Co-Writing with Opinionated Language Models Affects Users' Views

Maurice Jakesch, Advait Bhat, Daniel Buschek, Lior Zalmanson, Mor Naaman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

47 Scopus citations

Abstract

If large language models like GPT-3 preferably produce a particular point of view, they may influence people's opinions on an unknown scale. This study investigates whether a language-model-powered writing assistant that generates some opinions more often than others impacts what users write - and what they think. In an online experiment, we asked participants (N=1,506) to write a post discussing whether social media is good for society. Treatment group participants used a language-model-powered writing assistant configured to argue that social media is good or bad for society. Participants then completed a social media attitude survey, and independent judges (N=500) evaluated the opinions expressed in their writing. Using the opinionated language model affected the opinions expressed in participants' writing and shifted their opinions in the subsequent attitude survey. We discuss the wider implications of our results and argue that the opinions built into AI language technologies need to be monitored and engineered more carefully.

Original languageEnglish
Title of host publicationCHI 2023 - Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems
PublisherAssociation for Computing Machinery
ISBN (Electronic)9781450394215
DOIs
StatePublished - 19 Apr 2023
Event2023 CHI Conference on Human Factors in Computing Systems, CHI 2023 - Hamburg, Germany
Duration: 23 Apr 202328 Apr 2023

Publication series

NameConference on Human Factors in Computing Systems - Proceedings

Conference

Conference2023 CHI Conference on Human Factors in Computing Systems, CHI 2023
Country/TerritoryGermany
CityHamburg
Period23/04/2328/04/23

Funding

FundersFunder number
Bavarian Research Institute for Digital Transformation
Bavarian State Ministry of Science and the Arts
National Science FoundationCHS 1901151/1901329
Studienstiftung des Deutschen Volkes

    Keywords

    • Co-writing
    • GPT-3
    • opinion change
    • risks of large language models

    Fingerprint

    Dive into the research topics of 'Co-Writing with Opinionated Language Models Affects Users' Views'. Together they form a unique fingerprint.

    Cite this