UXmatters has published 5 editions of the column AI Minds.
In 1968, social psychologists John Darley and Bibb Latané published a landmark study. Participants had sat alone and heard what sounded like someone in an adjacent room having a seizure over an intercom. When participants believed that only they were aware of the medical emergency, 85 percent went to help. However, when they believed four other people also knew about it, only 31 percent helped.
Thus, Darley and Latané discovered and documented the phenomenon of diffusion of responsibility. While responsibility persists, it spreads out until it becomes thin enough that nobody claims specific ownership of it. Thus, when multiple people are nearby and possess the ability to act, each individual feels a lower sense of personal responsibility for taking action. Read More
Welcome to my new UXmatters column AI Minds. The design conversation around artificial intelligence (AI) often divides into two camps: technical capabilities and user-interface (UI) layouts. AI Minds will fill that gap. While UX designers typically focus on model parameters and prompt boxes, our most critical design work revolves around the people on the receiving end of an AI’s outputs. Now, more than ever, UX professionals must prioritize understanding users and how they navigate AI technology that is inextricably linked with high expectations, fears, and confusion. In my monthly column, I’ll explore the psychology of human-AI interactions, examining how cognitive biases, behavioral patterns, and psychological concepts shape the way people adopt, trust, resist, or oversee autonomous systems.
I’ll ground these discussions in established research—drawing from behavioral economics, cognitive psychology, and human factors—and apply these learnings directly to the design friction that UX professionals observe every day. Whether an incident manager is reviewing a database alert or a consumer is evaluating a suggested price, human judgment remains the primary safety check on the decisions software makes. In this installment of AI Minds, I’ll examine how anchoring operates in AI-assisted workflows, when it helps and when it harms, and what UX researchers and designers can do to preserve the independent, human judgment on which AI oversight depends. Read More
A client once asked me to help diagnose why a Sales team was not adopting an artificial-intelligence (AI)–assisted quoting tool that they had deployed six months earlier. The tool was accurate. The user interface was clean. They had provided training. But usage numbers were low—and those that weren’t low were superficial. People were opening the tool, then closing it without acting on its recommendations.
When I interviewed members of the Sales team to find out why, their explanations varied. One sales representative said the AI didn’t understand her customers. Another said he didn’t trust a system that he couldn’t see inside. A third said she had been doing quotes for eleven years and knew her accounts better than any algorithm. A fourth said he had caught the AI making a pricing error in week two and hadn’t fully trusted it since. Although the users’ feedback seemed to describe four different problems, there was actually just one problem that surfaced at four different points in the same task sequence. Read More