This table shows the numerical results we parsed from looking at the touch
sensor logs generated by the Geo.  The count is of the total number of reward
signals and punishment signals given (the number of times each sensor was
touched); positive:negative is the ratio of positive reward signals to negative
reward signals given.

From the results, we can draw several conclusions.  First, non-engagement Geo
produced approximately twice as many touches as engagement Geo. This result
surprised us, as we thought people would interact with engagement Geo more.
However, it turns out that the additional touches for the non-engagement Geo
were due to the lack of feedback responses from the Geo. This silence on the
part of the Geo caused people to not understand whether their reward and
punishment signals were being received, which in turn led to overcompensation.

We also observed that some people treated engagement Geo with some degree of
“fear” – as if the autonomous behavior was coming from a living entity, and as
such were reluctant to touch it extensively and unnecessarily. People in this
category typically became more willing to touch the Geo overtime, as they
“warmed” to it.

The second observation that we can draw from the experiment is that the ratio of
positive to negative rewards depend significantly on engagement. With engagement
mechanisms turned on, people gave a lot more positive rewards (relative to
negative rewards) than when engagement mechanisms were turned off. We believe
that this is because people viewed engagement Geo as a more social, “lifelike”
entity, and as such did not want to “hurt” it by giving too many negative
rewards.

A third and final observation is that with the engagement Geo, the ratio of
positive to negative rewards appear to skew towards the positive overtime, which
is reflected in the higher ratio (2.7:1 vs 1.6:1 positive:negative ratio) in the
freeform dance task, which always comes after the fixed dance task.  Meanwhile,
with non-engagement Geo, the ratio of positive to negative rewards did not show
any real difference between the hokey-pokey “fixed” dance teaching task and the
freeform dance teaching task. This could be because people were “warming” up to
engagement Geo overtime and began developing some sort of “social connection”
with it.