Back

Categorising Coercion in Assisted Dying. An AI-Assisted Analysis of the House of Commons Debate on the 2025 Terminally Ill Adults Bill in England and Wales

Hamarat, N.; Gonzalez-Hijon, J.; Wels, J.

2025-09-18 health policy
10.1101/2025.09.12.25335636 medRxiv
Show abstract

BackgroundThe debate on voluntary assisted dying (VAD) in England and Wales has long centred on safeguarding autonomy and preventing coercion. MethodsThis study analyses parliamentary debates (2024-2025) on the Terminally Ill Adults (End of Life) Bill to explore how Members of Parliament (MPs) framed coercion risks in assisted dying legislation. Using AI-assisted text analysis of Hansard transcripts, we identify seven coercion types: economic (financial pressure), ethnic (healthcare inequalities), familial (family influence), medical (clinician bias), mental (psychiatric conditions), poor care (inadequate palliative services), and self-coercion (internalised pressure). ResultsThese concerns were transpartisan, with no party disproportionately emphasising any single coercion type. Poor care and mental coercion were most frequently cited, reflecting concerns over UK healthcare underfunding and psychological vulnerability. Network analysis revealed strong co-occurrence between medical and mental coercion, as well as familial and economic pressures, suggesting MPs viewed these as interconnected risks. However, debates largely focused on individual-level safeguards (e.g., capacity assessments) rather than systemic monitoring of structural inequalities. ConclusionsWe argue that effective VAD regulation requires a dual approach: individual protections alongside population-level oversight to address disparities in access and prevent coercion across socioeconomic and ethnic groups. This framework could inform future legislation in the UK and other jurisdictions considering assisted dying reforms.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.