As we have seen, the results of this thesis raise several interesting questions which ought to be addressed in future work. There are a number of variations on the experimental designs that could yield other useful results related to the ethical implications of affective computing systems.

One extremely important aspect that this thesis neglected was the exploration of the physiology data collected. Pattern recognition on the skin conductance, mouse pressure, EKG, and video data collected has the potential to provide insights into not just subjects' reports but subjects' physiological reactions.

In the Quiz experiment it would be interesting to explore how subject results varied if they are not told that the average score was "21.7" and instead are made to feel positively about their performance. This might serve to make the control condition more of a neutral situation as opposed to one which manipulates subjects in negative ways.

Closer examination of the relationship between emotional involvement and unethical behavior as well as strong negative opinions would also be enlightening. Future work should not treat affective responses and ethical opinions as closed systems which do not interact with one another.

One problem that occurred frequently with the MixedEmotions program was the presence of noise in the sensor data due to motion artifacts and errors in the device drivers for the EKG system. It would be interesting to explore variations of the experiment which both reduce and increase the amount of noise in this data and subsequently examine how performance varies.

In the Interview experiment it would be interesting if the interviewer had "something else" (such as annoying sounds in the experimental space) to attribute discomfort to besides the unusual situation. It would also intriguing to see how subjects change their impressions if they are induced into other affective states prior to starting the experiment, perhaps by watching videos with compelling content.

Lastly, it would be worthwhile to repeat the experiments whilst varying how much information subjects have about the capabilities of affective systems used. In the experiments documented here, participants were deliberately not told about the capacities of the sensors they used. If instead, subjects had feedback or detailed information about the sort of emotional information transmitted and how it might be used, then the results are likely to be quite different. How might individuals feel if they are given everything in truth and in detail?