Living Through the Data: A Qualitative Study of Wearable Symptom Tracking, Self-Management and Family Participation During Chemotherapy

Abstract

Background: Artificial intelligence is increasingly capable of contributing to information processing, risk estimation, and clinical decision support in oncology, but its acceptability depends not only on technological performance but also on how patients and families understand its role within clinical relationships. Qualitative inquiry is needed to examine how trust, disclosure, accountability, explainability, and autonomy are experienced when algorithmic recommendations enter preference-sensitive treatment decisions. Objective: To explore how adults with cancer and family caregivers perceive artificial intelligence in oncology decision-making and to identify conditions under which AI is viewed as supporting or constraining informed participation and patient autonomy.Methods: An interpretive qualitative study was conducted with 14 purposively sampled participants, comprising 10 oncology patients and four family caregivers. Semi-structured interviews used clinical scenarios addressing AI-supported recommendations, disclosure, clinician–algorithm disagreement, responsibility, and treatment choice. Data were analysed using reflexive thematic analysis. Results: Five interrelated themes were developed: conditional trust rather than technological trust; consent as an ongoing conversation; human accountability as the boundary of automation; explainability as practical meaning; and algorithmic pressure on autonomy. Participants generally accepted AI as an additional source of information when clinicians remained visibly responsible for interpretation and decision-making. Meaningful disclosure was considered increasingly important when AI materially influenced consequential treatment choices. Participants preferred practical explanations of relevance, uncertainty, and alternatives over technical descriptions. Algorithmic recommendations could facilitate deliberation by clarifying options but could also make disagreement appear irrational when presented as objectively definitive. Conclusion: Participants perceived AI as most acceptable when it augmented rather than displaced clinical judgement, preserved opportunities to question recommendations, and maintained identifiable human accountability. Implementation should prioritise proportionate disclosure, clinically meaningful explanations, explicit responsibility, and protection of patient choice