AI agents are not your “coworkers”
AI Summary: A study by Boston University professor Emma Wiles reveals that framing AI tools as "coworkers" rather than software tools leads to decreased performance in error detection, with participants catching 18% fewer errors when AI was perceived as an employee. The research indicates that this perception shifts responsibility away from human users, resulting in a 44% increase in the likelihood of escalating questionable AI outputs to management instead of correcting them. This finding raises concerns about the implications of treating AI agents as colleagues, particularly in critical sectors like healthcare and education, where it may lead to misplaced accountability for failures. The study highlights the need for careful consideration of how AI is integrated into workplace dynamics to avoid unrealistic expectations and potential negative outcomes.