Monday 18 May 2015

Are academics open to hypotheses?

I once attended a garden party and found myself next to Kate Adie, a notable BBC reporter who had covered the Iranian Embassy siege in London, and then subsequently a wide range of international affairs, mostly in war zones.

At this particular time the big domestic story was the surge of a group of 121 children in Cleveland, UK who had been diagnosed as having suffered sexual abuse at the hands of their parents or relatives.  This was based on a particular, non-standard, diagnostic procedure called the “anal dilation test”. Opinion was split between those who doubted that so many parents and care-givers would bugger their children, and those who felt that child abuse was widespread and unacknowledged. The local authority believed the test was sound and removed the children from their parents, and put them into social care. A purely scientific approach would have been to be open to the hypothesis that child abuse might be more prevalent than previously estimated and then looked at the supporting evidence in a critical light. What was particular about the case was the extremely elevated estimate of anal intercourse implied by the test administered by two paediatricians. Anyway, at the parents eventually obtained a court judgement that the children could come home.

The BBC, Kate Adie said, had just put this as the lead item on the news as “Children re-united with their parents”. I thought about this, and it seemed a perfectly fair description to me. “Although I am not on duty” Kate continued “I rang up the news desk to correct them”. I was still bewildered, but kept listening. “The News should have said ‘The children were returned to their parents”. Finally, the penny dropped. We would not have said “The children were reunited with the paedophiles who had been abusing them”. “Returned” was factual, “re-united” implied a happy family, and that it had been wrong to take the children from their families.

Leaving aside the later knowledge that the supposed diagnostic test was deficient, this revealed to me the power of unexamined assumptions. Bias is most powerful when it seems reasonable and universal. In social science most researchers and most of the student audience share particular assumptions about society, about causal variables and even about values. In American academia, the most influential in the world, about 80% are of Liberal persuasion.  From a high minded perspective this should not matter, because methods should be strong enough to counter all assumptions. If researchers are open to all hypotheses, then all explanatory possibilities will be examined with equal rigour and rectitude. However, that is not always the case.

Much of the literature on maternal deprivation made light of possible genetic confounders. Some of the developmental literature avoids genetic group comparisons. As far as I know, no one is repeating Dan Freedman’s observational work on neonate behaviour, showing profound differences in behavioural reactions in the first days of life.

1 comment:

  1. I like your Adie story. I used to insist my students use objective language: effect X was not "improved", it was increased or decreased. Some of them never really understood, I'm afraid. But then I spent a lot of time in a world where most research students weren't remotely as bright as the brightest of the undergraduates. But maybe almost every academic works in such a world?