digital mirror _ critical computing _ jeana frost _ brian smith
frost@media.mit.edu

This research is about diabetics and their theories about their health practices. Daily decisions, such as eating and exercise habits, have clear consequences for a diabetic's health, yet many of them fail to change their behavior patterns despite knowing facts about the disease. We assume that they might change their practices if the relationship between their actions and their blood glucose levels was made explicit. To do this, we introduced photography into diabetes diagnosis, asking people to collect images of their daily activities to complement their glucose monitoring. By combining quantitative glucose measurements with qualitative portraits of action, we hope to make the relationship between physiology and behavior an object for discussion and reflection. More so, we hope that diabetics who viewed these data will begin to develop new interpretations of their lifestyles that will ultimately lead to healthier activities. The results of our initial studies suggest that photographing activities may have the potential to change the ways that diabetics understand and deal with their health.

More sample images
Further reading:
presented at dux2003 (designing user experiences)
presented at CSCL on preliminary results
presented at CHI on visualization techniques
Master's thesis

 

Diabetics  capture data: They shoot digital photographs of activities and events that affect their health and monitor blood-sugar on a regular basis as usual.

 

 

Example images
and the comments they inspired:	the deck could be a good place to relax.	stairs could provide an alternate form of exercise.	sitting on the couch watching TV may be bad for health.
	the salad looks healthy, but the ashtray next to it is suspicious.	you can exercise in your room as part of your daily schedule.	the diet pepsi is ok, but whose drinking that beer?

 

 

Then we built web - based tools to synchronize images with blood-sugar levels. Users uploaded images and blood sugar values to the web site where they were consolidated and visualized. Blood sugar data is organized to give an overall reading of the blood sugar. In the figure below, warm colors indicate high values, blues indicate lows, and yellow represents blood sugar values within target range. Each column is a day and hours proceed from top to bottom. The diabetic whose data is visualized below often has high blood sugar readings.

 

Alternate visualization tools elucidate different attributes in the data. For example, these radial plots have been incorporated to show the daily routine in how people measure. The circle represents a 24 hour clock with each petal-like polygon representing one test. The "petal" length represents the blood glucose levels. This is an example from a person who tests every day around 12 midnight and 9pm. The color coding is the same as in the grid visualization with blue low, yellow target, and red high.

Users “mouse over” a day of interest to see the photographs he or she shot that day. By matching blood sugar values with images, the user can begin to see how particular events and/or behaviors contribute to fluctuations in blood sugar. We collected 5 weeks of data for one type I diabetic. Below are two examples from his record. In the case on the left, pizza caused the most extreme rise in blood sugar tracked during those five weeks. The corn beef pictured on the right caused a similar climb in blood sugar.