---
source_url: https://arxiv.org/abs/2606.04543
ingested: 2026-08-03
sha256: 7cb38b2d6394f9c82176001af7a881e0c247547a5b84f54dcbb23b225dbe06c3
---

# Agentic AI and Pedagogical Best Practice: The Tension Between Automation and Learning

Steve Woollaston, Brendan Flanagan, Isanka Wijerathne, Hiroaki Ogata. arXiv:2606.04543 [cs.CY]. Accepted at AIED 2026 — Festival of Learning, HAI-Agency Workshop on Orchestrating Human and AI Agency for Proactive and Reflective Learning. Submitted 3 Jun 2026.

## Core argument

AI in education is evolving from passive chatbots to **proactive AI agents** capable of initiation and goal-directed interaction. While offering personalisation opportunities, this shift risks undermining **learner agency and cognitive effort**. The paper reviews six pedagogical principles through the lens of agentic AI and proposes design recommendations that keep AI supportive of — rather than supplanting — human learning.

## Six pedagogical principles vs agentic AI

1. **Prior knowledge activation** — agents that pre-fetch/surface content can bypass the retrieval practice that activates prior knowledge
2. **Collaborative learning** — agent initiative can crowd out peer-to-peer interaction and role negotiation
3. **Problem-based learning** — goal-directed agents may resolve the problem before the learner grapples with it
4. **Formative assessment** — agent-generated feedback can pre-empt the learner's own self-assessment loop
5. **Scaffolding** — automated scaffolds risk being static/one-size-fits-all rather than dynamically withdrawn
6. **Metacognition** — agent initiation can displace planning, monitoring, and evaluation by the learner

## Design recommendations

- **Intentional friction**: deliberately insert moments of productive struggle rather than maximising smoothness
- **Dynamic scaffolding**: scaffolds that adapt and are removed as competence grows
- **Human-in-the-loop oversight**: learners (and educators) retain control over agent initiation
- **Considered AI utilisation**: purposeful, pedagogically-justified agent use rather than maximal automation

## Key tension

The tension between **automation and learning**: the more an agent does, the less cognitive work the learner does. Proactive agents must be designed so their initiative supports learner agency (e.g., by prompting reflection) rather than replacing it.