From Human to Machine: The Ethics of How AI Is Reshaping Data in Scientific Research

There is a large body of literature on the ethical issues linked to the use of AI applications; however, less attention has been given to the ethical issues linked to using AI in scientific discovery, despite these systems being increasingly used. A particularly underresearched and important area to investigate is the role and the workings of AI in data collection, as the quality of data fundamentally determines the quality of research. Scientific research has traditionally relied on data collected through human interaction. Increasingly, however, AI systems are transforming data collection across disciplines. This shift alters the epistemic nature of research: data are less shaped by human judgment and relational context, and more by algorithmic design and interaction constraints. Given the importance of data in scientific research, it is important to understand precisely which types of AI systems can be used at the data collection or generation stage and what the associated ethical implications are. Healthcare illustrates this trend, as AI systems can now gather clinical information that was traditionally obtained by clinicians through conversations with patients. AI can also generate new data and clean and categorise existing data. These outputs can then be used to feed databases for researchers, such as national biobanks, secure analytics platforms, or clinical record search systems. In some cases, AI-mediated data could be of higher quality than data collected or generated by human researchers. For instance, in psychotherapy, people may disclose more to AI chatbots than to humans, enabling more accurate and comprehensive data. Well-designed AI also has the potential to reduce biases that might arise in human judgement. Furthermore, using AI to “clean” and categorise data may contribute to more efficient scientific research. Yet, there is potential for significant ethical concern. Like AI applications used outside of the research context, issues of bias introduced through inadequate design may persist. AI systems used to collect data may lack the ability to acquire contextual understanding and thereby lead to incomplete or misleading data. Furthermore, research participants may struggle to trust AI systems or may be uncertain about how their data will be used. As AI becomes more integrated into everyday life, scientific research is progressively drawing on data acquired and/or managed through AI-mediated interactions. This shift transforms the nature of the data itself, affecting its structure, context, and interpretability, and may have significant implications for the validity, ethics, and reliability of research findings. The central challenge, then, is to understand how AI alters the very nature of research data and what this means for the knowledge claims that follow. This research project, therefore, asks: how does the use of AI change the nature of data used in scientific research, and what are the associated ethical consequences? Using an empirical bioethics methodology and drawing on medical research as a case study, this study will generate novel guidelines to support policymakers, funders, and researchers in promoting responsible scientific research.