Using AI to give feedback to students can reduce performance anxiety, improving confidence and emotional resilience, according to a study led by Min He from Shaanxi Normal University, Xi’an, China. Gill Ragsdale explains.
Love it or hate it, there’s no doubt that AI in the classroom is here to stay and on the increase. Compared to the many studies on methodologies using AI, studies focussing on less tangible aspects, such as its influence on learner resilience, are comparatively rare.
Min He’s study looked at different kinds of AI feedback: corrective and motivational, and their impact on students’ self-reflection, creativity, performance anxiety and emotional resilience.
A survey using semi-structured and structured questions was given to 205 Chinese undergraduate EFL students. The strength of the relationships between the variables under study were then analysed using structural equation modelling. This model also took into account students’ familiarity with AI.
The resulting model showed that use of both types of AI feedback had a positive impact, especially on confidence and emotional resilience. Corrective feedback, related to grammar and vocabulary, had a stronger impact, while motivational feedback, i.e. encouragement and progress tracking, promoted self-reflection, enabling students to identify skills in need of improvement and setting appropriate goals for themselves.
Students became more relaxed about making mistakes and more inclined to take risks without fear of failure. This follows on from previous findings also showing the usefulness of the highly personalised support that AI can offer.
Min He recommends expanding AI capabilities with training on, for example, cultural nuances, idiomatic expressions and tone. While the results appear to be mediated by reduced performance anxiety, the influence, as measured in the model, only bordered on statistical significance and Min He suggests that might be improved by having, for example, a more conversational style of feedback. To further improve emotional resilience, AI could suggest personalised coping strategies such as mindfulness exercises.
Many of these recommendations seem to converge on essentially making AI more human-like and adds to the ongoing debate over the impact of AI on the role, employment and conditions of real human teachers.
Improved confidence and resilience are welcome impacts and the potential for further personalisation is impressive, even though it is interesting that motivational feedback was less impactful than corrective feedback. Could a more human like AI be as genuinely motivational as a real teacher? Could future studies compare the impact of an AI reporting ‘Well done! I really think you are making good progress, keep it up!’ with the same from an enthusiastic human teacher?
Similarly, it would be interesting to establish whether the confidence and resilience experienced using AI translates fully into real-world contexts – or does performance anxiety reappear when faced with real, diverse and unpredictable humans.
Reference
He, M., Abbasi, B. N. & He, J. (2025) AI-driven language learning in higher education: an empirical study on self-reflection, creativity, anxiety, and emotional resilience in EFL learners. Humanities & Social Sciences Communications 12: 1525 https://doi.org/10.1057/s41599-025-05817-5





