What Sets Us Apart? Identifying AI Enhanced Expertise in Human-Machine Interaction

Application summary

Artificial Intelligence (AI) has been insinuating into every aspect of our modern world for decades. With the recent release of highly fluent Large Language Models (LLMs), including deepfake technologies, the pace and range of AI’s integration into our lives has taken a quantum leap. With this leap, the importance of developing the knowledge required to ethically and effectively utilize AI, and to critically assess its impact on society is becoming widely recognized. My research focuses on how members of different expert groups distinguish between human and AI output. What kind of competences can be mimicked by a machine? What type of expertise is exclusively human? Focusing on expertise rather than intelligence, means conceiving human knowledge and skills as shared and maintained within social groups and acquired through socialization. I employ the sociological Imitation Game (IG) method. This novel experimental technique enables deep ethnographic probes into human cultures. In the IG, a participant playing as “judge” communicates with anonymized purported experts in a specific domain. The judge’s task is to discern genuine experts from laypersons imitating experts, or from AI portraying either role. The experiment illuminates how human and artificial expertise are expressed and recognized in interactions. One of the significant strengths of the IG is its empowerment of participants to act as proxy researchers, which ensures that knowledge is generated through the diverse voices of participants themselves. The plan is to conduct IG experiments with rock climbers, vegans and scientists, against LLMs, such as ChatGPT. These groups were chosen to delve into the known limits of AI: embodiment, morality and tacit knowledge. The objective of this research is to improve the human condition by charting how expertise emerges from human-machine interaction and ultimately, attend to the perennial question of what makes us human.