Impression transference from AI to human: The impact of AI's fairness on interpersonal perception in AI-Mediated communication

Jiahao LI, Yang CHU, Jie XU*

*Corresponding author for this work

Research output: Journal PublicationsJournal Article (refereed)peer-review

8 Citations (Scopus)

Abstract

Artificial Intelligence-mediated communication (AI-MC) is a new form of communication where AI algorithms modify, augment, or generate messages for its human principal. However, little is known about how AI-MC influences impression formation and interpersonal dynamics in the literature. This study examines how AI mediation affects impression formation in AI-MC using a moderated mediation model. We conducted an experiment with 66 participants who played the Alternated Repeated Ultimatum Game with a human-agent group. We manipulated the fairness of the AI agent and the entitativity of the human-agent group. The results show that the fairness of the AI agent influenced the participants’ impressions of both the AI agent and the human. This effect was mediated by the impression of AI and moderated by the entitativity of the group. The quantitative evidence shows that AI can influence impressions in AI-MC and suggests that modifying AI behavior can enhance social perceptions and relations.
Original languageEnglish
Article number103119
Number of pages11
JournalInternational Journal of Human Computer Studies
Volume179
Early online date25 Jul 2023
DOIs
Publication statusPublished - Nov 2023
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2023

Funding

This research was funded by the National Natural Science Foundation of China (grant number: T2192931 and 31800931 ).

Keywords

  • AI-mediated communication
  • Fairness
  • Impression formation
  • Trait transference
  • Ultimatum game

Fingerprint

Dive into the research topics of 'Impression transference from AI to human: The impact of AI's fairness on interpersonal perception in AI-Mediated communication'. Together they form a unique fingerprint.

Cite this