[Air-l] Great Ethical disasters in Internet Research?
waern at dsv.su.se
Thu Jul 18 12:25:36 PDT 2002
Sorry to be so late in reacting. The issue caught my interest after
three weeks on a Swedish Island - Gotland. Sunny! So I may need some
intellectual change! Here is my reaction to the discussion.
>Thanks for putting the homenet study as a warning example! I have
>used it myself in a course in methodology. Useful to have flawed
>studies! Did you see, for example, that the web-formt of the paper
>does not allow you to see the figures? I had to get the original
I also saw that somebody asked for the "flaws" in the study. I have
seen a couple:
1) the sample of respondents is not random. This is common in
psychological research, but here, the sample is probably rather
biased. People did get a computer to try. How many have not already
got one? Who are the people who got the computer?
2) the data are mostly based on ratings. Ratings have a tendency to
get biased if they are repeated. There is something that is called
"regression toward the mean". This means for instance that if you are
very positive at the beginning, the average will be lower at a
repeated measure, just because the average was above the (supposed)
mean from the beginning. Thus, one possible bias can be the
following: people were happy to get a computer. Later, they, probably
got disappointed with the utility of the computer. This affects all
3) the actual "significant" differences are VERY small! You have to
read the original paper. I usually do not recommend my students to
talk about effects in this range at all!
Good methodology and ethics go together in research, as many have
proposed. We are responsible to our subjects as well as to other
researchers to do as good a research as possible. Why do we else have
research at all?
>1. Carnegie Mellon's HomeNet study
Computer and System Sciences
Stockholm University /KTH
Forum 100, S-164 40 Kista, Sweden
Phone (home): +46 8 500 307 18
More information about the Air-l