Abstract / Summary
There are over five-hundred-thousand individuals with recorded dementia diagnoses in England, highlighting an urgent need for scalable and effective care solutions to ease pressure on health and social care systems. Artificial Intelligence (AI)-enabled smart home systems are emerging as promising digital health innovations, offering cognitive support, real-time monitoring, and decision-making assistance. However, concerns around trust, user agency, and misconceptions about AI continue to limit acceptance and may hinder adoption. This study examined how stage of dementia care, system design, and level of system involvement shape public attitudes towards AI-driven care technologies. A repeated-measures design was employed, using vignettes that varied across three dimensions: stage of care (aligned with NHS England’s Dementia Well Pathway), system centrism (AI-based versus human-based), and level of involvement (moderate support versus full control). Participants aged 55–64 years – a group at elevated risk of developing dementia or encountering these technologies within the next decade – rated each scenario on both acceptability and perceived likelihood of actualisation/plausibility. Findings showed that acceptability was sensitive to both care stage and system type. A significant interaction indicated that full human involvement was consistently rated as more acceptable, while AI involvement was viewed more favourably only at moderate levels. This interaction effect intensified across the dementia care pathway, with the largest discrepancies observed in later stages such as Living well and Dying well, where full-control AI was rated least acceptable. In terms of perceived likelihood, scenarios were judged more likely under conditions of moderate involvement, with human-centric scenarios rated as more likely than AI-centric ones, particularly in later stages. These results highlight the importance of trust, autonomy, and public understanding in AI adoption. Acceptance was highest when AI was positioned to augment rather than replace human input, supporting hybrid models that preserve agency while enhancing scalability.