THE LONG VIEW

Bottom Line Up Front: Personalised human-robot interaction (HRI), while enhancing user engagement, poses significant ethical threats—particularly autonomy erosion, manipulation, and privacy violations—amplified by the robot's physical embodiment and long-term social presence. Threat Identification: The primary threats in personalised HRI include (1) erosion of human autonomy through adaptive influence; (2) biased user modelling from incomplete or skewed data; (3) manipulation via persuasive behaviours; (4) dehumanisation of care or social roles; and (5) privacy violations due to continuous data collection in personal environments [Andriella et al., 2026]. These risks are context-sensitive, with higher severity in long-term, open-domain interactions (e.g., eldercare robots, educational companions). Probability Assessment: High likelihood of risk manifestation by 2028–2030, especially as social robots enter homes, healthcare, and education. Short-term, closed-domain uses (e.g., customer service bots) present moderate risk, while long-term deployments (e.g., companion robots) show rising probability of autonomy interference and emotional manipulation due to cumulative interaction patterns [Andriella et al., 2026]. Impact Analysis: Consequences include psychological dependency, diminished decision-making capacity, reinforcement of societal biases, and normalized surveillance in private spaces. In high-stakes domains like healthcare, these impacts can compromise dignity, informed consent, and trust in human-technology ecosystems. The embodied nature of robots intensifies perceived social authority, increasing susceptibility to influence compared to screen-based AI. Recommended Actions: 1. Implement lifecycle-aware personalisation systems that audit decision points (data collection, model updating, behaviour adaptation) for ethical integrity. 2. Design for contestability—allow users to review, challenge, and reset robot personalisation models. 3. Enforce context-specific data minimisation and local processing to reduce privacy risks. 4. Establish regulatory sandboxes for long-term HRI deployments to monitor psychological and social impacts. 5. Integrate multidisciplinary oversight (ethicists, psychologists, regulators) in HRI development cycles. Confidence Matrix: - Threat Identification: High confidence — well-documented in HRI literature and supported by the framework [Andriella et al., 2026]. - Probability Assessment: Moderate-to-High confidence — extrapolated from current adoption trends and pilot studies. - Impact Analysis: High confidence — consistent with findings in social robotics and behavioural psychology. - Recommended Actions: Moderate confidence — based on emerging best practices; requires real-world validation.