On this page

Synthesis: AICoFE (AI-based Collaborative Feedback) is a multi-Large Language Models (LLMs) feedback generation system for higher education that combines independently fine-tuned language models with teacher-in-the-loop mediation, producing diverse feedback perspectives while preserving pedagogical authority through Learning Analytics dashboards.

Key Findings

  1. Multi-LLM diversity improves feedback quality. AICoFE uses three independently fine-tuned LLMs (GPT-4.1-mini, Gemini 2.5 Flash, Llama 3.1) to generate diverse feedback perspectives from the same input data — quantitative rubric scores, validated qualitative observations, rubric level descriptions, and instructional materials. Each model produces independent drafts that the teacher can inspect and curate.
  2. Teacher mediation is central, not an afterthought. Rather than automating feedback end-to-end, the system positions educators as active curators. The teacher dashboard enables instructors to compose final feedback by selecting individual sentences or paragraphs from AI outputs, with a visual legend indicating the proportion of content contributed by each LLM. This preserves pedagogical judgment while reducing feedback workload.
  3. Dual-purpose transparency supports both practice and research. The system tracks the extent of teacher modification and curation, serving simultaneously as (a) a reflection tool for individual instructors examining their curation patterns, and (b) a dataset for large-scale analysis of the teacher mediation role in AI-assisted feedback.
  4. Role-specific dashboards close the feedback loop. The student dashboard provides access to video recordings of presentations, self-evaluation rubrics, visual comparisons of self vs. external evaluations, teacher-curated AI feedback, and the ability to rate perceived agreement and usefulness — completing a full assessment-for-learning cycle.

System Architecture

Module Purpose
Visualization Role-specific dashboards (student, teacher, evaluator)
Management Hybrid SQL + MongoDB data infrastructure (traceability + semi-structured feedback versions)
Feedback Generation Multi-LLM pipeline synthesizing rubric scores + qualitative comments
Recording Video/audio capture of student presentations (opt-in, GDPR-compliant)

The system treats AI as a draft generator, not a final deliverer — educators curate and compose the final feedback before it reaches students, so AI augments rather than replaces peer and teacher judgment.

What this means for practice

  • Students. Rate the AI-assisted feedback you receive for agreement and usefulness: those responses are the loop that tells instructors whether multi-model drafts add anything.
  • Students. Compare your self-evaluation against the external evaluation in the dashboard and use your presentation recording as evidence for that reflection.
  • Students. Treat the AI text you receive as teacher-composed feedback rather than raw model output, since sentences and paragraphs are selected and edited before delivery.
  • Students. Check the legend showing each model's contribution when a comment seems off-target, and say so in the embedded questionnaire.

Limitations

  • The reported deployment involves approximately 80 students and 5 teachers across one undergraduate and one master's course at Universidad Autónoma de Madrid.
  • Evaluation evidence is student-perceived coherence and usefulness of the feedback plus system usability scores; effects on feedback quality, reflective learning, or presentation performance are stated as future analyses.
  • Comparison between AI-mediated and traditional manual feedback is also deferred, so the paper cannot yet show that AICoFe improves on existing practice.
  • Presentation recording is opt-in and GDPR-compliant, so only consenting participants' data enter the pipeline.

Citation

Becerra, Á., Palma, A., & Cobos, R. (2026). AICoFe: AI-Based Collaborative Feedback System for Higher Education.

Embed this page

Copy the code below to embed a chromeless version of this page in a learning management system or other website. The embedded view hides the site header, navigation, and footer.