Stay Hydrated, My Friend

In a paper at the 2009 Ubicomp conference, Hao-Hua Chu and his students and colleagues presented results of a trial of a system that encourages people to drink more water. They built a tracker that sensed water consumption from a special bottle.

Some subjects got an individual feedback app. When they drank enough, their tree avatar was healthy; when they didn’t, it lost its leaves.

Five subjects got a version with two social features:

1. They could see small representations of the other four users’ tree avatars
2. They could send reminders to other users (in the form of heart icons, suggesting “I care about you, so drink!”)

The study had too few subjects to yield conclusive results, but it appeared that the subjects in the social version had a greater increase in water intake, compared to personal baseline), than the subjects in the individual condition.

It appeared that the social reminders are an active ingredient. The median time to drinking was about 12 minutes in response to social reminders, about 20 minutes in response to system-generated reminders (the system generated reminders in both the individual and social versions of the app).

(Aside: the paper reports that the difference was statistically significant, but this is due, I think, to a mistake in the analysis. The analysis treats each subject-day as a independent observation, which is not reasonable since each subject contributed 20+ observations. The analysis should have clustered by subject-id, or another equivalent technique for compensating for the correlation between observations for the same subject. The statistical significance would almost certainly disappear. Shame on the authors and conference reviewers for missing this. Still an interesting project and paper, though; glad it was published.)