An AI-supported E-mentoring model to develop EFL pre-service teachers' self-efficacy and emotional intelligence

Synthesis: Ismael, Luo, and Li (2026) test a 10-week AI-supported e-mentoring model with 50 second-year EFL pre-service teachers during their practicum in Egypt, pairing structured human mentoring, collaborative digital platforms, and AI-driven feedback via Gemini. The experimental group outperformed a conventional-practicum control group on both Self-Efficacy and trait emotional intelligence, with large effects, while interviews and weekly reflective logs show a move from control-oriented to student-centered, reflective practice. The authors frame AI as an augmentation of human mentoring rather than a replacement, attributing the gains to an integrated ecosystem of reflection, peer collaboration, and feedback.

Key Findings

Self-efficacy grew substantially more in the AI-supported group. Teachers using the e-mentoring model reported markedly stronger beliefs in their ability to manage classrooms, engage students, and apply instructional strategies, well beyond the modest gains seen in the control group.
The gains spread across all three self-efficacy subdomains. Student engagement, classroom management, and instructional practice all improved, signaling a broadened rather than a narrow sense of teaching competence.
Trait emotional intelligence rose as well. Teachers in the experimental group reported higher confidence and emotional competence, consistent with mentoring that deliberately worked on the affective side of teaching alongside the technical.
Teachers shifted from moderate to genuinely high perceived efficacy. They reported greater confidence in handling disruptive behavior, sustaining engagement, and teaching independently rather than leaning on scripts or supervisor direction.
Qualitative data show changed pedagogy. Four themes emerged: classroom management beliefs moving from "making students silent" toward rapport-building, a shift to student-centered engagement, growth in communicative and inductive teaching practices, and a joint development of self-efficacy and emotional intelligence.
Three mechanisms linked the outcomes — continuous reflective practice, collaborative peer learning, and AI-supported instructional and emotional feedback — which the authors argue produced a mutually reinforcing effect on confidence and emotional AI Regulation in Education rather than two separate developmental tracks.

Study Design & Method

An embedded mixed-methods design within a quasi-experimental framework anchors the study in a higher-education Professional Development context. Participants were 50 second-year EFL pre-service teachers at the Faculty of Education, Ain Shams University (Egypt), in the first semester of 2025-2026, placed across ten public schools. Randomization was conducted at the school level by lottery: five schools to the experimental condition and five to the control. All participants were female, reflecting gender-segregated practicum placement.

The experimental group received the AI-supported e-mentoring model across pre-practicum, during-practicum, and post-practicum phases: Google Classroom for weekly tasks and resources, Nearpod for interactive content and lesson modeling, Facebook as a collaborative mentoring space for sharing teaching videos and peer Feedback, Padlet for anonymous reflection and scenario analysis, and Gemini for personalized, context-sensitive feedback. The control group followed standard practicum procedures — school-based supervision by the department head and cooperating teacher, informal oral feedback, and no structured online platforms or AI tools. The intervention ran for about ten weeks, aligned to the practicum schedule.

Outcomes were measured with the 24-item Teachers' Sense of Efficacy Scale (TSES) and the 30-item Trait Emotional Intelligence Questionnaire (TEIQue), administered before and after the intervention, with good internal consistency and no meaningful baseline differences between groups. Effects were tested with mixed (group-by-time) analyses of variance, and the qualitative strand — pre- and post-intervention semi-structured interviews plus weekly reflective inputs — was analyzed thematically with a hybrid coding approach. Because the sample was set by availability rather than a priori power analysis, a sensitivity analysis indicated the design could reliably detect only relatively large effects.

Interview data trace a shift from wanting to avoid teaching out of fear of classroom challenges toward later reports of reduced anxiety — a change the authors read as emerging emotional regulation supported by guided reflection.

The theoretical Scaffolding combines Bandura's Social Cognitive Theory (mastery and vicarious experiences, social persuasion, emotional states), a Hattie-and-Timperley Feedback division of labor in which AI works at the task and process levels while human mentors address self-regulation and self-level concerns, and a trait-based conception of emotional intelligence. AI is explicitly positioned as augmenting, not replacing, the human mentor.

