Scholarlink

Purpose <p>Providing effective peer feedback plays an important role in collaborative learning and is an essential professional skill for students to develop. This study investigates whether generative AI can improve the quality of peer feedback, and whether it influences students’ perceptions of the feedback received.</p> Methods <p>An experimental design was employed involving 129 third-year Doctor of Pharmacy (PharmD) students. Participants were randomized into two groups: self-generated (SG) feedback and AI-assisted (AI) feedback. The SG group provided feedback independently (i.e., the feedback was crafted on their own, without the use of an AI prompt) using the task, gap, action (TGAP) framework. The AI group utilized generative AI to create feedback based on a given prompt that aligned with the TGAP framework. The prompt was partially completed, requiring students to add behaviors, professional skills (e.g., communication, interpersonal skills), or knowledge related to each student they were evaluating. Feedback was coded and analyzed, and students’ perceptions were measured using the Feedback Perceptions Questionnaire.</p> Results <p>A total of 353 peer feedback comments were analyzed (162 self-generated, 191 AI-assisted). The AI-assisted group achieved significantly higher median scores across all TGAP criteria (Task: H(1) = 27.32, <i>p</i> &lt; 0.001; Gap: H(1) = 89.32, <i>p</i> &lt; 0.001; Action: H(1) = 86.86, <i>p</i> &lt; 0.001). Specifically, 37.0% of students in the SG group compared to 61.3% of students in the AI group provided specific feedback on what their peers did well. For areas of improvement, 13.6% of students in the SG group provided specific areas to improve upon compared to 55.0% of students in the AI group. Only 22.8% of students in the SG provided feedback on how their peers could improve moving forward, compared to 72.8% in AI group. Student perceptions of feedback were positive, with only 3% of the AI group reporting negative feelings about the feedback they received compared to 12.7% in the SG group; this difference was not statistically significant (χ²(1) = 2.97, <i>p</i> = 0.085).</p> Conclusions <p>The study demonstrates how the strategic use of AI can be used to improve the quality of peer feedback. Future research should explore the long-term effects of AI-assisted feedback on students’ independent feedback skills and investigate behavioral changes resulting from improved feedback quality.</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

AI-ding peer feedback: a randomized study of self-generated vs. ai-assisted peer feedback

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月25日

查看全文
  • Zachary R. Noel,
  • Heewoo Lee,
  • Christina H. Sherrill,
  • Ian B. Hollis,
  • Jacqueline E. McLaughlin

摘要

Purpose

Providing effective peer feedback plays an important role in collaborative learning and is an essential professional skill for students to develop. This study investigates whether generative AI can improve the quality of peer feedback, and whether it influences students’ perceptions of the feedback received.

Methods

An experimental design was employed involving 129 third-year Doctor of Pharmacy (PharmD) students. Participants were randomized into two groups: self-generated (SG) feedback and AI-assisted (AI) feedback. The SG group provided feedback independently (i.e., the feedback was crafted on their own, without the use of an AI prompt) using the task, gap, action (TGAP) framework. The AI group utilized generative AI to create feedback based on a given prompt that aligned with the TGAP framework. The prompt was partially completed, requiring students to add behaviors, professional skills (e.g., communication, interpersonal skills), or knowledge related to each student they were evaluating. Feedback was coded and analyzed, and students’ perceptions were measured using the Feedback Perceptions Questionnaire.

Results

A total of 353 peer feedback comments were analyzed (162 self-generated, 191 AI-assisted). The AI-assisted group achieved significantly higher median scores across all TGAP criteria (Task: H(1) = 27.32, p < 0.001; Gap: H(1) = 89.32, p < 0.001; Action: H(1) = 86.86, p < 0.001). Specifically, 37.0% of students in the SG group compared to 61.3% of students in the AI group provided specific feedback on what their peers did well. For areas of improvement, 13.6% of students in the SG group provided specific areas to improve upon compared to 55.0% of students in the AI group. Only 22.8% of students in the SG provided feedback on how their peers could improve moving forward, compared to 72.8% in AI group. Student perceptions of feedback were positive, with only 3% of the AI group reporting negative feelings about the feedback they received compared to 12.7% in the SG group; this difference was not statistically significant (χ²(1) = 2.97, p = 0.085).

Conclusions

The study demonstrates how the strategic use of AI can be used to improve the quality of peer feedback. Future research should explore the long-term effects of AI-assisted feedback on students’ independent feedback skills and investigate behavioral changes resulting from improved feedback quality.