Scholarlink

<p>As artificial intelligence (AI) becomes increasingly integrated into workplace training, its role in delivering performance feedback, particularly negative feedback, has garnered growing interest. While prior research highlights AI’s potential to reduce the emotional discomfort associated with human-delivered critiques, emerging evidence suggests AI feedback can also demotivate. A key issue lies in perceived reliability: the black-box nature of AI complicates individuals’ ability to assess its reliability. Shifting focus from comparisons between AI and human feedback, this study draws on attribution theory to investigate whether and under what conditions higher levels of AI-delivered negative feedback motivate employees to engage in further learning. Using field data from tens of thousands of employee practice sessions at a global pharmaceutical company, we find that a higher degree of AI-generated negative feedback is associated with greater learning willingness. Yet, this positive effect diminishes when individuals encounter two observable front-end interface cues that weaken perceived reliability: processing error and output subjectivity. Two follow-up experiments confirm these results, revealing that processing error and output subjectivity decrease internal attribution and ultimately hinder feedback acceptance. By examining how front-end interface cues of AI systems influence reliability perception, this study enhances the understanding of AI feedback in organizational settings and provides actionable insights for designing AI coaching systems that promote learning motivation.</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

The promise and pitfalls of AI-generated negative feedback on learning willingness: the moderating role of AI reliability cues

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • Rongyu Zhong,
  • Yicheng Zhang,
  • Shengming Liu

摘要

As artificial intelligence (AI) becomes increasingly integrated into workplace training, its role in delivering performance feedback, particularly negative feedback, has garnered growing interest. While prior research highlights AI’s potential to reduce the emotional discomfort associated with human-delivered critiques, emerging evidence suggests AI feedback can also demotivate. A key issue lies in perceived reliability: the black-box nature of AI complicates individuals’ ability to assess its reliability. Shifting focus from comparisons between AI and human feedback, this study draws on attribution theory to investigate whether and under what conditions higher levels of AI-delivered negative feedback motivate employees to engage in further learning. Using field data from tens of thousands of employee practice sessions at a global pharmaceutical company, we find that a higher degree of AI-generated negative feedback is associated with greater learning willingness. Yet, this positive effect diminishes when individuals encounter two observable front-end interface cues that weaken perceived reliability: processing error and output subjectivity. Two follow-up experiments confirm these results, revealing that processing error and output subjectivity decrease internal attribution and ultimately hinder feedback acceptance. By examining how front-end interface cues of AI systems influence reliability perception, this study enhances the understanding of AI feedback in organizational settings and provides actionable insights for designing AI coaching systems that promote learning motivation.