Scholarlink

<p>AI integration in everyday decision making raises critical questions about how agency and affect are shaped when control is shared with machines. We investigated human–AI collaboration while choosing and executing a physical effort to earn rewards. AI was either involved in choosing (AI-choose) or executing (AI-execute) the effort. Both AI conditions reduced sense of agency (SoA), confidence, and happiness relative to baseline, with the effects being amplified following task failure. Across the two AI roles, SoA and happiness were lower in AI-execute, suggesting a greater negative impact when losing executional control. SoA predicted confidence and happiness, suggesting the importance of maintaining control in positive human-AI interaction. SoA could also be modulated by AI attitude. Participants with more favorable attitudes toward AI reported greater agency, even after failures. These findings highlight the need for collaborative AI systems that preserve human agency while maximizing the benefits of assistance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

When AI decides or acts: role-specific effects on human agency during decision making

  • 期刊论文

  • 发表日期：

    2026年6月22日

查看全文
  • Roya Mohammadsadegh,
  • Youngbin Kwak

摘要

AI integration in everyday decision making raises critical questions about how agency and affect are shaped when control is shared with machines. We investigated human–AI collaboration while choosing and executing a physical effort to earn rewards. AI was either involved in choosing (AI-choose) or executing (AI-execute) the effort. Both AI conditions reduced sense of agency (SoA), confidence, and happiness relative to baseline, with the effects being amplified following task failure. Across the two AI roles, SoA and happiness were lower in AI-execute, suggesting a greater negative impact when losing executional control. SoA predicted confidence and happiness, suggesting the importance of maintaining control in positive human-AI interaction. SoA could also be modulated by AI attitude. Participants with more favorable attitudes toward AI reported greater agency, even after failures. These findings highlight the need for collaborative AI systems that preserve human agency while maximizing the benefits of assistance.