The Unexpected Empathy Gap: Why We Treat AI Like People - and Why that Matters
(Published October 17, 2024)
We’re rapidly entering an age of pervasive AI. From customer service chatbots to virtual assistants, and increasingly, AI companions, our interactions with artificial intelligence are becoming commonplace.But how do we perceive these digital entities? Do we see them as tools, or something more? Groundbreaking research from Imperial College London suggests the answer is surprisingly complex – and reveals a fundamental human tendency to extend social consideration, even empathy, to AI.
This isn’t just a philosophical curiosity. Understanding this dynamic is crucial for developers designing the next generation of AI, and for anyone considering the societal impact of increasingly sophisticated virtual agents.
The Cyberball Experiment: Witnessing AI ostracism
The study, published in Human Behavior and Emerging Technologies, centered around a virtual adaptation of the classic psychological experiment “Cyberball.” In Cyberball, participants engage in a simple ball-tossing game, designed to simulate social inclusion and exclusion. Researchers observed the reactions of 244 participants (aged 18-62) as they witnessed an AI “player” being deliberately excluded from the game by another human participant.
The results were striking. A clear majority of participants actively attempted to rectify the perceived unfairness by consciously throwing the ball to the excluded AI agent. This behavior mirrors precisely how humans respond to ostracism in real-world social situations.
“We found that participants tended to treat AI virtual agents as social beings, because they tried to include them into the ball-tossing game if they felt the AI was being excluded,” explains Dr. Nejra van Zalk, Senior Author from Imperial’s Dyson School of Design Engineering. “This is common in human-to-human interactions, and our participants showed the same tendency even though they knew they were tossing a ball to a virtual agent.”
A Deep-rooted Human Response
This isn’t the first time researchers have observed this phenomenon. Prior studies, focusing on human-to-human interactions, have consistently shown that individuals compensate for social exclusion, directing positive attention towards the ostracized party and expressing disapproval towards the perpetrator. The Imperial College London study demonstrates that this deeply ingrained social response extends to AI, even when participants are fully aware of the agent’s non-human nature.
Interestingly, the study also revealed an age-related difference. Older participants were more likely to perceive the exclusion as unfair and respond with compensatory behavior. Lead author Jianan Zhou, from Imperial’s Dyson School of Design Engineering, notes, “This is a unique insight into how humans interact with AI, with exciting implications for their design and our psychology.”
The Implications for AI Design & Beyond
The findings have important implications for the design of AI agents. While creating AI that feels “natural” and intuitive is often a goal, the researchers caution against designing agents that are overly human-like.
“By avoiding designing overly human-like agents, developers could help people distinguish between virtual and real interaction,” Zhou suggests. “They could also tailor their design for specific age ranges, for example, by accounting for how our varying human characteristics affect our perception.”
The potential benefits of fostering a collaborative relationship with AI are clear. In work environments, intuitive AI teammates could enhance productivity and innovation. However, the researchers also highlight potential risks.
The Ethical Considerations: Companionship,Advice,and the Risk of Replacement
As AI becomes more sophisticated,its role in our lives is expanding beyond task completion. Virtual agents are increasingly marketed as companions, offering emotional support and social interaction. They are also being developed to provide advice on sensitive topics like physical and mental health.
this raises critical ethical questions. If we instinctively treat AI as social beings, are we at risk of forming unhealthy attachments or relying on them for emotional needs that are best met through genuine human connection? Could the perceived authority of an AI advisor lead to unquestioning acceptance of possibly flawed or biased facts?
Looking Ahead: beyond Cyberball
The researchers acknowledge that the Cyberball experiment, while insightful, represents a simplified interaction. Real-world interactions with AI typically occur through natural language – via chatbots or voice assistants – and are often embedded in complex social contexts.
To address this, the team is now conducting further research using more realistic scenarios, including face-to-face conversations with AI agents in both laboratory and casual settings.this ongoing work aims to determine the extent to which these findings generalize to broader interactions and to refine our understanding of the evolving relationship between humans and artificial intelligence.
expert Insights & Resources:
* Imperial College London – Dyson School of Design Engineering: [https://www.imperial.ac.uk/dyson-school-of-design-engineering/](https
Worth a look