Tuesday, October 4, 2022

Lost in translation

 A few days ago, the New York Times had an opinion piece by Huw Green, a clinical psychologist, which said "A clear causal link between psychiatric illness and gun violence has not been established..."   I followed the link, which was an interview with Ragy Girgis, a professsor of psychiatry at Columbia University.  That story had a caption saying "Findings from the Columbia database help dispel the myth that having a severe psychiatric illness is predictive of who will perpetrate mass murder."  It also contained a link to an article by Dr Girgis and others using the database (an attempt to compile a comprehensive list of mass murders since 1900), which said "the prevalence of psychotic symptoms among mass murderers is much higher than that in the general population (11% v. approximately 0.3-1%)."   That is, people with psychotic symptoms were between 10 and 30 times more likely to commit mass murder than people without psychotic symptoms.  

How did we go from 10 to 30 times more likely to "dispel the myth"?  The interviewer asked "Are people with mental health disorders more likely to commit mass shootings or mass murder?"  The answer started "The public tends to link serious mental illnesses, like schizophrenia or psychotic disorders, with violence and mass shootings. But serious mental illness—specifically psychosis—is not a key factor in most mass shootings or other types of mass murder.."  That is, it didn't answer the question that had been asked, but a different question:  whether most mass murders are committed by people with severe mental health disorders.   The answer to this second question is no, according to the information in the database.  But the answer to the question that the interviewer had asked was yes (at least for one kind of mental disorder, psychosis).  Apparently the interviewer didn't notice the difference, and followed up by asking "why does the public erroneously link mental illness with mass shootings and with violence in general?"

This could just be a case of miscommunication--anything involving probabilities can be confusing.  However, I think it's an example of a more general problem:  sometimes a focus on making sure that people don't draw the wrong conclusions comes at the expense of explaining what the research actually found.  I first noted this when writing  a post on a study of coffee consumption, where accounts emphasized a point that wasn't supported by the data:  that benefits only occurred with moderate consumption, not high consumption.  I saw another example later in the summer, when a study of diet and exercise was described as showing "that healthy eating and regular workouts do not, in isolation, stave off later health issues. They need to be done together."  In fact, the study suggested exactly the opposite--exercise and diet had additive effects on mortality, and no interactions were found.  The reason seemed to be a goal of getting people to think of exercise as a way to improve one's overall health rather than a way of getting away with a bad diet--"the study highlights the importance of viewing food and exercise as components of holistic health, Dr. Ding said, instead of calculating how many miles can 'cancel out' a cookie."  With mass shootings, there's concern about increasing the stigma against mental illness. As these examples suggest, the problem is more prevalent in medicine and public health, where researchers know that people might use their findings to make decisions.   

4 comments:

  1. IMO this is an understandable explanation but I'm not sure it's accurate. In my view, the critical question isn't about the *probability* of people with mental illness committing mass murder. The critical question is question, rather, is whether "severe psychiatric illness is predictive" - the key word is **predictive** - of who will perpetrate mass murder. It's not clear if the fact that one in ten people (11%) with "severe psychiatric illness" are likely to commit "mass murder" constitutes a **predictive** condition. I think most groups of ten people in which one of them has committed a crime would feel justifiably wronged if some sanction were imposed on all of them because of the crime of the one. So, in fact, it's probably safe to say that most people would not accept 1:10 odds of an offense as a "predictive" condition.

    I suggest that the real issue isn't about preventing stigma for the mentally ill. There is already substantial stigma and that will - probably to some degree justifiably - remain. The real issue is rather preventing the unjust punishment of innocent people. It's true that the research was incorrectly described, but this was a muddling of words. If the original question was actually **predictability** - a much more concrete concept than mere probability - it's appropriate to say that severe mental illness *does not* constitute a predictive condition with regard to mass murder, and thus the article flubbed some aspects explaining the research but *did not* send the wrong message about predictability.

    ReplyDelete
  2. The problem is that "predictive" doesn't have any definite meaning. If the rate of some action differs between groups, which it does, than knowing which group someone is in does help to predict whether they will do it. Of course, with mass shootings the chance is very low for both groups. But if this is the point someone wants to make, they should just say it directly.

    People also sometimes say that mental illness isn't (or rarely is) a "primary cause" or a "root cause," or a "key factor" (as in the quoted passage). But like "predictive", these don't really mean anything.

    ReplyDelete
  3. David: I agree, the lack of specific meaning of "predictive" and "primary cause" are quite a problem. The discussion should have identified this problem, so I guess there was quite a bit of muddling of different concepts and ideas.

    I suppose the question is then: what sensible meaning could these terms **inherently** have? IE what meaning could be ascribed to them that could be widely accepted, that people could accept as what they actually mean? I suspect odds somewhere between 7 and 9 in 10 might be high enough to persuade the general public that a condition is "predictive" or a "primary cause", although I don't know what the exact level would be. But I do think odds of 1:10 are probably too low for the general public to accept as a form of conviction, especially given the overall poor reputation of psychological sciences.

    ReplyDelete
  4. Or it could be that the New York Times s trying to defend the official narrative that guns fire themselves whilst being held in the hands of innocent individuals.

    ReplyDelete