On the Facebook research scandal
There’s a lot of people in a lather about the Facebook research, in which 600,000 people were fed happy content, and 600,000 sad, so as to find out whether some people being happy on Facebook made other people sad. It’s unethical, says a law professor, because Facebook didn’t get consent*. Alex is so incensed he’s going to give up Facebook.
This makes me wonder why there isn’t the same big lather about Randomised Control Trials (RCTs), which can actively do down the life chances of participants, without those participants ever knowing what happened and – even if they do know they’ve been in an RCT – unable to reverse the effects. As this government paper recommending the widespread use of RCTs says:
Sometimes interventions which were believed to be effective turned out to be ineffective or even actively harmful. This can even be the case with the policies we think will be guaranteed to work. For example, incentives have been used to encourage adult learners to attend literacy classes, but when an RCT of this policy was conducted, it was found that participants attended approximately 2 fewer classes per term than the non-incentive group…..we could have done harm to the adult learners with the best intentions, and without ever knowing we were doing so (p.17).
True, RCTs tend not to be designed so that intuitively harmful things are done to people, but then the whole point of them of an RCT is that they may do harm, however counter-intuitive that may be to those who propose the study in the first place.
So how exactly is the Facebook research more unethical than widespread government research and consequent social policy? Or have RCTs just become so much part of the social policy furniture that no-one questions them anymore.
Maybe, if this question is provoked, the Facebook research will turn out to be massively socially useful, just as the research team intended, I’m sure.
* The issue of informed consent is course vital, even though bad things happening to people are still bad things, whether or not they are consented to.
In the adult learner incentive RCT referred to in the government paper linked to above, consent from individuals was obtained (I looked it up). However, in some RCTs, especially clutster RCTs, it becomes impractical for all individuals to provide informed consent and ethics committees have to weigh up the pros and cons, using available guidance from national and international-level experts who make the ethics of RCTs their life’s work.
So, for example, I’m not sure from the literature whether the government funded RCT into whether texts sent to people who owed fines to courts sought consent from all those who ended up as subjects of the study. More seriously, in the ongoing RCT into the Safeguarding Assessment and Analysis Framework (SAAF) child protection assessment tool, conducted at cluster level with social work teams as the unit of randomisation, it’s easy to see how difficult it would be to get informed consent on all families (and who in the family?) randomly assessed or not assessed using the tool being evaluated, especially as a decision not to give informed consent would then presumably provide social work teams with a tricky decision, both in terms of ethics and management, as to whether they should use the tool or not. Arguably, therefore, we have a trial which actively harms children, without their or their families’ knowledge, by dint either or using the tool at (cluster) random, or not using the tool at random.
This is in no way meant to be a criticism of the evaluators commissioned by DfE, who I’m sure have gone through all the correct ethical processes, and probably a few steps beyond too. I raise it simply because it highlights the tricky issues with consent. FWIW, I’ll be looking out for the results of this RCT with interest, as my hunch is that the use of the SAAF tool may turn out to be less useful than its promoters suggest it is, compared (say) to flexible and autonomous professional judgment which moves beyond the parameters of the its ‘stages’ which, of themselves, are grounded in contestable assumptions about what child protection is and should be.
What the RCT will show up will depend to a great extent on how the desirable outcomes are defined (what, after all, is a protected child, and is that the same as a happy one with a good chance of growing up happy?)