669146 event 1692892817 1693229946 <![CDATA[Ethics & Coffee: Does ChatGPT support confirmation bias? Designing a study]]> The way ChatGPT picks up the language used in the question-prompt both in the beginning of an answer and then again in the summary indicates that it takes the framing of the problem provided by the user for granted. Thus, it might not be able to react appropriately to the possibility that the user’s framing is inadequate or based on faulty assumptions. This leads to the more general question of whether ChatGPT and similar systems support confirmation bias.

Test this hypothesis in a discussion with Michael Hoffmann, professor of philosophy in the School of Public Policy. Hoffmann will present his ideas for designing an experiment, but will also open the floor to a discussion about how this might be done. Hoffmann is looking potential collaborators and would welcome your ideas and contributions.

Co-sponsored by the Ethics, Technology, and Human Interaction Center (ETHICx).

]]> A discussion about how to design an experiment to test whether ChatGPT supports confirmation bias.

]]> Michael Hoffmann
michael.hoffmann@pubpolicy.gatech.edu

]]> <![CDATA[]]> https://bit.ly/44nGjpW 1281 1289 1788 179080