GenAI Use Behavior and Post‐Failure Perceptions Among People With Functional Disabilities: A Multimethod Study

GenAI Use Behavior and Post‐Failure Perceptions Among People With Functional Disabilities: A Multimethod Study

Summary

This multimethod paper studies how people with functional disabilities (hearing, visual and mobility impairments) adopt generative AI (GenAI) and how they react when services fail. Two studies were run: Study 1 surveyed 284 participants to identify motivational drivers of GenAI use and to test whether perceived value moderates the intention–use link. Study 2 used an experiment (231 participants) to compare responses to service failures attributed to GenAI versus human officers and tested how attitudes and actual GenAI use moderate those reactions.

The main findings: habit, promotional benefits, trust and behavioural intention all increase GenAI use behaviour among people with functional disabilities. However, higher perceived value weakens the effect of intention on actual use (it acts as a decision gate). In failure scenarios, GenAI-attributed errors reduce inferential judgment (i.e. users judge future GenAI services more harshly) more than human errors. Positive attitudes and greater prior GenAI use buffer these negative reactions. The paper frames results using UTAUT2, behavioural reasoning theory, attribution theory, algorithmic transference and the social model of disability to highlight structural and experiential drivers of adoption and post-failure perceptions.

Key Points

1. Habit, promotional benefits, perceived trust and behavioural intention positively drive GenAI use behaviour for people with functional disabilities.
2. Perceived value moderates the intention→use relationship: higher perceived value can weaken the translation of intention into actual use.
3. Study 1 combined genetic-algorithm variable selection with Bayesian regression and IPWRA robustness checks on survey data (n=284).
4. Study 2 shows GenAI-attributed failures produce lower inferential judgement than identical human errors, signalling stronger distrust when AI is blamed.
5. Attitudes toward GenAI moderate post-failure reactions: more positive attitudes reduce harsh inferential judgements after GenAI failures.
6. Frequent GenAI users show greater resilience: prior use behaviour buffers negative generalisation after failures.
7. Findings are interpreted through the social model of disability: structural and design barriers shape expectations, trust and abandonment risks.
8. Practical recommendations include inclusive onboarding, transparency, targeted promotions and error-handling designed with disabled users.

Why should I read this?

Short version: if you build, buy or regulate GenAI services and care about inclusion, this paper tells you who actually uses GenAI among disabled users, why uptake sometimes stalls, and how a single AI cock-up can hurt trust — unless you design for it. It’s written with real methods (Bayesian models, experiments) and gives usable fixes: better onboarding, honest comms and co‑design with disability groups. Saves you time — and potential reputational pain.

Author style

Punchy: the authors treat people with functional disabilities as active agents (not passive recipients) and stress that perceived value can backfire when high expectations meet inaccessible implementations. If you’re in product, policy or marketing, the paper amps up why inclusion must be operational — not just rhetorical.

Context and Relevance

Why it matters: about 15% of the global population live with a disability; GenAI’s multimodal features hold strong promise for personalised accessibility. But this study shows that technology adoption among disabled users is shaped by structural barriers, prior experiences and expectations. The work is highly relevant to developers, accessibility teams, marketers and policymakers because it ties behavioural theory to practical interventions (training, transparency, accessible defaults and error recovery). It also extends debates about algorithmic distrust by showing disabled users’ responses are nuanced and moderated by attitudes and experience.

Source

Source: https://onlinelibrary.wiley.com/doi/10.1002/mar.70028?af=R