Dissertation Title: Investigating Interventions in Fine-grained Contexts for Habit Formation

Abstract:

Behavior change is critical for well-being, but is often hard to sustain. Habits are automatic responses to specific contextual cues, and can help sustain behavior change. Habit-development interventions have been explored using automatic interventions in broad contexts like time, geolocation, and physical activity, but not in fine-grained contexts such as eating, brushing teeth, or using social media. This thesis investigates the potential of habit-formation interventions in fine-grained contexts.

We explore habit formation in both mobile, physical-world and digital, computer-based contexts, making three key contributions for each: a survey to identify behavior change needs, a prototype system designed to deliver fine-grained context-specific interventions, and a study to investigate habit-formation using interventions in fine-grained contexts, compared to interventions in less fine-grained contexts. We use the standardized Self-report Habit Index and Self-Report Behavioral Automaticity Index to measure habit formation.

For mobile, physical-world behavior change, the survey of needs (N=53 participants) indicated that participants want diverse and personalized behavior change support in diverse and specific contexts. We created a wearable device with on-device visual context detection and open-ear audio for real-time mobile interventions in egocentric visual contexts. In a 4-week study (N=10 participants), interventions using egocentric visual contexts led to more habit formation than interventions in course contexts based on time, geolocation, and physical activity. The habits also persisted in the post-study evaluations 1 and 10 weeks later.

For computer-usage behavior change, the survey of needs (N=63 participants) indicated that participants want to reduce excessive/unnecessary use, e.g., social media and entertainment. We created a Chrome extension to deliver web interventions based on specific web activities. In a 6+2-week study (N=31 participants), interventions in web-activity-based contexts led to more habit formation than interventions in course contexts based on intervals or random contexts. Habits also persisted 2 weeks after the 6 weeks of interventions.

Due to the small participant pool, the quantitative results are not statistically significant. Nonetheless, the qualitative results showed that interventions in fine-grained contexts occur in more user-desired and less disruptive or ignorable moments. Our research represents a first step toward understanding the potential of habit-formation support using interventions in fine-grained contexts.

Committee

Pattie Maes, Germeshausen Professor of Media, Arts and Sciences, MIT Media Lab
Rosalind Picard, Professor of Media, Arts and Sciences, MIT Media Lab
Shakir Mohamed, Research Director, Google DeepMind