People Hold Smart AI Assistants Partially Responsible
A new study reveals that individuals attribute partial responsibility to AI assistants for outcomes, even when viewing them purely as tools.
Research from Ludwig-Maximilians-Universität München (LMU) indicates that people hold smart AI assistants partially responsible for outcomes, even when they perceive them primarily as tools. The study, involving 940 participants, examined how individuals assessed the responsibility of AI assistants used by human drivers.
The findings demonstrated an ambivalence: participants consistently viewed smart assistants as tools but also assigned them partial responsibility for both successes and failures experienced when a driver consulted them. This division of responsibility was not observed when participants evaluated a non-AI navigation instrument.
Furthermore, the study found that AI assistants were considered more responsible for positive outcomes than negative ones. The researchers suggest this might stem from people applying different moral standards for praise and blame, making it easier to assign credit to non-human systems than to attribute fault.
The research, led by Louis Longin at LMU's Chair of Philosophy of Mind, also explored whether the type of assistant (verbal or tactile) made a difference, finding no significant variation. The researchers argue these findings have implications for the design and social discourse surrounding AI assistants, urging developers to consider the impact on social and moral norms.