How phone habits could predict a stressful day, study finds

Your phone might know you are having a rough day before you admit it. Researchers at KAIST, the Korea Advanced Institute of Science and Technology, have built a system that reads stress from ordinary phone habits: how often you type, how you move between screens, how frequently you chat, search, or pay for things. According to the research team, the tool could become an early warning system for days when stress is piling up, and the university announced the findings on October 15. In short, phone habits may carry the signal long before a mood journal would.

The technology is called SHIRBT-MMF, and it does not care which apps you open. Earlier stress-tracking tools counted screen time, app launches, or usage by category, which breaks down the moment someone switches from one messaging app to another. This system asks a different question: what is the person actually doing? Chatting on KakaoTalk, LINE, or WhatsApp all count as the same routine, which the team labels "messenger chatting." Web browsing, financial payments, and reading or writing email get the same treatment. The team defines these routines as smartphone human interaction routine behavior tasks, mines them with a pattern-finding technique called multi-level sequential pattern mining, and then uses a large language model to assign labels a person can understand. In testing, the same behavior was classified consistently across different email, browser, and messenger apps more than 95 percent of the time.

How well the stress detector works

To build personal stress models, participants shared their phone data alongside daily self-assessments of how stressed they felt. The best-performing model separated stressful days from stress-free ones with an average accuracy of 75.9 percent for the short-term group. For the long-term group, accuracy landed at just under 74 percent. The work was published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, the field's main journal for this kind of research.

The approach has a built-in honesty about its limits. The study shows that phone habits and stress move together, not that any particular habit causes stress. A frantic burst of typing on a bad day could reflect the stress, the workload behind it, or a dozen other things. Either way, the shift shows up in phone habits before most people name the mood behind it. Still, for a technology that works passively in the background, the numbers are strong enough to make researchers think about practical uses: a nudge to take a break when your patterns look like your worst days, or a weekly summary that flags when stressful days start clustering together.

Privacy was part of the design

The researchers deliberately avoided the most invasive data. The system never recorded what anyone typed or read the content of any message. It logged only behavioral traces of phone habits, such as how often a person typed and how they moved between screens. That constraint matters because the most useful stress signal is also the most personal. A tool that reads your texts to decide you are stressed would be a non-starter for most people, and the team built the privacy boundary in from the start.

For now, the system lives in the lab rather than the app store. But the logic behind it is something you can try yourself. The study's core finding is that routine changes reveal strain: when your own phone habits shift, with more frantic switching, bursts of late searching, or messaging that looks nothing like your normal rhythm, treat it as data. Pair that self-check with one deliberate response, like a short walk or ten minutes away from the screen, and the phone goes from being a stress source to a stress signal. According to the researchers, the long-term goal is steadier tracking of mental health over months and years, so that small shifts get noticed before they turn into something bigger.

Why did the team ignore app names entirely? Because people switch apps constantly, and a stress model tied to a specific icon would decay every time the trending app changed. Earlier work measured screen time, launch counts, or usage by app category, and those signals drifted whenever interfaces changed. Behavior purpose is far more stable. Whether someone chats on one platform or another, the underlying routine stays the same, which is why the researchers expect this version to hold up over months of use instead of weeks.

Sources: TechXplore and Korea JoongAng Daily reported on the research.