Personalized Micro Rewards

Micro-incentives for digital wellbeing

Have you ever promised yourself to stop doomscrolling, only to keep swiping?

If so, you're far from alone. Late-night scrolling through endless Reels or Shorts has become a modern habit with real consequences. This behavior, known as "doomscrolling," isn't just a minor distraction. It affects sleep quality, mental focus, and overall wellbeing.

So how can we gently regain control of our screen time without resorting to harsh digital restrictions or blocking apps altogether? KAIST's Interactive Computing Lab (ICLab) explored a softer, smarter alternative: personalized micro-financial incentives. Their study, recently presented at the CHI 2025 conference, suggests that even small nudges like a few cents can make a real difference.

A Gentle Nudge with the WellbeingWallet App

Main screen of WellbeingWallet

To test their idea, they developed an Android app called WellbeingWallet. The app breaks the day into hourly segments and automatically presents users with a behavioral mission: to keep their phone usage under 10 minutes during that hour. If the user succeeds, they receive a small cash reward ranging between 0 and 100 KRW (roughly 0–8 cents).

The interesting part? The reward amount isn’t fixed. It changes dynamically based on the user’s behavior and context. Some hours might offer 25 KRW, others 75 KRW. The app also provides real-time feedback on usage, visual progress tracking, and even automatically records usage during a baseline phase before rewards are introduced.

How the Rewards Are Determined

Multi-Armed Bandit using Thompson Sampling

The app does not pick reward amounts randomly. Instead, it uses a machine learning technique called the Multi Armed Bandit algorithm with multi objective optimization, which is based on reinforcement learning. To understand this concept, imagine a row of slot machines in a casino. Each machine has a different chance of giving a payout, but you do not know which one is best. You could keep trying different machines. This is called exploration. Or you could stick with the one that has worked well so far. This is called exploitation. The challenge is finding the best balance between these two strategies.

Casino Slot Machine

In WellbeingWallet, each possible reward amount such as 0 KRW, 25 KRW, 50 KRW, 75 KRW, or 100 KRW is treated like a different slot machine arm. But rather than simply looking for the reward that leads to the highest success rate, the algorithm also considers cost. It uses a multi objective approach to find the smallest amount that is still effective, balancing the chance of success with the need to spend less. It evaluates how well each reward works in different situations such as work hours during weekdays, evenings, or weekends, and selects the most efficient reward accordingly.

For example, the algorithm may determine that during weekday mornings, a user often succeeds with just 25 KRW, while on Friday nights, they may need 75 KRW to stay on track. Instead of offering high rewards all the time, the algorithm learns and adapts in real time to minimize cost while still encouraging success. It aims to give just enough motivation, not too little and not too much.

This method avoids a one size fits all solution. Instead, it creates personalized strategies that adjust to each person's daily routines, preferences, and environment

The Four Week Field Study

This was not just a lab experiment. The research team recruited 72 participants and ran a four week real world study. In the first week, users did not receive any rewards. The app just tracked their screen time to establish a baseline. In weeks two and three, participants were split into three groups. One group received a fixed reward for each success, another received randomly chosen rewards, and the third group received personalized rewards decided by the Multi Armed Bandit algorithm. In the fourth and final week, all rewards were removed to see whether the changes would stick.

Randomized Controlled Experiment

All three groups reduced their screen time during the intervention period. Interestingly, this reduced usage persisted even after the rewards were removed. This suggests that the app not only helped people change in the moment, but also encouraged longer term behavior shifts.

What stood out the most was how cost efficient the personalized approach was. While the fixed reward group received an average of 7,193 KRW and the random group received about 6,891 KRW over the two weeks, the personalized group received only 3,774 KRW and still achieved similar reductions in screen time. In other words, personalization helped achieve the same impact at a much lower cost.

Contribution

This study provides evidence that even small financial rewards can meaningfully change behavior, especially when those rewards are personalized and delivered at the right time. Importantly, it shows that behavior change does not require massive interventions or drastic restrictions. Sometimes, all we need is a well designed nudge.

In a world where digital habits are increasingly difficult to manage, this research opens up a new path for improving digital wellbeing. By combining behavioral science and smart algorithms, we can help people regain control of their attention without taking away their autonomy.

Citation

Paper Title: Like Adding a Small Weight to a Scale About to Tip: Personalizing Micro Financial Incentives for Digital Wellbeing
Conference: CHI 2025, Yokohama, Japan
Authors: Sueun Jang, Youngseok Seo, Woohyeok Choi, Uichin Lee
Link: https://dl.acm.org/doi/full/10.1145/3706598.3714208

About the Author

Somin Park(som2n@kaist.ac.kr) is a master’s student at KAIST School of Computing, conducting research at the Interactive Computing Lab with advisor Professor Uichin Lee. Her research interests are Human Computer Interaction, Sleep Health, Smartphone Usage Analysis, and DUI sensing.