Research Summary: LearnAdapt Praxis: Controlled AI Assistance and Evidence Traces for Adult Workplace Learning

The LearnAdapt Praxis architecture addresses the challenge of assessing independent reasoning in adult workplace learning when AI assistance is utilized. It proposes a five-stage learning workspace (Frame, Learn, Build, Validate, Transfer) designed to provide controlled AI coaching and maintain an evidence trail of a learner's progression. This system aims to facilitate the production of plausible workplace artifacts while also allowing for inspection of the learning process through stored responses, versioned specifications, and coaching records, although it acknowledges limitations in detecting external AI use.

Why it matters

This research addresses the critical strategic challenge of integrating AI into learning and development without compromising the assessment of individual competency. It is crucial for organizations looking to leverage AI for productivity gains while ensuring that employees retain and develop essential skills and independent reasoning capabilities. This system could provide a framework for future AI-enabled learning platforms that prioritize verifiable skill acquisition.

Key insights

• AI assistance in the workplace can obscure a learner's independent reasoning and understanding.
• The LearnAdapt Praxis system offers a structured five-stage learning workspace: Frame, Learn, Build, Validate, and Transfer.
• The system records initial responses, versioned specifications, coaching interactions, validation observations, and a separate transfer response to trace learning.
• Project-level server controls can restrict AI coaching, access to previous work, and learner export during active transfer attempts.
• The architecture does not claim to prevent learners from seeking AI assistance outside of the application itself.
• The technical report describes the implemented architecture and evaluates selected aspects including workflow, authority, and failure recovery.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2610.02699