The Information Framing Effect of “AI Unemployment”

Subjects: Psychology >> Social Psychology submitted time 2024-05-03

WU, Jiahua ZHANG, Yuyan XU, Liying YU, Feng

Abstract： The advancement of artificial intelligence (AI) technology significantly contributes to enhancing productivity; however, concerns regarding potential technological unemployment have garnered considerable attention. The uncertainty surrounding the occurrence, timing, and scale of AI-induced unemployment impedes definitive conclusions. This uncertainty may also lead the public to be influenced by encountered information concerning AI-induced unemployment. Media coverage on AI-induced unemployment often presents extensive information regarding affected industries, occupations, and probability scales, establishing two numerical information frameworks: one emphasizing factors influencing unemployment distribution across industries and another emphasizing the probability of unemployment occurrence. Comparatively, the probability framework, as opposed to the factor framework, allows individuals to formulate judgments indicating a reduced likelihood of AI-induced unemployment, thereby mitigating the perceived threat of AI, especially among individuals with high ambiguity tolerance. Building upon the foundational assumption that the probability framework alleviates AI threat perception, this study, comprising seven recursive experiments, investigates the mediating role of judgments on AI-induced unemployment likelihood and the moderating role of individual ambiguity tolerance. Experiment 1 juxtaposes AI threat perception elicited by general AI-induced unemployment descriptions, factor frameworks, and probability frameworks. Experiment 2 validates the mediating role of likelihood judgments. Experiments 3 and 4 respectively eliminate potential influences of probability values and unemployment scale. Experiment 5 explores ambiguity tolerance’s moderating effect. Experiments 6 and 7 examine subsequent AI threat effects, including support for AI development policies and willingness to recommend various occupations. The primary findings are as follows. Firstly, introducing AI-induced unemployment through a probability framework effectively diminishes AI threat perception (Experiments 1-7). Secondly, this effect is mediated by perceived likelihood, whereby the probability framework prompts individuals to form judgments indicating decreased AI-induced unemployment likelihood, thus reducing AI threat (Experiments 2-5). Thirdly, the information framework effect is moderated by ambiguity tolerance, primarily manifesting among individuals tolerant of ambiguous information (Experiment 5). Fourthly, individuals influenced by the probability framework demonstrate increased support for policies related to AI development, with AI threat playing a mediating role (Experiment 6). Lastly, individuals influenced by the probability framework exhibit a heightened willingness to recommend jobs involving frequent AI interaction (Experiment 7). This study extends prior research by elucidating how external factors such as information frames contribute to variations in AI threat perception. Unlike the extensively studied valence information frame, numerical information frames impact AI threat perception by altering individuals’ likelihood judgments. Our findings shed light on the effects of the numerical information framework on AI-induced unemployment threat perception, policy support, and job recommendation willingness.

Peer Review Status:Awaiting Review

 Hits 766  Downloads 175  Comment 0