Mapping Moves, Modes and Methods: Designing Socratic Conversational Agents for AI-Enhanced Learning

Abstract

Conversational agents powered by Large Language Models (LLMs) are increasingly proposed as scalable tools for personalised learning support. Yet much existing research focuses on algorithmic capability rather than the nuanced human learning dialogues that shape educational practice. This leaves a gap in empirically informed frameworks for translating rich instructional conversation into actionable design principles for Socratic AI partners. This paper addresses this need through a secondary analysis of live, online design critiques conducted at a South African university - an environment reflective of many Global South contexts marked by resource constraints, student diversity, and socio-economic pressure. Building on the author’s doctoral research, the study synthesises previously collected empirical material, including surveys, a focus-group interview, and recorded critique sessions. A composite conceptual lens (Conversation Theory, Experiential Learning Theory, and Cognitive Apprenticeship) guided the interpretive analysis. The findings identify four recurring student–tutor relationship archetypes and four interaction dimensions that position critiques as formative, iterative, formal, and immersive. These insights are consolidated into a “moves–modes–methods” matrix that captures how knowledge is negotiated, feedback is scaffolded, and agency is fostered in the critique space. Mapping this matrix onto current scholarship on LLM-based tutors reveals both alignments, such as the value of probing questions, and tensions related to contextual sensitivity, including bandwidth limitations, student diversity, and socio-economic realities. By integrating detailed empirical insight with emerging work on AI-supported learning, the study offers an evidence-based framework to inform the design of conversational agents that augment human expertise while preserving the pedagogical integrity of the online critique in under-resourced, highly diverse settings.