Mark Warschauer, hoogleraar aan University of California, Irvine, deelde vandaag een pre-print van een interessante studie waarin feedback werd gegeven door een getrainde leerkracht of door ChatGPT op teksten die door leerlingen waren geschreven.

Dit is de prompt die de onderzoekers gebruikten voor ChatGPT: “Pretend you are a secondary school teacher. Provide 2-3 pieces of specific, actionable feedback on each of the following essays…that highlight what the student has done well and what they could improve on. Use a friendly and encouraging tone. If needed, provide examples of how the student could improve the essay.”

Hun studie laat zien dat feedback van de getrainde leerkrachten op vier van de vijf door de onderzoekers gebruikte kwaliteitscriteria beter scoort dan de feedback van ChatGPT. Warschauer en zijn team concluderen ook dat de verschillen in kwaliteit van de feedback opvallend klein zijn.

Warschauer en zijn collega’s concluderen op basis van hun bevindingen dat ChatGPT (nog) geen vervanging is voor feedback van een leerkracht, maar dat leerlingen ChatGPT zouden kunnen gebruiken als aanvulling op feedback van een leerkracht.

Het abstract:

Offering students formative feedback on drafts of their writing is an effective way to facilitate writing development. This study examined the ability of generative AI (i.e., ChatGPT) to provide formative feedback on students’ compositions. We compared the quality of human and AI feedback by scoring the feedback each provided on secondary student essays (n=200) on five measures of feedback quality: the degree to which feedback (a) was criteria-based, (b) provided clear directions for improvement, (c) was accurate, (d) prioritized essential features, and (e) used a supportive tone. We examined whether ChatGPT and human evaluators provided feedback that differed in quality for native English speakers and English learners and for compositions that differed in overall quality. Results showed that human raters were better at providing high-quality feedback to students in all categories other than criteria-based. Considering the ease of generating feedback through ChatGPT and its overall quality, practical differences between humans and ChatGPT were not substantial. Feedback did not vary by language status for humans or AI, but AI and humans showed differences in feedback based on essay quality. Implications for generative AI as an educational tool are discussed.