Beyond artificial encouragement: Why pre-service teachers value critical over positive feedback

Abstract

As generative artificial intelligence (GenAI) tools proliferate in higher education, critical questions arise about their effectiveness in providing assessment feedback, particularly in professional preparation contexts where feedback shapes future practice. This mixed methods study examined the quality and student experience of GenAI-generated versus human feedback among 15 first-year pre-service teachers. Using ChatGPT-4 to produce rubric-aligned feedback compared with expert human assessment, the results showed GenAI consistently awarded higher marks and revealed a “praise paradox”, where its uniformly positive tone undermined educational value by appearing excessive and developmentally limited. Student surveys showed a strong preference for human feedback across all quality dimensions, especially accuracy, actionability and professional contextualisation. Key themes included superficial praise concealing learning gaps, reliance on pedagogical expertise over algorithmic judgement and confusion from score disparities. While GenAI demonstrated strengths in personalisation and positivity, it lacked evaluative judgement and contextual understanding essential for teacher education. The study concludes that GenAI should complement, not replace, human expertise, advocating a human-in-the-loop model that preserves the relational and developmental dimensions critical to effective feedback in professional preparation programmes.

Implications for practice or policy:

• Teacher education programmes should not use GenAI to replace human feedback in primary assessment.
• Teacher educators should retain responsibility for evaluative judgement, professional context and developmental guidance.
• Course leaders may use GenAI to support formative feedback drafting, but only with human review and refinement.
• Institutions should establish transparent policies and professional learning so GenAI supports, rather than substitutes, relational feedback practice.