A live web tool that coaches students through structured thinking on an assignment — planning, questioning and evaluating their own reasoning — instead of generating answers for them.
The Assignment Thinking Assistant is a live, publicly deployed web application I built to put my research principles into working software. Rather than producing text a student could submit, it is designed to scaffold the student’s own thinking about an assignment — prompting planning, questioning assumptions and structuring reasoning — so that AI assistance strengthens epistemic agency instead of substituting for it.
Try it live: assignment-thinking-assistant.vercel.app
Most generative-AI writing tools optimise for producing finished text quickly. That is exactly the failure mode my research is concerned with: it can displace the thinking assessment is meant to develop and certify. I wanted to test whether a tool built the other way around — optimising for the quality of the student’s thinking process, not the fluency of an output — was practically buildable and usable.
Designed and built the application end-to-end: interaction design, prompting strategy, and deployment.
Can an AI-assisted tool be designed so that its primary output is improved student reasoning — planning, self-questioning, evaluative judgement — rather than submittable text, while remaining genuinely useful enough that students would choose to use it?
Applied design research: building a working prototype, using it against real assignment prompts, and iterating the interaction flow based on where it either produced usable text (a failure mode to close off) or produced genuine thinking prompts (the intended behaviour).
TypeScript web application, deployed on Vercel; prompt design grounded in the Academic Integrity Decision Framework and the Responsible Human–AI Content Workflow developed elsewhere in this portfolio.
The assistant guides a student through an assignment brief by asking structuring questions — what is actually being asked, what counts as evidence, what the student already thinks and why — rather than drafting content on the student’s behalf. This mirrors the “AI assists, human authors” boundary at the centre of my academic-integrity research.
A documented, human-led workflow in which AI assists with drafting, organising, coding and formatting while humans remain responsible for meaning, accuracy, ethics and final approval.
Workflow design · LLM-assisted drafting · Verification checklists · Documentation
Last updated 28 July 2026
A six-question framework that structures fair, consistent decisions about student generative-AI use in assessment.
Conceptual synthesis · Case testing · Policy analysis
Last updated 28 July 2026