New App Can Detect Mental Stress Using Your Smartphone

Researchers have developed a smartphone-based system that can detect stress by analyzing everyday activities such as chatting, browsing and making payments.

Key points

• While another 10 provided approximately 105 days of data, sixteen participants contributed around 30.5 days of data on average.
• The researchers tested the system on 26 university students and office workers in their 20s and 30s.
• The researchers plan to present the work at UbiComp/ISWC 2026 in Shanghai.
• Researchers at the Korea Advanced Institute of Science and Technology (KAIST) developed the system, called SHIRBT-MMF.
• The research additionally involved only Android users in their 20s and 30s, limiting how widely the results can be applied.

Unlike earlier approaches that focus mainly on screen time or app launches, the system looks at the tasks a person performs. In a small study involving 26 individuals, personalized models achieved accuracy of around 75%. System Focuses on Tasks Instead of Apps.

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) developed the system, called SHIRBT-MMF. The paper first appeared in December 2025.

Previous stress-detection research often relied on signals such as screen time, app launches, and typing behavior. Nevertheless, these signals can change when apps or interfaces are updated, making long-term tracking less reliable.

Meanwhile, the KAIST team instead focused on activities that remain similar throughout different apps. For example, chatting through KakaoTalk, LINE or WhatsApp is treated as the same general task.

Professor Uichin Lee, who leads the research team, remarked the smartphone should not simply be viewed as a device for measuring usage time, but as a place where activities such as communication, searching and payments repeatedly occur. GTA 6 Will Let Players Carry and Snort Cocaine. GPT-4o Helps Identify Repeated Behaviors. The system detects screen-transition patterns that appear repeatedly over a number of days.

For context, a sizeable language model then analyzes each pattern and assigns it a simple label, such as messenger chatting, web browsing or email reading. The researchers applied OpenAI's GPT-4o for this part of the system.

In tests involving email, browser and messenger apps, SHIRBT-MMF identified the same task correctly more than 95% of the time. Researchers Did Not Store Message Content.

Meanwhile, the system was designed to avoid collecting the actual content users typed.

Instead, it recorded behavioral information, such as whether someone was typing, how often they typed and which screens they moved between.

Information regarding message recipients and call partners was hashed so the original identities could not be directly disclosed. Personalized Models Reached Around 75% Accuracy.

Notably, the researchers tested the system on 26 university students and office workers in their 20s and 30s.

Participants rated their own stress levels each day. Researchers then created personalized models to distinguish stressful days from calmer ones.

While each participant regularly performed around seven of them, the study identified 29 types of repeated behavior.

For context, the best-performing model achieved average accuracy of 75.9% for the short-term group and 73.8% for the long-term group, or roughly three correct predictions out of four.

Even using fewer signals produced useful results. Models based only on chatting and payment behavior, or chatting alone, performed better than the conventional approach.

Notably, the researchers now plan to reduce the amount of data collected to create the system less demanding.

In the long-term dataset, repeated-behavior indicators additionally changed less over time than signals employed in conventional methods. Amazon's Bee AI Gadgets Can Remember Everything For You. Small Sample Limits the Findings.

For context, the study discovered an association between smartphone behavior and stress but did not show that certain behaviors cause or reduce stress.

Meanwhile, the research additionally involved only Android users in their 20s and 30s, limiting how widely the results can be applied.

Meanwhile, the team intends to expand future research to include more age groups, occupations and smartphone environments.

Personalization additionally appears to be notable. A single model designed for everyone performed significantly worse than models built for individual users.

For context, the researchers plan to present the work at UbiComp/ISWC 2026 in Shanghai.

For now, SHIRBT-MMF remains a research project, but the study suggests that analyzing repeated smartphone tasks could offer a more stable way to track stress even as individual apps and interfaces change. Stay Connected with ProPakistani.

Obtain the latest tech news, telecom insights, and product launches wherever you prefer. Follow on Google Discover. Follow on Google News Join WhatsApp. See more ProPakistani stories in Google Search and Top Stories.

Technology and Automotive Specialist covering the latest cars, smartphones, AI breakthroughs, and.

For now, new App Can Detect Mental Stress Using Your Smartphone remains the part of the story worth watching, and further updates are likely as more details are confirmed.