Design Tacit Knowledge Capture

Design decisions disappear while the interface remains.

UI/UX designers make many small decisions while working: what to prioritise, what to compare, when to revise, and why one option is preferred over another. Much of this knowledge is tacit. It is expressed through action and judgement but may never be written down.

Retrospective interviews and think-aloud methods can help recover design rationale, but they rely on memory and require designers to explain their work after or during the task. Continuous questioning can also interrupt the activity being studied.

This research investigated another source of evidence already produced during digital design work: fine-grained Figma interaction traces. The aim was to examine whether those traces, combined with the design brief, could surface candidate design knowledge for the designer to review.

AI proposes. The designer decides.

The research purpose was to examine whether Figma interaction traces and a task brief could be transformed into candidate tacit-knowledge cards for expert review, and to compare the resulting material with retrospective think-aloud reflection.

The system does not claim to reconstruct a designer’s private thoughts. It uses observable actions and task context to propose possible goals, concerns, strategies, and reasoning. Designers then decide whether each proposal should be kept, edited, or discarded.

Turning an invisible process into a reviewable system.

I translated the research problem into a six-stage system that connects behavioural evidence with expert judgement. The prototype records fine-grained Figma actions and combines them with the design-task context. It groups related actions into design episodes and selects the most informative episodes for analysis.

For each selected episode, the language model generates a short candidate knowledge card. Each proposal remains connected to its source episode so the designer can inspect the basis of the inference.

Three decisions shaped the system.

Testing the workflow with professional designers.

Professional UI/UX designers with at least four years of experience completed a design task in Figma while the recorder captured their interaction traces. After a short break, each participant reflected through retrospective think-aloud, reviewed the generated cards, and completed a questionnaire.

The analysis compared reviewed cards with retrospective material. It examined what each candidate concerned, the goal or concern it described, the strategy it proposed, and the interaction evidence supporting it.

This evaluation treated quality as more than whether a sentence sounded convincing. It examined whether the proposal was supported, whether the designer accepted it, and whether relevant knowledge was missing.

A reviewable capture process, with clear limits.

The completed prototype turned fine-grained design interactions into evidence-linked candidate knowledge that professionals could inspect, correct, and reject. Its main contribution is the combination of episode-based capture, selective language-model interpretation, and structured expert review.

The comparison with retrospective reflection provides a way to examine both inaccurate proposals and missing knowledge. This makes human review part of the knowledge-capture method rather than a final approval step added after generation.