AITentive: AI-supported Attentive User Interfaces (FWF) 

Human attention is a valuable resource, yet unlike the increasing power of the tech competing for it, attention is limited. Some view our “always-on” culture as akin to addiction, driven by platforms trying to maximize screen time. Today, many people juggle multiple conversations and media, a behavior intensified by digital notifications that constantly interrupt. This sequential multitasking incurs switch costs, as shifting focus demands time and effort. Such interruptions add stress, increase errors, and reduce performance, cutting productivity by up to 40% in workplaces and leading to severe risks, like driver distraction, in safety-critical situations. 

What if computer notifications, warnings, or updates arrive only when users are most receptive, ensuring minimal disruption? The project AITentive is developing AI-based systems to determine the best moments to interrupt users and schedule activities/tasks using Reinforcement Learning (RL) in combination with computational rationality (CR). These models learn optimal timing for notifications, managing attention in ways that help users transition smoothly between tasks. 

RL, a machine -learning method that does not need labeled training data, adapts flexibly to various problems and can learn “superhuman” multitasking strategies that are not restricted by the limits of individual cognitive abilities. Results obtained within the project have already shown that RL and CR can significantly improve multitasking effectiveness compared to human abilities alone. Beyond that, we are exploring how users’ work and learning behavior changes with AI collaboration, focusing on which attention management features users prefer for optimal productivity and minimal stress.