📎 Webclip
Happy and productive at work: Predicting opportune moments to switch tasks and take breaks
Microsoft Research describes work on balancing productivity and well-being by predicting when an information worker should continue a task, switch to another task, or take a break. The paper uses emotion signals, digital activity, and task information to identify opportune moments for transitions while keeping work moving.
Reading notes#
- The research focuses on helping people move between tasks and breaks without losing productivity or emotional well-being.
- The paper studies the first step of the problem, which is deciding when a transition or break should happen.
- The input data includes digital activity, a daily task list, and emotion data based on facial expressions.
- The study collected data from 25 people for three weeks with a multimodal sensing platform.
- The models estimate a separate expected value for continue, transition, and break, and recommend the action with the highest value.
- Personal models were built for each participant, along with job-cluster models and a general model for comparison.
- ARIMAX models helped identify important features, but they were too slow for real-time use.
- Random forest regression was the best fit for real-time recommendations and stayed close to the ARIMAX results.
- Interaction data was the most important feature on average, followed by task information and emotion.
- Feature importance varied a lot by person, and job-cluster patterns did not always match individual patterns.
- In testing, participants agreed with the timing of transition recommendations 85.7 percent of the time and break recommendations 77 percent of the time.
- Agreement with the recommendation did not mean people could always comply right away.
- The authors present the work as a step toward AI assistants that help people act on recommendations and manage work more effectively.
