Is Feedback In The Eye Of The Beholder? Chat Gpt Vs. Google Gemini In Giving Feedback On Students’ Oral Presentation Scripts
Is Feedback In The Eye Of The Beholder? Chat Gpt Vs. Google Gemini In Giving Feedback On Students’ Oral Presentation Scripts

Summary/Abstract: Aim. The paper aims to demonstrate how integrating innovative technologies like Chat GPT and Google Gemini can offer a transformative approach to giving feedback to students.Methods. This qualitative research explores the effectiveness of Chat GPT and Google Gemini in assessing students’ oral presentation scripts. The sample consists of 12 oral presentation scripts written by medical students in years 5 and 6 for their Medical English class. The scripts are ordinarily evaluated based on the rubrics including clarity, cohe- sion, medical terminology, grammar, and syntax. Also, the Medical case report bench- marks are considered in the evaluation. The identical assessment criteria were used as prompts for both Chat GPT and Google Gemini and the feedback produced by each tool was then compared manually. The feedbacks were analysed for comprehensiveness, sug- gestions, problems and tone to determine which tool was more suitable for this task.Results. The results indicate differences in the comprehensiveness and the focus of the assessment. Whereas Chat GPT relies for the most part on medical case report benchmarks and how they were/were not met by the students, Google Gemini analyses strengths, areas for improvement (clarity, concision, flow and transitions) and provides specific feedback by slide.Conclusions. The results can contribute to a better understanding of the capabilities and limitations of Chat GPT and Google Gemini in providing feedback on students’ oral presentation scripts. Furthermore, the findings can help educators select the most ben- eficial tool, with which they can provide personalised and immediate insights and thus support student learning and growth in medical English. Nevertheless, what is necessary for that potential to be adequately harvested with the lowest probability of error, bias and hallucination is quality-based prompt engineering and coaching on the part of the user.

• Details
• Contents