Paper Reading #31: Identifying emotional states using keystroke dynamics

References
Identifying emotional states using keystroke dynamics by Clayton Epp, Michael Lippold, and Regan L. Mandryk.  Presented at the CHI '11 Proceedings of the 2011 annual conference on Human factors in computing systems.

Author Bios

• Clayton Epp is currently a software engineer for a private consulting company and holds a master's degree in CHI from the University of Saskatchewan.
• Michael Lippold is currently a masters student at the University of Saskatchewan.
• Regan L. Mandryk  is an Assistant Professor in the Interaction Lab in the Department of Computer Science at the University of Saskatchewan.

Summary

• Hypothesis
• It is possible to determine a person's emotional state based on their keystrokes.
• Methods 
• The researchers used a software program to collect keystroke patterns of the participants.  Based on the user's level of activity, the program prompted the user with an emotional state questionnaire and another short piece of text to type.  Users with fewer than 50 responses were eliminated on the basis that they didn't provide enough variance in response to be useful.   The raw data collected included key press and release events, codes for each key, and a timestamp on key events.
• Results
• The researchers used undersampling on many of the models to help make the data more meaningful in terms of detectable levels of emotion.  They found that two of their "tired" models performed most accurately with the most consistency, and that models utilizing the undersampling performed better overall.
• Contents
• In this paper the researchers describe how minute measurements of a person's keystrokes can be calibrated to give a reasonably accurate representation of their current emotional state.  They discuss some of the related work in human emotion and computing, and go on to describe their experiment process in detail.  They also describe some of the ways that this paper might be expanded upon in the future.

Discussion
I found this paper extremely interesting and was pleased to see that they were able to come up with a fairly consistent model for their research.  They mentioned that their results might be improved upon in a laboratory setting with elicited emotions, and I agree.  Their data was a little bit weak simply due to the nature of the study.  However, I feel that they accomplished their overall goal and I am quite convinced of the results.