• Figure 2. Data mining for social norms.

directions to the nearest cupcake shop, or need to tell our friend we are running late, our technology should know that what we ask for is not always in our best interests.

And when it comes to important personal transitions, it is often difficult to know the desired destination or even how to frame a question: We may simply have an idea we want more novelty, more fun, or stronger relationships. Our apps and devices need to listen for our goals over time, challenge us when our immediate requests run counter to them, and steer us in the right direction.

Apply social influence. Our technol-
ogies know a tremendous amount
about us and others. This knowl-
edge, which now guides Google
searches, Amazon purchases, and
a wide array of other consumer
profiling, can also drive healthful
change. For example, personal data
can be used to reference peer groups
who engage in healthful habits.
The most relevant and specific peer
group should be identified for such
referencing. Social comparison has
been used to influence behavior in
many contexts, from clinical inter-
ventions for substance abuse to
social psychology experiments on
environmentalism. To curtail sub-
stance abuse, it has been effective
to inform heavy drinkers about the
lower consumption of their peers,
using feedback such as “You drink a
six-pack every Friday and Saturday;
most men your age have only two
drinks a week” [ 7]. Similarly, hotel
guests are far more likely to reuse
their towels for the length of their
stay when they are told that the
majority of other guests who stayed
in the same room had done so [ 8].