Overview
Human-Centred Guided Agency is a design principle for AI-supported learning and decision-making. It begins from the position that the purpose of educational AI should not be to remove productive struggle from the learner. It should help the learner see options, test reasoning, improve expression and make better choices while keeping responsibility for the thinking with the human.
Why agency matters
A system can produce a polished answer while leaving the user intellectually unchanged. In education that is a serious design failure. The relevant outcome is not only the quality of the product but the development of the person who produced it. Agency matters because learning requires choices, errors, revision, reflection and ownership.
Guided rather than automated learning
Guided agency favors prompts, choices, examples, feedback and next-step suggestions over automatic completion. The system can reduce unnecessary friction without erasing the cognitive work that builds competence. The learner remains the author, problem-solver and decision-maker.
AI as scaffold
AI can function as a scaffold when it helps a learner notice gaps, organize ideas, compare approaches, rehearse explanations or generate possibilities for consideration. The scaffold should be removable. If the learner cannot function without it, the design may have created dependency rather than capability.
Design principles
Human agency first. Make AI involvement visible. Preserve opportunities for judgment. Offer explanation rather than only output. Design for transfer beyond the tool. Respect privacy and local context. Build in moments where the learner must decide, justify or revise.
Application to writing and numeracy tools
This philosophy informs the direction of projects such as Next Move and StudyUp Studio. The purpose is not to claim empirical validation at this stage, but to make the design intention explicit: AI should help users move forward without taking over the thinking they need to develop.
Responsible AI implications
Responsible AI is not only a governance issue. It is also a product-design question. Systems can be safer and more educationally defensible when they are designed around human agency, bounded assistance, transparency and the development of capability.
Research questions
How much assistance supports learning before it becomes substitution? What forms of AI feedback increase independent transfer? How should guided-agency systems differ by age, discipline or cultural context? These questions form part of the continuing research programme.
This work forms part of the wider AI Masters Barbados scholarship programme. For lectures, seminars, institutional discussions or research collaboration, request a session.
Request a Session →