Friday, 4 July 2014

Facebook Does NOT Do Science

This past week has seen quite a bit of noise relating to the ill-fated, ill-conceived, and ill-received experiment Facebook and a couple of US universities ("universities"?) conducted on Facebook users in 2012.

Quite a lot of digital ink has been spilled trying to divine if what has been done was plain wrong, potentially criminal, simply misguided, just poorly communicated, none of the above or even all of it mixed up then spat out by news agencies of various levels of scientific (and journalistic) competencies.

What seems to have got lost in the noise is the following set of simple facts:

  1. The research has been conducted by Facebook's research "scientist" and two real scientists, both hailing from reasonably decent American universities.
  2. The research has been published in a bona fide scientific publication, PNAS (it's here, for your reading pleasure).
Call me old fashioned, but with these two simple, indisputable - and undisputed - facts, should come a third, or they should have at least implied (and enforced, by at least PNAS, if not everybody else involved), that each and every participant has given their informed - and explicit - consent to take part in a targeted psychological study. Period. No ifs and no buts.

The Facebook mounts a defence based on their Terms and Conditions of Service where to use their "product" (even if the product here is, effectively, you) you agree your data to be used in various (but presumably not nefarious) ways within the company, including for research purposes. And yes, I am sure that's what people sign away when they join. I can't really vouch for it myself as I don't use their "product".

However, what most people - and, I submit, any reasonable person on this planet - assume this entails is, for want of a better word, "passive" use of such data. For example, harvesting your data to better target ads. Or even to better target news items you see.

But surely nobody in their right mind would read the Ts & Cs (if they do in the first place) and assume "data use" to involve manipulating their emotions. Messing with their minds, in other words. I would really like to see if Facebook's argument would stand up in the court of law. And, as the thing seems to be going, I shall live to see it thus (PDF behind this link) tested, thanks to the good people of Epic.

Now, I can - sort of - understand how Facebook would not care - even understand - what and where the problem here is. They are, after all, a for-profit corporation and not a research organisation of any real scientific kind (even if they could afford to pay someone to teach them a few basic principles).

On the other hand, one has to truly wonder at the PNAS, who decided not just to publish the study on the word from Facebook that informed consent was obtained, but to then defend itself by essentially quoting Facebook's public position on the matter. To say nothing of the two universities who - one hopes - should now be busy reviewing their policies, procedures, and positions of the two "researchers" involved.

So, again, and in a nutshell: no explicit informed consent has been sought from the study subjects, ergo, no science took place or, better, no science should have taken place, let alone validated by publishing in PNAS.


For those in need of a bit of a further clarification: no, it is not necessary to obtain explicit informed consent to the exact target of a psychological experiment. In most cases, this would skew the results to the point of uselessness. However, and this is standard practice, such consent is sought for something that is unrelated yet revealing enough that the consent can still be considered as informed. Vague and overreaching T&Cs do not count. I am not an expert in the field by any stretch of imagination so I can't offer a qualifying counterexample, but I am sure the field of experimental psychology can be canvassed for a reasonable one in no time, and probably for no fee either.