Dissertation Lara Bernasconi

Abstract

This PhD investigates the ethical, legal, and practical implications of introducing artificial intelligence (AI) into clinical trial management, with a particular emphasis on recruitment processes and natural language processing (NLP) applications. The overarching research question guiding this work was: How can we ethically harness the potential of AI to empower clinical research management? To address this question, the thesis examined which AI applications in clinical trial management are perceived as both relevant and ethically acceptable by the public, researchers, and other professionals, and analysed the implications of these applications for research participants’ rights, with particular attention to issues of equity and autonomy.

The research was structured into four projects, each progressively narrowing the scope from general attitudes to specific applications, moving from theoretical exploration to practical use cases. This approach ensured a coherent progression and balanced breadth with depth in examining AI in clinical trial management. Project 1 mapped current uses of AI in clinical trial management and assessed professionals’ attitudes toward these tools. Project 2 focused on NLP in patient recruitment, combining a scoping review with stakeholder interviews. Project 3 explored the development of a large language model (LLM) trained on Swiss electronic health records, analysing its ethico-legal challenges through a case study embedded in a national AI initiative. Project 4 examined professional and public acceptance of LLM-supported communication with research participants, particularly during recruitment. It also examined how individuals from diverse backgrounds responded emotionally to AI-generated versus human-written patient information, focusing on perceived clarity, empathy, and overall impact.

Methodologically, the thesis employed a mixed-methods design, combining literature reviews, surveys, stakeholder interviews, and case study analysis. The findings across projects indicate that, although stakeholders broadly acknowledge AI’s potential, its adoption in clinical research remains limited. The research showed that inadequate regulation, low AI literacy, and unresolved ethical concerns pose major barriers to responsible implementation. Applications such as AI-driven participant screening, chatbots, and synthetic control arms were seen as particularly sensitive, given risks related to unreliability, privacy, and transparency.

The thesis identifies blind spots in the current discourse, where ethical dimensions of AI in clinical trial management are often underrepresented relative to technical considerations. Furthermore, it demonstrates the importance of stakeholder engagement to uncover hidden risks and challenge assumptions underlying AI adoption. Finally, it offers normative guidance for understanding and shaping how AI may transform clinical research. The thesis argues that AI in clinical trial management should not be viewed merely as a technical innovation but as a sociotechnical development shaped by governance, trust, and institutional practices. By combining empirical studies with ethical and legal analysis, it provides actionable insights for policymakers, clinical researchers, and healthcare institutions seeking to harness AI responsibly in support of clinical research.

Lara Bernasconi