Can Humans cooperate with Artificial Intelligence? An Examination of Virtual AI Agent Experiments Using Crowdsourcing

• Published in: Socio-Informatics Vol. 12; no. 1; pp. 1 - 17
• Main Author: GOTO, Akira
• Format: Journal Article
• Language: Japanese
• Published: The Society of Socio-Informatics 30.06.2023; 一般社団法人 社会情報学会
• Subjects: Artificial Intelligent; Crowdsourcing; Intimacy; Public Goods Experiment; Social Value Orientation; クラウドソーシング; 人工知能; 公共財ゲーム実験; 社会的価値志向性; 親密性
• ISSN: 2187-2775; 2432-2148
• DOI: 10.14836/ssi.12.1_1

Artificial intelligence (AI) is steadily developing in today's society. Someday, AI will influence every aspect of our lives, and human–AI collaboration will be a significant concern. This paper aims to explore such cooperation via public goods game experiments with virtual AI agents and participants recruited through a crowdsourcing service.The findings revealed that trust and social value orientation toward humans did not impact the public goods game contributions. However, social value orientation toward AI did have an effect, and humans cooperated more with cooperative AI agents. Although general trust toward AI agents was higher than trust toward humans, prosociality and intimacy were still stronger toward humans. Interestingly, intimacy toward AI agents increased after engaging in public goods games with them.These results suggest that while individuals exhibiting cooperative behavior toward humans may not necessarily display the same behavior toward AI, social preferences still can apply to human-AI relationship studies. Furthermore, enhancing human-AI cooperation opportunities is thought to improve individuals' attitudes toward AI.