AI-Mediated Communication： How the Perception that Profile Text was Written by AI Affects Trustworthiness

The paper explores the impact of AI-mediated communication (AI-MC) on online self-expression for the first time. It specifically investigates how the credibility of Airbnb hosts changes when their profiles are believed to be created by AI. The study conducted three experiments and observed a new phenomenon called the “Replicant Effect.” The following points elaborate on the paper’s key aspects:

1. Background and Objectives of the Study

   • Definition and Impact of AI-Mediated Communication
     • AI-MC refers to communication facilitated by algorithms that optimize, extend, or generate messages beyond traditional computer-mediated communication (CMC).
     • AI-MC has the potential to significantly alter individual self-expression and communication methods.
   • Airbnb and Credibility
     • Previous research has highlighted the significance of how the profile information provided by Airbnb hosts influences guest trust ratings.
     • This study examines how the evaluation of credibility changes when the host’s profile is believed to be written by AI.

2. Experimental Methods

   • Experiment 1: Overall Evaluation of AI-Generated Profiles
     • Objective: To compare the credibility evaluations when all profiles are believed to be generated by AI versus when believed to be written by humans.
     • Procedure: 527 participants evaluated Airbnb host profiles, divided into groups randomly believing all were “AI-generated” or “human-created.”
     • Results: Belief in whether profiles were generated by AI did not impact credibility evaluations.
   • Experiment 2: Evaluation in a Mixed Source Environment
     • Objective: To examine credibility evaluations in an environment where some profiles are believed to be AI-generated.
     • Procedure: 286 participants evaluated profiles in an environment where some were believed to be “AI-generated,” although all profiles were actually written by humans.
     • Results: Host credibility ratings decreased when participants believed profiles were generated by AI.
   • Experiment 3: Influence of Labeling and Priming
     • Objective: To investigate the impact of labeling and priming on credibility evaluations.
     • Procedure: 323 participants were divided into groups believing profiles were “all human-created” or “partially AI-generated.” They evaluated AI-like and human-like profiles under different labeling and priming conditions.
     • Results: Labeling and priming AI-like profiles led to lower credibility evaluations.

3. Significance and Conclusion of the Study

   • Replicant Effect: The study revealed that profiles believed to be generated by AI in a mixed source environment are less trusted, termed the “Replicant Effect.”
   • Theoretical Contribution: The study provides empirical evidence on the impact of AI-MC on credibility evaluations and suggests new research directions for CMC theories.
   • Practical Implications: The need for design guidelines and policies to ensure user trust in systems utilizing AI technology was implied.This research is an important step in understanding the impact of AI technology on communication. It provides valuable insights that will be useful for future research and practical applications.