Agents In The Classroom: A Multi-Agent AI Environment For Developing Critical Thinking, Digital Competence, And Professional Readiness Among Pre-Service Teachers

Abstract

This study investigated the effect of a Multi-Agent AI-based learning environment on pre-service teachers’ critical thinking, digital competence, and professional readiness for teaching. A randomized pretest–posttest controlled experimental design was employed. The participants were 100 fourth-year pre-service teachers enrolled in the General Education Program at the Faculty of Education, Damietta University, during the first semester of the 2022/2023 academic year. Following stratified random selection, participants were randomly allocated to an experimental group that used the Multi-Agent AI-based learning environment (n = 50) or a control group that studied the same Field Training course content through a conventional Moodle environment (n = 50). The 10-week intervention was delivered entirely online from October 1 to December 3, 2022, with one three-hour session each week. The experimental environment was implemented as a custom Moodle plugin connected to the OpenAI Completions API using the text-davinci-002 model. Its coordinated agents provided differentiated support for pedagogical planning, critical analysis, digital production, reflective practice, and professional decision-making. Data were collected using a critical-thinking situational test, a digital-competence performance test, a professional-readiness scale, a professional-readiness situational judgment test, a digital instructional product rubric, an agent-interaction log rubric, and an environment usability, trust, and satisfaction questionnaire. After controlling for pretest scores, the experimental group significantly outperformed the control group in critical thinking, F(1, 97) = 99.22, p < .001, η²ₚ = .506; digital competence, F(1, 97) = 71.99, p < .001, η²ₚ = .426; professional readiness measured by the scale, F(1, 97) = 156.08, p < .001, η²ₚ = .617; and situationally assessed professional readiness, F(1, 97) = 115.01, p < .001, η²ₚ = .542. The experimental group also achieved substantially higher digital product scores. These findings indicate that pedagogically differentiated and coordinated AI agents can strengthen the cognitive, digital, and professional dimensions of pre-service teacher preparation when embedded in authentic field-training activities.