Imagine stepping into your office only to learn that a new team member, named Alex, will be reporting to you. However, Alex is not a human colleague but rather an AI tool touted as an ’employee’ with designated roles and responsibilities. Recent research conducted by Emma Wiles, a business professor at Boston University, reveals that this framing may lead to diminished performance. Managers treating AI as a ‘coworker’ rather than a tool found themselves making 18% more errors in their assessments of work attributed to these so-called AI employees. This raises critical questions about the implications of how we label and integrate AI into our workplaces.
The tech industry is actively pushing towards a future where AI is envisioned as a type of ‘digital human.’ Since last year, major companies like Nvidia, Microsoft, and Google have introduced tools that position AI as team members, often with the cognitive abilities akin to human employees. Alarmingly, nearly a third of the 1,261 managers surveyed by Wiles indicated that their organizations already categorize AI agents as employees, with a surprising 23% including them in organizational charts. This shift in perception can lead to unrealistic expectations about AI capabilities, which could ultimately undermine human workers’ responsibilities and effectiveness.
Wiles’s findings suggest that framing AI as an employee can distort accountability, making individuals less inclined to trust their judgment when assessing AI-generated outputs. Participants were 44% more likely to refer questionable work to a manager instead of rectifying it themselves, negating the purported efficiency benefits of employing AI. This issue transcends mere office dynamics; as AI becomes embedded in sectors like healthcare, education, and defense, there is an increasing risk of misplacing blame on these technologies for failures that stem from human oversight and decision-making. Daron Acemoglu, a Nobel Prize-winning economist, argues that AI should enhance human capabilities rather than attempt to replace them, a perspective echoed by ongoing research at Stanford, which indicates that workers prefer AI assistance in specific areas rather than full automation. As we move forward, it’s crucial to recognize that while tools like Alex can offer support, they should never replace the agency and decision-making power of human employees.
Source: AI agents are not your “coworkers” via MIT Technology Review