, a problem of satisfactory explanation provision alone, but interwoven with questions of competing epistemological and ethical choices and related to the ways in which we choose sociotechnical arrangements and offer consent to be governed by them.