What this means for practice

Faculty developers. Run AI-supported e-mentoring as continuous infrastructure across pre-, during- and post-practicum phases rather than as an add-on tool; the 10-week experimental group outperformed the conventional practicum on both Self-Efficacy and emotional intelligence, and the gain came from one ecosystem of structured mentoring, collaborative digital interaction, reflective practice and AI support that bridged the theory-practice gap.
Faculty developers. Divide feedback labor deliberately: let AI carry task- and process-level feedback while human mentors handle self-regulation and identity-level work, keeping human oversight in the design.
Instructors. Develop confidence and emotional competence in the same activities rather than as separate tracks, since reflective cycles moved participants from control-oriented classroom management toward rapport-building, student-centered practice and the two dimensions reinforced each other.
Instructors. Treat AI literacy as entangled with emotional literacy: because Gemini was used to reframe emotionally charged classroom situations and design communicative activities, preparation should build the judgment to use AI suggestions reflectively rather than as verdicts.
Administrators. Target settings where supervision is thin: the authors present the model as scalable, context-sensitive professional development for under-resourced contexts with irregular, evaluation-focused mentoring, including Global South settings where mentoring resources are limited.

Limitations

Small, single-institution sample. Fifty pre-service teachers at one institution, which limits generalizability and did not permit a robust examination of the instrument's structural validity; the study was adequately powered only for large effects, so smaller but meaningful ones may have gone undetected.
Composition and stage. All participants were female, a consequence of gender-segregated school placements and a predominantly female English cohort, and all were second-year pre-service teachers at the preparatory stage, so the findings may not transfer to more advanced or in-service teachers or to other contexts.
Clustering and attribution. Randomization at the school level means participants within a school are not fully independent, and multilevel modeling is recommended for future work with more schools; the intervention was also multi-component, with a control condition that differed in being face-to-face and less structured, so improvements cannot be attributed to the AI tools alone.
Short intervention with thin data. A single semester restricts claims about long-term sustainability past the practicum, and some participants were initially hesitant to record or document classroom activities, slightly limiting the observational and reflective data.

Connected Concepts

Connected Articles

Artificial and Emotional Intelligence: Two Key Forces for Teachers' Professional Development in an Era of Uncertainty — AI and emotional intelligence as forces in teacher professional development
Human-AI Co-Mentorship in Project-Based Learning: A Case Study in Financial Forecasting — Human-AI co-mentorship designs in education
AI-Generated versus Human-Developed Assessment Tasks in EFL Context: Insights from TPCK Model — AI versus human assessment and EFL teacher knowledge
Self-Efficacy and Favorability Shape Learning from Tutoring Systems and Paper Practice — Self-efficacy in tutoring and learning support
Positioning Generative AI in EFL Peer Feedback: Training Feedback Literacy and Enabling Uptake in Speaking Classes — EFL peer feedback literacy
A Systematic Review of Emerging Technology Applications for Teaching English as a Foreign Language Across Different Educational Levels — Emerging technologies in TEFL
AI literacy-related domains and AI-TPACK readiness among preservice mathematics teachers: A factor-informed structural equation modelling study — AI-TPACK development in pre-service teachers
Conceptualizing pre-service teachers' readiness for AI integration into teaching practices: An intelligent-TPACK approach — Conceptualizing pre-service teachers' AI readiness
Gen-Mentor: A Human-in-the-Loop Instructional Framework for Dental Radiography Using Generative AI — AI mentoring in a professional practicum setting

Citation

Ismael, I., Luo, X., & Li, S. (2026). An AI-supported E-mentoring model to develop EFL pre-service teachers' self-efficacy and emotional intelligence. Frontiers in Psychology, 17, 1853510.

Metadata

Study design	Quasi Experiment
Discipline	Language Learning, English Education
Level	Higher Ed, Teacher Education
Intended audience	Faculty Developers, Researchers, Instructors
Learning and instruction	Self-Efficacy, Social-Emotional Learning
Research methods and evaluation	Mixed-Methods Research
Foundations of AI in education	AI Literacy
Page kind	Framework

Source History Edit Chat