Context-Aware Delivery of Ecological Momentary Assessment

Area of Science:

• Behavioral science
• Computer science
• Human-computer interaction

Background:

• Ecological Momentary Assessment (EMA) collects real-time data, reducing recall bias and enhancing ecological validity.
• Optimizing EMA prompt timing is crucial for participant engagement and data quality.
• Traditional time-based prompting can be intrusive and lead to low response rates.

Purpose of the Study:

• To introduce SEP-EMA, a machine learning method for context-aware EMA prompt timing based on activity transitions.
• To evaluate the effectiveness of SEP-EMA compared to traditional time-based prompting.

Main Methods:

• Developed SEP-EMA, a machine learning model for adaptive EMA prompt scheduling.
• Compared SEP-EMA with time-based prompting in a study with 19 smart home residents.
• Collected data on participant response rates and EMA effectiveness.

Main Results:

• SEP-EMA increased participant response rates by 7.19% compared to time-based prompting.
• Context-aware, transition-based prompting significantly improved EMA usability and effectiveness.
• Reduced data loss associated with low response rates.

Conclusions:

• SEP-EMA offers a more effective and usable approach to EMA prompting.
• Timing EMA prompts during activity transitions enhances engagement and data quality.
• Machine learning-based adaptive prompting is a promising direction for improving EMA systems.