This seminar explores the implications of privacy and anthropomorphisation within human-AI interactions, emphasizing how AI technologies can influence our understanding of privacy and emotional connection.
We discuss why anthropomorphisation – the attribution of human-like qualities to AI – is likely to be widespread in contexts such as virtual therapy or elder care bots. Anthropomorphisation fosters rapport and comfort, encouraging individuals to share personal information more readily than they might with a non-anthropomorphic system. While this can enhance therapeutic outcomes, it also presents risks related to “privacy fatigue,” where the effort to maintain privacy becomes overwhelming, leading to the unintentional sharing of sensitive information.
The seminar addresses the tension between the clinical need for engaging and empathetic AI and the potential for manipulation. While friendly AI can improve patient outcomes, it also risks blurring the line between therapeutic support and privacy infringement, especially if third-party data access is involved.
Finally, we propose tentative guidelines for addressing these issues: respecting privacy norms aligned with user expectations, recognizing signs of privacy fatigue or inhibition, and ensuring transparency regarding the AI’s capabilities and third-party access to data.
We will explore these questions through examples, including well-known cases such as the “Alice” care bot in the Netherlands and other elder care bots. These cases illustrate both the benefits and ethical challenges posed by anthropomorphised AI in healthcare, showing how care bots can improve comfort and care quality but also complicate privacy boundaries due to their empathetic designs.

www.ethox.ox.ac.uk/events/anthropomorphisation-and-ai-in-the-healthcare-context