About

Who I am, how I approach research, and the working principles behind both — plus a CV rendered from the same data as this site.

Bio

I'm XIANG JINWEI — an HCI researcher at Kyushu University, working on ageing, care, and AI.
I study how technology can hold memory and dignity for people who are usually designed around rather than designed for: AI-mediated reminiscence in dementia care, personalised memory environments for mild cognitive impairment, and clinical communication across language barriers.
I build my own research instruments — paper-reading workbenches, real-time translation tools, tiny macOS utilities — because the fastest way to understand a method is to build the tool that runs it.

Research approach

I work on ageing, care, and AI — with older adults, people living with dementia or mild cognitive impairment, and the people who care for them. These are people who are usually designed around rather than designed for, and that gap is where my research questions come from.

My default methods are participatory and qualitative. The things I care about — memory, dignity, connection — do not reduce cleanly to metrics, so I would rather sit with people in the settings where care actually happens than measure them from a distance.

Ethics is a design input, not a compliance step. Working with cognitively vulnerable participants means consent, dignity, and the cost of a system failing someone are constraints from the first sketch — not items for a later checklist.

How I work

01Primary sources first
Judgements are built on the actual paper, the actual data, the actual code — never on an abstract or a second-hand summary.
02Main problem first
Deal with what decides whether the work stands — validity, argument, risk — before touching wording, formatting, or polish.
03Direct, not agreeable
If something doesn't hold, I say so and propose a fix. Critique of the work stays separate from support for the person.
04Never fabricate
No invented citations, numbers, or results. When evidence is thin, the honest answer is “not enough evidence” — plus what it would take to know.
05Reuse before rebuild
Look for a mature, maintained solution before writing a new one. Building from scratch is the last resort, and it has to be argued for.
06Build the instrument
The fastest way to understand a method is to build the tool that runs it — so I build my own research instruments.
07Debug from first principles
Root causes come from what the data and control flow actually do, not from guessing and stacking patches.
08No silent fallbacks
Failures should be loud and visible where they happen. Quiet defaults and silent degradation are how bugs hide.

Curriculum vitae

xiang-cv.pdfRendered from the same data files as this site — always in sync.