Threat Modeling UX Research for Vulnerable Populations

WHY IT EXISTS: Traditional UX research assumes participants can read dense consent forms, advocate for themselves, and leave a session without social or economic penalty. That assumption breaks down for people with cognitive disabilities, trauma histories, low literacy, or severe economic need. When standard protocols ignore these asymmetries, they produce invalid data and can cause real harm.

THE MENTAL MODEL: Treat your research protocol like a threat model. Every standard method, think-aloud tasks, screeners, NDA signatures, and incentive payments, has a failure mode that can exploit a vulnerable participant. Your job is not to be charitable; it is to identify where power, comprehension, or physical ability gaps turn your study into a source of stress or coercion.

HOW IT WORKS: Start with a risk audit. Replace jargon-heavy consent with plain language or visual consent cards. Build multiple exit ramps so participants can stop without explaining themselves or losing compensation. Match incentive value to the effort and local cost of living, but keep it low enough that it does not become coercive for someone in desperate need. Use trauma-informed language, avoid surprise tasks, and allow support persons or caregivers at the participant's request. Debrief every session and provide referral resources. Finally, validate your findings with community advisors rather than treating individual participants as spokespeople for an entire group.

WHEN TO USE IT: Use these practices whenever your screener captures participants with diminished autonomy, such as people with intellectual or physical disabilities, older adults with cognitive decline, children, incarcerated individuals, or people in acute economic distress. Also apply them when the research topic itself is sensitive, such as domestic violence, addiction, grief, or mental health, because trauma can temporarily reduce anyone's capacity to consent or withdraw.

WHEN NOT TO USE IT: Do not use vulnerability as a reason to exclude diverse voices from your research. Do not apply a blanket label to an entire demographic; vulnerability is situational and intersectional, not an identity trait. And never let institutional review board paperwork substitute for genuine ethical care; compliance is the floor, not the ceiling.

ONE CANONICAL EXAMPLE: A team testing a banking app recruits adults with intellectual disabilities. The standard think-aloud protocol fails because literacy assumptions break task comprehension. The ethical adaptation uses pictorial consent cards, allows a support person present only at the participant's request, splits sessions into twenty-minute chunks to prevent fatigue, pays the same rate as non-disabled participants, and validates findings by member-checking with disability advocates rather than treating the participant as the sole source of truth.