EmpaText: An LLM-Powered Companion for Bias-Mitigated Project-Based Learning Support

Autores

  • Ahmed Si abdelhadi LRIAS-Artificial Intelligence and Systems Research Laboratory, University of Tiaret, Algeria
  • Abdelkader Ouared Computer Science Department, University of Tiaret, Algeria
  • Abdelhafid Chadli Computer Science Department, University of Tiaret, Algeria
  • Bendaoud Mebarek LRIAS-Artificial Intelligence and Systems Research Laboratory, University of Tiaret, Algeria

DOI:

https://doi.org/10.5935/jetia.v12i61.3799

Resumo

In Project-Based Learning (PjBL), psychological safety is essential for effective teamwork, while peer-review collaboration relies on strong interpersonal and communication skills. However, emotional reactions and cognitive biases such as impulsive judgments, confirmation bias, and decision fatigue can disrupt communication, weaken team dynamics, and compromise project outcomes. To address these challenges, we introduce EmpaText, an LLM-powered support system designed to enhance social awareness and foster communication, conflict resolution, and negotiation skills. EmpaText provides empathetic, context-aware, and constructive feedback to guide learners during peer-review activities, improving feedback clarity while reducing cognitive and emotional bias. Our framework comprises three core components: (i) Activity Enhancer, which improves learning deliverables; (ii) Review Facilitator, which guides assessment and promotes feedback quality; and (iii) Bias Mitigator, which employs linguistic markers and vocabulary analysis to detect and reduce cognitive and emotional biases. Through realistic PjBL scenarios and user-centered evaluations with 45 participants over 8 weeks, we show that EmpaText achieves a 60.9% reduction in cognitive biases, increases specific, actionable feedback from 32.7% to 71.1%, and enhances learning outcomes by 8.3-10.7 points. User satisfaction reached 86.7% positive responses, demonstrating the system’s effectiveness in fostering psychologically safe, bias-mitigated collaborative learning environments.

Downloads

Os dados de download ainda não estão disponíveis.

Downloads

Publicado

2026-09-30

Edição

Seção

Articles

Como Citar

EmpaText: An LLM-Powered Companion for Bias-Mitigated Project-Based Learning Support. (2026). ITEGAM-JETIA, 12(61), 512-528. https://doi.org/10.5935/jetia.v12i61.3799

Artigos mais lidos pelo mesmo(s) autor(es)