Microsoft Word hci-notes-11 doc


Projects with user interfaces



Download 189.31 Kb.
View original pdf
Page39/40
Date21.11.2022
Size189.31 Kb.
#60003
1   ...   32   33   34   35   36   37   38   39   40
HCI2010
Projects with user interfaces
All of the above applies to projects that do not include user interfaces. If the project does includes a user interface, or is an interactive system, then some kind of user study will be appropriate. It is not expected that user studies at undergraduate level should be large or time consuming, to the extent of those carried out in professional HCI research projects. This means that number of participants is likely to be small, and there maybe substantial variation between them. A statistical significance test of an experimental comparison, for an undergraduate project addressing a problem of realistic complexity, is unlikely to reach the 95% significance level expected of professional research. You (and your supervisors and directors of studies) should not be worried about this – most undergraduate machine learning projects do not achieve 95% statistical accuracy either, and even undergraduate compiler implementations seldom succeed in 19 out of 20 independent trials (because most of them never compile 20 different programs) In cases where it is not feasible to gather sufficient quantitative data fora statistically significant comparison at the 95% level, it is perfectly appropriate to report the mean of two samples, along with a confidence interval. This can be expressed graphically as error bars on the estimated mean. Your main concern should be to estimate the effect size, rather than be too concerned with variability in your sample – you want to know what impact does your system have on user performance Evaluation of user interfaces need not be quantitative, of course. In cases where speed and accuracy measurements are not meaningful ways to assess your project, any of the


45 qualitative methods described above could provide useful evaluations think-aloud studies, interviews, questionnaires, observational studies, or even field trials. Where there is a lot of variation between users, it can be a good idea to interview them and ask why.

Download 189.31 Kb.

Share with your friends:
1   ...   32   33   34   35   36   37   38   39   40




The database is protected by copyright ©ininet.org 2024
send message

    Main page