A few weeks ago I put up a post about the research challenges in studying online communities, in which I suggested that online surveys are an essential but flawed tool for the study of communities that are largely defined by their internet rather than their physical presence. My post was in the context of the controversy over Lewandowsky’s analysis of data from online climate “skeptics,” but in contrasting online communities with marginalized or stigmatized groups, I implied that online surveys are part of a broader problem in research, dating from before the internet, in how to access people who are not easy to trace or very rare.
Today’s issue of the journal PLOS Medicine has an article about ethical considerations in online research, which is published in the context of online surveys of medical treatments and genome studies. The paper, available open access here, has a nice description of the types of bias that enter online studies, and also a discussion of the ethical implications that arise both from the nature of these biases and from the general properties of online surveys. They mention the possibility that data will be shared and the importance of telling participants, something which I suspect neither Lewandowsky nor the follow up survey at WUWT explained to their participants; they also mention the possibility of data sales and the additional complexities of obtaining consent from non-identifiable participants.
The article doesn’t make any revolutionary claims or present any strong judgments about whether online research is bad, good or better than other forms of research, but it does note the growth of this type of research, discusses its applicability and generalizability, and points out some of the ethical implications that arise from the ease with which people can conduct online research. I was certainly surprised to read the kind of data people are willing to exchange with online research organizations of the type identified in the article and I wonder if there is a broader problem here in that our technological ability to collect private and sensitive data from strangers has outpaced the community’s understanding of the risks and ethics of such practice. If the wash-up of the Lewandowsky affair has you wondering about the broader issues surrounding that type of research, then the article is certainly worth reading.
Compromise and Conceit 
November 2, 2012 at 7:27 am
Statistical research needs to catch up with where the world is now up to and where it’s heading. Hundreds of years back the king’s men needed to ride around the country and collect census data. 50 years ago telephone polling sped up data capture and enabled modern poll driven politics [1]. Today the Internet allows specific communities to be targeted. [2]
And in just a couple of years statistical research will be conducted by just asking the Facebook corporation, which will have such a large collection of privacy shunning fools that comprehensive studies into formerly difficult to discuss topics such as bowel polyps will be the work of minutes. At the same time, Facebook avoiding paranoids will be declared a non-human species, thereby enabling 100% data collection rates from the human populace.
Of course, the real challenge that will come from this bold new world will be “What will statisticians do with the extra time they will have available?” I suggest more blog posts and maybe taking up a MMO hobby [3]
[1] Thanks for that.
[2] And I know that other readers will be looking forward to indepth analysis of the furry subculture just as much as I am
[3] Given that’s basically an infinite time sink.
November 2, 2012 at 10:42 am
I think I speak for all of us when I say that the day where everyone’s bowel polyps are discussed openly on facebook cannot come soon enough. Also, if you can make any suggestions as to how this statistician can find that fabled “extra time” without playing silly buggers with quantum mechanics, I’m all ears. At the moment I’m completely snowed under with work (hence the low posting rate here). I may even have to cancel tomorrow’s role-playing so I can work all weekend. The humanity!
November 2, 2012 at 9:09 pm
Just make sure that you post all details of your life onto Facebook. It shoulds like it shouldn’t take long as posting “Working on weekend” is generally considered poor form.
Maybe just perform a self assessment of your risk rate of Type II diabetes and leave it on Facebook so that some other statistician can easily assemble a data set. It’s a karmic system approach to data gathering…