Margaret Burnett.

I attended the CS Colloquium on Thursday September 6th. The speaker was Margaret M. Burnett. Burnett's work focuses on "end-user programming" and on gender issues in programming and human-computer interaction. Her colloquium talk on mainly addressed her "WYSIWYT" ("What You See Is What You Test") paradigm for encouraging and/or enabling end-users to test their programming for correctness.

Burnett clarified the phrase, "end-user programming" with some assumptions that she made throughout the talk: 1) by "end-user" she means users who are not programmers in the traditional sense, eg. engineers, accountants, managers, etc., and 2) by "programming" she means very high-level programming tasks such as designing spreadsheets.

The motivation for this research is that there are undeniably a large number of end-user-created spreadsheets in existence, and that according to Burnett's findings as much as 90% of all such spreadsheets contain errors. I was glad to hear that she and others are addressing software design from this perspective; engineers have historically been notorious for ignoring the needs of the layperson.

Burnett's work is specifically targeted toward users who do not necessarily enjoy programming and do not necessarily want to do testing. To overcome this challenge, she has developed a methodology called, "Surprise-Explain-Reward." Instead of training users to use their testing suite, she and her colleagues have integrated the testing into the user experience of programming a spreadsheet. While a user is working, the testing software will code certain cells by color based on their level of "testedness." A curious user may click by choice to learn more about the testing, and the software will suggest test values for certain formulas. Often these test values will be chosen as unusual or "out-of-bounds" cases that force the user to acknowledge constraints on permissible values. The more a user chooses to engage in the testing process, thon gets rewarded for thon's efforts as the tested cells change color to indicate increasing levels of "testedness."

I feel that Burnett's WYSIWYT testing methodology is a useful and highly important development in "end-user programming." I am certain that many businesses would benefit from the technology. However, to be fully taken advantage of it must be integrated into mainstream applications such as Microsoft Office. The main challenge will be to make this transition to widespread commercial use. Once the technology is actually in front of a user, the "Surprise-Explain-Reward" system should take care of the rest.

Last modified 18 October 2007 at 8:11 am by danknights