DIGITAL LIFE

Smartphone chatting and payment habits help ai identify daily stress, small study finds

Your smartphone may be able to tell you when you are having a stressful day based on how you use it, a KAIST research team has found.

The AI-assisted technique, developed by a team at the KAIST School of Computing, assesses users' daily stress based on the everyday tasks they repeatedly perform on their phones, the university said Thursday.

The best-performing model identified stressful days with an average accuracy of 75.9 percent among participants tracked for about a month, and 73.8 percent among those tracked for about 105 days.

The research ties mental health monitoring to the smartphone, a device most people now carry everywhere. Because phone usage records capture changes in everyday behavior over time, they are increasingly used as a digital indicator of mental health conditions such as stress and depression.

Earlier studies mostly measured screen time or which apps people used, but those numbers can change whenever someone switches apps. The researchers gave the example of someone who moves from KakaoTalk to WhatsApp or Line. The app changes, but the task stays the same: chatting on a messenger.

The KAIST team instead focused on what users do on their phones rather than which apps they use, so the data stays consistent even when people switch apps.

The team analyzed data from 26 university students and office workers in their 20s and 30s, divided into 16 short-term participants, who participated for 30.5 days on average, and 10 long-term participants, who took part in the study for 105.1 days on average. The participants provided both phone data and daily self-assessments on stress, which were used to create a personal model.

The best-performing model distinguished stressful days from stress-free ones with an average accuracy of 75.9 percent for the short-term group and 73.8 percent for the long-term group.

The study showed a link between phone habit and stress, not that any habit causes it. The research also did not record what users typed or what their messages said, to protect privacy. Instead, it logged how often a person typed and how they moved between screens.

"We looked at the smartphone not simply as a device that measures usage time, but as a living space where a person's communication, information searches and financial activities repeat," said Prof. Lee Ui-chan of KAIST's School of Computing.

“By turning complex smartphone records into units of behavior that people can understand, this study lays the groundwork for analyzing long-term changes in mental health more reliably,” Lee said.

Three types of smartphone users...Participants were divided into three groups according to their habits. “Pure social” users mainly used their phones to communicate, maintain relationships, and form new social connections.

“Light process” users spent less than four to five hours a day checking news, viewing entertainment, using social platforms, and passively scrolling. About one third were classified as “heavy process” users who spent four to five hours or more on their phones each day, primarily consuming content rather than talking with others.

“Our main finding is that personality alone—such as a high level of neuroticism—does not make someone addicted to their phone. People who are more prone to negative emotions like anxiety, depression, or stress do not automatically become addicted. What really matters is how well they can control their behavior, emotions, and impulses and how strongly they fear missing out,” said Dr. Johanna Takács, senior research fellow at the Faculty of Health Sciences at Semmelweis University and lead researcher of the study.

Passive scrolling strains attention...The findings point to passive content consumption as a particularly concerning behavior. Social feeds and short videos deliver a rapid succession of novel stimuli, continually pulling attention from one item to the next. Prolonged exposure to that pattern may make sustained concentration more difficult and contribute to mental fatigue.

On cognitive tests, “heavy process” users performed worse than expected for their age on measures of attention and memory. They also slept more poorly, sought new stimulation more intensely, and found it harder to resist the urge to check their phones or satisfy an impulse immediately.

Time still mattered, but primarily for physical health. Longer use was associated with posture-related effects that may extend beyond neck discomfort to balance, reaction time, and everyday movement.

Because the research identified associations rather than assigning participants to different smartphone habits, it cannot prove that phone use directly caused the observed mental or physical differences. The results instead highlight patterns that may help identify which behaviors deserve closer attention.

How text neck affects the body...“When people hold their phones at waist or chest level and look down at them, the head moves forward, creating a forward head posture, also known as ‘text neck.’ This puts extra strain on the neck and can cause pain or early wear and tear of the neck joints. The neck contains receptors that tell the brain how the head is positioned. If a forward head posture is maintained for long periods, these signals can change and lead to balance problems. These effects are not limited to phone use but can also influence everyday movements,” said Beáta Seregély, assistant lecturer and expert responsible for the physical measurements in the study.

The head becomes harder for the neck to support as it shifts forward, increasing the load on muscles and joints. At the same time, prolonged changes in head position may interfere with sensory information that helps the brain coordinate posture and balance.

Building healthier phone habits...The researchers do not recommend banning smartphones or blaming young people for using them. Their findings instead support helping users become more deliberate about when, why, and how they reach for their devices.

Practical steps include silencing unnecessary notifications, especially alerts from social media and messaging apps; setting limits on social media use; avoiding phones before bedtime; exercising regularly; and keeping the screen closer to eye level to reduce strain.

Mental health strategies may be just as important. Behavioral training and psychological support can help young people strengthen self-control and become less anxious about missing social events. As those pressures ease, people who are sensitive, anxious, or prone to stress may be far less likely to develop problematic smartphone habits.

A new study by researchers at KAIST (Korea Advanced Institute of Science and Technology) shows that artificial intelligence can identify daily stress with up to 75.9% accuracy by analyzing routine smartphone habits rather than just total screen time:

• Focus on Intent, Not Apps: Instead of tracking specific app names (which change or get swapped out by users, like moving from KakaoTalk to WhatsApp), the AI focuses on what a user is trying to accomplish—such as messenger chatting, web browsing, financial payments, or reading and writing emails.

• SHIRBT-MMF Technology: The research team developed a system called "Smartphone Human Interaction Routine Behavior Tasks" (SHIRBTs) using a technique named Multi-Level Sequential Pattern Mining (ML-SPM) combined with a large language model (LLM) to categorize everyday screen transitions.

• Accuracy Rates:

• Reached 75.9% average accuracy for short-term participants (tracked for an average of 30.5 days).

• Reached 73.8% average accuracy for long-term participants (tracked for an average of 105.1 days).

• This performance was about 3.5 to 5.5 percentage points higher than traditional app-usage tracking methods.

• Participant Group: The study analyzed real-world data from 26 university students and office workers in their 20s and 30s using Android devices.

• Privacy Protection: The system did not store actual message text or contents; it only logged behavioral patterns (like frequency and screen transitions), and contact identifiers were processed using hashing (converting data into anonymous values).

• Important Caveat: The researchers note that the study shows an association between behavioral patterns and stress, not a direct causal relationship

A research team at KAIST (Korea Advanced Institute of Science and Technology) developed an AI-assisted technique that identifies daily stress by tracking routine smartphone usage, achieving an accuracy of up to 75.9%.

Key findings of the study:

• Accuracy: The top-performing personal AI model distinguished stressful days from stress-free days with an average accuracy of 75.9% for short-term participants (tracked for roughly 30 days) and 73.8% for long-term participants (tracked for about 105 days).

• Behaviors Tracked: The system analyzes everyday actions such as typing frequency, screen transitions, communication patterns, information searches, and financial/payment activities.

• Privacy-First Design: To protect user privacy, the AI did not read the content of text messages or record what users typed, focusing instead on behavioral metadata like movement between screens and interaction rhythms.

• Participants: The study tracked Android smartphone users—consisting of 26 university students and office workers in their 20s and 30s—who logged their phone habits alongside daily self-assessments of stress.

• Correlation, Not Causation: The researchers noted that the findings show a strong association between phone habits and stress, rather than proving a direct causal relationship where a specific habit causes stress.

by mundophone