Motivation

Problem Formulation

Implementation
System Overview
Hardware
Interaction Dialog

Discussion

Discussion & Future Direction

Due to the longitudinal nature of this device and some of the initial issues I had with the hardware I have a paucity of statistically viable results. However, I have learned a few things through my trials. For the experiments I ran, the association task gives nice results as far as negative and positive valence goes. Unfortunately I need to tweak the score to get the neutral state better. Most of the informal trials I ran saw people answering neutral to the validation dialog using Russell's model. However, I did not give the subjects privacy as I should have and thus we have a problem with the self report as would be expected.

As far as the model goes, I would love to get a better model of emotions for this task. I looked into other measures of affect and mood and found that two popular ones are Nolis' MACL and the POMS-BI. I would like to look into these, but they are hard to get cheap.

Heart Rate is not the most reliable measure. In addition to me missing samples and the average varying more than I would like over time due to noise, I also cannot get emotions that do not exhibit a noticeable change in heart rate such as the admiration discussed in the McCraty paper. One way to go is to use a GSR. The other way is to implement the Heart Rate Variability algorithm.

Beyond that, I would like to couple this with a more complicated system architecture that promotes healthy behavior change  over a long period of time. This will require the system to be a little more dynamic, establish trust, and fullfill its promises so that the user does not get board and stop using it after the novelty wears off.