Translation is difficult.
From anthropometric data to product decisions.
/RESEARCH PROBLEM
Body measurements still leave choices about population, geometry and fit. Designers need to understand and explain how the evidence supports a product decision.
RESEARCH QUESTION
/WHY AI
Design requests vary. The support has to adapt.
AI can help clarify a request, find relevant evidence and coordinate calculations. The designer checks assumptions, chooses the design direction and retains approval.
/WORKED EXAMPLE
What does body data tell us about a chair?
A public waiting chair for Italian adults requires suitable evidence, an accommodation strategy and a traceable design decision.
01 / EVIDENCE
Choose the evidence
Population match and measurement coverage are separate choices.
Italian sample: 0 / 7 required measures. DINED 2004 proxy: 7 / 7. Record the population mismatch before calculating.
02 / ACCOMMODATION
Choose who it accommodates
Fixed, adjustable or multiple sizes?
Define the target and exclusions for a fixed chair; the range for adjustment; or the dimension relationships for size groups.
03 / DECISION
Translate and record
453 mm body reference → design decisions → fit check.
Allowance, final geometry and approval remain unestablished. The report image illustrates the intended output; its dimensions and approval stamps are not validated results.
A measurement becomes defensible when its source, assumptions and design decisions remain visible.
103 retained sources
53 anthropometry + design
50 AI + HCI
RAG supported the literature review and linked evidence tree.
Four research lenses
Is this the right data for these users?
Can this data support the product decision?
Can we still see how the decision was made?
How should AI guide and support the designer?
Where translation breaks
What the system must support
Population mismatch. A correct calculation may still represent the wrong users.
Deep dive into challenges and requirements
/DESIGN BRIEF
Ask
Clarify consequential gaps.
Calculate
Use defined numerical tools.
Warn
Expose mismatches and limits.
Explain
Connect evidence to decisions.
Record
Preserve assumptions and reasoning.
The designer confirms assumptions, chooses the direction and approves.
Build, test, learn, consolidate
AI proposes
Clarifies, retrieves evidence and calculates with defined tools.
Designer inspects and edits
Checks assumptions, evidence limits and the design direction.
Designer approves
Approval and physical validation remain explicit boundaries.
The project remains connected
/TECHNICAL IMPLEMENTATION
Next.js / CopilotKit / AG-UI · FastAPI / PydanticAI · Defined numerical tools · Typed artefacts and project memory
The thesis reports 120 API, 5 contract and 29 web tests for fixture-mode integration. These are reported technical checks, not new portfolio tests or evidence of a live deployment.
AI support was preferred; confidence was more divided.
01 / PREFERENCE
AI support was preferred in the task.
Brief A preferences favored AI support. The chart retains the thesis aggregates projected to N=8; these are bounded task preferences.
02 / CONFIDENCE
A useful answer still needs inspection.
Six coded responses asked for less density. Structure, visuals and report-style explanations helped; fluent AI output could still create false confidence.
03 / NEXT ITERATION
Make the next action clear.
Brief B and expert critique called for task-appropriate explanations: identify who a warning affects, what a measurement supports and what the designer should do next.
/RESEARCH ANSWER
/THESIS CONTRIBUTIONS — FOUR CORE PILLARS
01 / FRAME THE PROBLEM
Translation Gap
Defining the translation gap from anthropometric values to design decisions.
02 / EVIDENCE CHAIN
Sources To Design Req.
Insights · Challenges · Requirements · Prototypes · Validation.
03 / DEFINE RESPONSIBILITIES
Design Brief
Design requirements and 5 support behaviours dividing AI calculation and designer judgment.
/LIMITATIONS AND FUTURE WORK
“The thesis establishes the direction.”
[ ■ LIMITATIONS ]
01 / CONTROLLED INTEGRATION
Complete workflow needs to be tested and refined with live data. Final integration used controlled fixture data.
02 / SHORT AND BOUNDED STUDIES
Supports directional validity; does not yet support wider educational, clinical or industrial adoption.
03 / INTERFACE TO BE OPTIMISED
Interface design was not an integral part of the thesis; focus was the inspectable reasoning model.
[ ■ NEXT STEPS ]
01 / PRODUCTION READY PRO
Hardened, production-ready tool testing complete end-to-end workflows with live anthropometric datasets.
02 / TEST WITH REAL DESIGN PROJECTS
Longer longitudinal studies deployed in design education and professional industrial design practice.
03 / INTERFACE AND INTERACTION STUDIES
Optimize the interface and recommendation ergonomics based on user preferences and task walkthroughs.