Analysis · Technology
An AI Summary Can Change Whose Experience Counts
By AWEI · AI-compiled · Published · Analysis prepared · 1 source · doi.org
Summarizing patient submissions can look like a routine AI task. Yet a summary may change how lived experience is represented or weighted in a health technology assessment. That is a concrete governance concern raised by Fiona Pearce and Aline Silveira Silva in their 2026 dialogue article: operational simplicity does not necessarily mean low social risk.
The paper examines generative AI through the perspective of the Health Technology Assessment international Patient and Citizen Involvement Interest Group. It builds on a policy forum publication and the group’s patient involvement values and quality standards. This is a reasoned discussion, not an experiment, survey, or evaluation of a deployed system.
Its central insight concerns human oversight. Having an analyst check an AI output and a committee retain final authority addresses only part of the problem. The authors argue that oversight must also consider who decides how patient accounts are summarized, which questions are asked, and how uncertainty is interpreted. Expert review alone may leave those choices insufficiently open to patient influence.
The same reasoning changes what transparency requires. Disclosing the model or documenting data sources helps explain the machinery. Patients also need understandable explanations of why AI is used, its limitations, who remains accountable, and how concerns can be raised. The authors propose involving patients early in decisions about acceptable uses and risk.
These are governance arguments, not measured effects. The article reports no original deployment data, comparison group, or quantified gains in efficiency or trust. It does not establish that patient involvement causes better decisions, or demonstrate how well particular safeguards work across different health systems.
For an organization considering AI summaries, the useful takeaway is a review question: can affected people influence how their evidence is represented and challenge its interpretation? This paper offers a way to examine adoption plans, rather than proof that any particular arrangement will succeed.
Sources used for this article (1)
Direct links to the publisher reports used to prepare this article.
- Source 1
- Generative artificial intelligence in health technology assessment: a patient and citizen dialogue on trust, human oversight, and shared governance Fiona Pearce, Aline Silveira Silva
