AI evidence extraction

Augmenting dissertation mentorship through multi-modal generative AI: adapting language and visuals to diverse learning styles

Daren Scerri, Alan Gatt, Stephane Role, Gerard Said Pullicino · Artificial Intelligence in Education · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

7/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

1.76

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1108/aiie-05-2025-0085

Methodology & findings

Study design

Classroom experiment with a multi-modal GAI prototype evaluated at a vocational college where learners queried the AI mentor about research proposal forms, methodology assistance, and theory-context connections..

Main result

The study found that "51% of learners sitting for the pilot session preferred the kinesthetic mode, a finding in line with the college's vocational structure." Additionally, "Results for the AI-driven experiment confirm that increased accessibility, a reliable and/or verified data corpus and the AI mentor ability to explain with language adapted to each learning style are the three main features that distinguish the specialized prototype from generic chatbots like ChatGPT or Gemini."

Reports effect sizes.

Research paradigm

pragmatism

Author conclusions

The authors conclude that "while human interaction with the lecturer and/or mentor remains critical, the proposed AI solution augments mentorship provision beyond classroom hours and to a level of detail, which is difficult to achieve in traditional classroom or mentor meeting settings." They also note that "our work is distinct in applying a unified text-table-image pipeline to an educational tutor and combining this with learning style-based conditioning."

Risk of bias

Selection bias - sample from single vocational college may not generalize; Lack of control group comparison; Small pilot sample size not specified; No baseline measurements reported; No control group mentioned - single-group design with no comparison; Potential selection bias in vocational college participants; No information on blinding or randomization; Single institution evaluation may limit generalizability; Small pilot session sample size not explicitly reported in abstract; Selection bias: participants from single vocational college may not represent broader learner populations; Lack of control group comparison mentioned; Potential self-selection bias in learner participation in pilot session

Open questions raised

The authors note that while prior multi-modal RAG systems exist, their work is distinct in applying a unified text-table-image pipeline to an educational tutor and combining this with learning style-based conditioning.
The authors identify that prior multi-modal RAG systems exist but note their work fills a gap by specifically applying a unified text-table-image pipeline to educational tutoring while combining learning style-based conditioning.
The authors identify that prior multi-modal RAG systems (MuRAG and SAM-RAG) exist but lack application to educational tutoring combined with learning style-based conditioning.

Data: not_statedCode: not_statedExtracted from: pdfAgreement 66%

Explore related topics

AI Literacy Training Doctoral Training Design Science