Multi-agent tutoring architectures show superior behaviour modelling and feedback compared to single-agent or rule-based tutors¶

Claim · All claims
Evidence · 1 study · review r3 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Multi-agent systems exhibit superior behaviour modelling, finer-grained individualised learning organisation, richer real-time feedback, and stronger metacognitive enhancement than traditional single-agent or rule-based tutoring technologies. → Simon Baradziej 2026

Evidence¶

Simon Baradziej 2026¶

Simon Baradziej. (2026). A systematic review of the roles of agentic AI in higher education. Social Sciences & Humanities Open. https://doi.org/10.1016/j.ssaho.2026.103418

q2 · i? · review · r3

The review attributes this comparative conclusion to Gupta and Heggond's (2026) review of multi-agent cognitive architectures for personalised adaptive learning, covering specialised agents for profiling, content selection, difficulty adjustment, and engagement monitoring. It is a review-reported comparison, not an effect-size finding.

"Their analysis demonstrates that multi-agent systems exhibit superior behaviour model - ling, more fine-grained organisation of individualised learning processes, richer real-time feedback, and stronger metacognitive enhancement capacities compared to traditional single-agent or rule-based tutoring technologies."

Discussion¶

Related Claims¶

• Most published studies of AI agents in higher education concern task-specific tools or single-agent systems, with only about 8% covering agentic multi-agent systems — related
• A constrained architecture combining prerequisite enforcement and a minimum cognitive-demand floor substantially reduces reward hacking, lowering RHSI from 0.317 (MO) to 0.102 (ST) — related
• Intelligent tutoring systems improve achievement over non-individualised instruction — a broader claim this one bears on
• Among deep learning KT models, simpler DKT performs comparably to complex architectures under limited dialogue data — related
• The RHSI condition ordering is robust across a 20-cell window-size and demand-floor grid, with EO strictly worst in all cells, though ST–MAS ordering reverses at demand floors ≥0.45 — related