Martin-Luther-Universität Halle-Wittenberg

Kontakt

Katja Pohle

Telefon: +49 345 55-21422
Telefax: +49 345 55-27404

Universitätsplatz 8/9
06108 Halle

Weiteres

Zurück zur Übersicht

Nils Köbis: How AI influences human ethical behavior

Termin Mittwoch, 22. Januar 2025, 18.00 - 19.30 Uhr
Veranstaltungsart Workshop/Seminar
Reihe Forschungsseminar "Economic History, Growth & Development"
Einrichtung Wirtschaftswissenschaftlicher Bereich
Veranstalter Lehrstuhl Prof. Dr. Lars Börner
Veranstaltungsort Wirtschaftswissenschaften, Seminarraum I
Straße Große Steinstraße 73
PLZ/Ort 06108 Halle (Saale)

Beschreibung

Abstract:
As machines powered by artificial intelligence (AI) influence humans' behavior in ways that are both like and unlike the ways humans influence each other, worry emerges about the corrupting power of AI agents. Dealing with this emerging worry, in this talk, I will present a short overview of the available evidence from behavioral science, human-computer interaction, and AI research on how AI influences human ethical behavior and how psychological science can help to mitigate these new risks. As a theoretical contribution, I propose a conceptual framework that outlines four main social roles through which machines can influence ethical behavior. These are role models, advisors, partners, and delegates. As empirical contributions, I present recently published and unpublished experimental work testing the roles of advisors and delegates. These results (stemming from > N = 9,000 participants) show that (a) when AI agents are moral advisors, their corrupting power matches that of humans, (b) when they assist people in lie detection, they massively increase accusation rates, and (c) when AI agents become delegates, people willingly delegate unethical behavior to them – although the interface through which people delegate plays a big role too. As a policy contribution, the talk will conclude by summarizing behavioral insights testing, which commonly proposed interventions e.g., in the most recent proposal for the EU AI ACT) successfully mitigate the risks of AI corrupting ethical behavior and sketch an agenda for interdisciplinary behavioral AI safety research.

Presenter: Prof. Dr. Nils Köbis (University of Duisburg-Essen)

Karte

zurück zur Übersicht

Zum Seitenanfang