AI Chatbots Are Making Attention a Design Problem

Gloria Mark’s warning about AI chatbots is not that every prompt makes people less capable. It is that frictionless help can quietly replace the mental effort that attention, memory, and judgment need to stay strong.

Mark, a University of California, Irvine psychologist who studies human attention and technology use, told MIT Technology Review at SXSW London that people have “lost control” of their brains in the sense that digital environments keep pulling attention away from intentional focus. MIT Technology Review interview

The numbers she cited are stark: in her early work, Mark said people focused on one thing for about two and a half minutes in 2003; by 2012 that average had fallen to about 75 seconds; and in work conducted from 2014 to 2020, it fell to about 47 seconds. MIT Technology Review interview

The broader research record supports the idea that work has become highly fragmented. In a CHI 2005 field study of 24 information workers, Mark and co-authors found that people spent little time in working spheres before switching and that 57% of working spheres were interrupted. Mark, Gonzalez, and Harris, CHI 2005

Another Mark study found that interrupted workers could sometimes finish tasks faster without losing quality, but the speed came with higher stress, frustration, time pressure, effort, and workload. Mark, Gudith, and Klocke, CHI 2008

AI changes the attention problem

The chatbot twist is not just more notifications. Mark’s concern is that tools such as ChatGPT, Claude, and Gemini can take over the hard parts of reading, summarizing, evaluating, and writing. When users outsource that work too often, they may skip what psychologists call deeper processing — the active engagement that helps people understand and retain information. MIT Technology Review interview

That does not make AI useless or inherently harmful. It does mean the healthiest use of AI may look less like delegation and more like resistance training: ask for help, but still read the original, challenge the summary, rewrite the answer, and make the tool explain its reasoning instead of accepting the first polished paragraph.

The same rule applies to emotional life. MIT Technology Review reports that Mark is concerned synthetic companions could reduce the effort people put into real relationships, where empathy, patience, and misunderstanding are part of the work. MIT Technology Review interview

The goal is not abstinence

Mark’s prescription is not to abandon technology. In the interview, she said she loves technology and that people need new routines rather than a fantasy of going offline. MIT Technology Review interview

That is the practical takeaway for AI users. Use the chatbot to remove friction where friction is pointless: formatting, brainstorming, translation, first-pass organization. Keep the friction where it builds the mind: deciding what matters, checking the evidence, reading long enough to be changed by an idea, and talking to people without needing a bot to smooth every edge.

The danger is not that AI answers questions. The danger is that it can make thinking feel optional. Attention survives when users make effort part of the workflow on purpose.