Saturday, 18 June 2011

Science Is Hard (Says Microsoft)

As I warned last week, Microsoft seems to have branched (PDF) away from pure IT...

To be fair, the paper is mainly targeted at proving how (most, or ta least many) cyber crime surveys give misleading results. The interesting bit is that to corroborate the findings, the researchers also analyse self-reported surveys of number of sexual partners, which also seem to give misleading results for men and women, respectively. To be even fairer, I couldn't really fault the findings (do have a look; it's surprisingly readable).

Still, I have to take issue with a couple of things.

First, the authors seem to imply that the large difference in numbers of sexual partners between men and women is "of course[, ] impossible". Now, the authors do not offer neither any further detail, nor do they offer any references in support of this claim - even if it were really "of course" true. From what they say it can (must?) be implied that there should not be any, or at least great, difference. But is that really so? Methinks not. Due to evolutionary pressures men are adapted to be more prone to zipless sex (to borrow a Jong quote, here), while women are adapted to be much more picky. This would imply that it is more likely that more men have more sexual partners, and conversely that a comparatively smaller proportion of women are so "liberated" as to be more promiscuous than evolutionarily sensible. Which would mean that the bell curves of numbers of sexual partners for men and women are necessarily skewed - and skewed in opposite directions. How much skewed is open to investigation (and better surveys), but to imply otherwise is just wrong, if not irresponsible. As a side note, if neo-feminists had their way, the skewedness would be even greater with most women just plain having no sex at all. But I digress...

The second problem I found in the paper is probably worse from the point of view of Microsoft, the authors, and those who'd use it to some good end, equally. This problem is in the part which is the most important, the Recommendations. It opens, somewhat facetiously, by telling us that "[survey] science is hard". No shit! It then goes on to say that "no weight can be given to surveys that fail to disclose methodology". Well, hello! Whoever looks at a survey without knowing what they look at deserves what's coming to them. Only it shouldn't take a handful of academics to spend time and money on this paper to get to that conclusion. Again, in all fairness, most of the other "hard" recommendations make perfect sense. However, IMHO, every single one of them can be deduced by anyone clever enough to know that a) science is hard, b) surveys have to have appropriate methodology, and c) care about the above two. So, essentially, the authors use 11 pages to tell us to be careful when designing and perusing surveys, and consult relevant literature. D'oh!

In conclusion: yes, science is hard, but Microsoft should really try harder...