Recent generative AI systems have demonstrated more advanced persuasive capabilities and are increasingly permeating areas of life where they can influence decision-making. Generative AI presents a new risk profile of persuasion due the opportunity for reciprocal exchange and prolonged interactions. This has led to growing concerns about harms from AI persuasion and how they can be mitigated, highlighting the need for a systematic study of AI persuasion. The current definitions of AI persuasion are unclear and related harms are insufficiently studied. Existing harm mitigation approaches prioritise harms from the outcome of persuasion over harms from the process of persuasion. In this paper, we lay the groundwork for the systematic study of AI persuasion. We first put forward definitions of persuasive generative AI. We distinguish between rationally persuasive generative AI, which relies on providing relevant facts, sound reasoning, or other forms of trustworthy evidence, and manipulative generative AI, which relies on taking advantage of cognitive biases and heuristics or misrepresenting information. We also put forward a map of harms from AI persuasion, including definitions and examples of economic, physical, environmental, psychological, sociocultural, political, privacy, and autonomy harm. We then introduce a map of mechanisms that contribute to harmful persuasion. Lastly, we provide an overview of approaches that can be used to mitigate against process harms of persuasion, including prompt engineering for manipulation classification and red teaming. Future work will operationalise these mitigations and study the interaction between different types of mechanisms of persuasion.

Machine learning predictive analytics (MLPA) are utilized increasingly in health care, but can pose harms to patients, clinicians, health systems, and the public. The dynamic nature of this technology creates unique challenges to evaluating safety and efficacy and minimizing harms. In response, regulators have proposed an approach that would shift more responsibility to MLPA developers for mitigating potential harms. To be effective, this approach requires MLPA developers to recognize, accept, and act on responsibility for mitigating harms. In interviews of 40 MLPA developers of health care applications in the United States, we found that a subset of ML developers made statements reflecting moral disengagement, representing several different potential rationales that could create distance between personal accountability and harms.