In need of a framework

Following the initial dinner at the causal inference workshop, it is clear that there are both insights and mistakes flying around. Epidemiologists are way ahead of philosophers in some respects, especially in their appreciation for the varieties of causal situation that can arise, and in their treatment of general causation and especially quantitative treatments of it. On the other hand they make some persistent errors, some of which a philosophical training would prevent, and some arising from the intrinsic difficulty of the subject. The problem, and perhaps the place where philosophical input could be most useful, is that the issues aren’t clearly partitioned, and discussions roam from one point to another, with interlocutors not clearly recognizing the logical links, or lack of them, and thus ultimately not able to separate their errors from their insights. Perhaps what I have just described sounds like many philosophical discussions too, but philosophers do characteristically–to a fault, perhaps–distinguish issues. “Ah, but that’s a different question,” is a common retort.

In the present debate, a framework within which certain issues could be separated from others, some parked and some revved up, would be very helpful. I’m not sure how to achieve it, without a huge synthesic survey of everything that has been written. Perhaps that is the only way.

How much do I owe Pinocchio?

I suspect that I would not have made it through my PhD without the institution pictured below:

During “Fall” 2006, when I was a Visiting Fellow at the above-pictured bus’s destination, Pinnochio’s was my lunchtime sustenance, the thing I looked forward to all morning as I sat in the library of the Philosophy Department wondering whether the debate between Stalnaker and Lewis on counterfactual excluded middle had any bearing on Schaffer’s contrastivist treatment of causal selection.

But wait: when I say that I would not have got through without Pinnochio’s, what exactly am I envisaging? A situation where Pinnochio’s was there but I never discovered it? A situation where it had never existed? –or where perhaps it closed just as I found it, like my second favourite, Burritos? (I still sometimes wonder what might have been.) Or perhaps a situation where Pinnochio’s existed, but didn’t do such good pizza; pizza such as this:

It is interesting to be back in the same place thinking about the same questions ten years later, and satisfying to have come quite an unpredictable route, both intellectually and geographically. In 2006 I had never heard of epidemiology, much less contemplated working on it; and when I got into epidemiology, I thought I had decided to move away from working on counterfactuals and causation. It is curious how things turn out.

I am here for a Radcliffe Institute workshop mostly centering around an upcoming special issue of the International Journal of Epidemiology on causal inference. Previously I posted links to papers in IJE earlier this year, and the special issue promises to be really interesting–a real resource, for sociologists, philosophers and perhaps historians, as well as for epidemiologists.

On the long journey from Johannesburg to Boston I read or re-read as much as I could of what has been published or written on this during 2015 and 2016. I can see some real progress. It is clear that nobody intends to restrict causal claims to those that can be relativized to a humanly feasible intervention, for example. It is also clear that at least some of the debate concerns matters of presentation. It is important to me, for example, not to be cast as a methodological Luddite who insists on putting verbiage in the way of honest, hardworking methodologists who just want to get on with the job. It is important to those methodologists, on the other hand, not to be cast as any sort of methodological fascist, intent on forcing everyone to employ a preferred set of methods, and excommunicating everyone else.

On the other hand, serious questions remain. From what I have read, the main ones are as follows.

  1. What exactly are the commitments of the Potential Outcomes Approach? Is it just a set of methods? Or something more? And if more, exactly what? I have been trying to understand this since I first started working on the topic, and in the latest paper by Vandenbroucke, Pearce, and myself, we try to set the elements of the position out. I will be hoping to find out if we got it right.
  2. The distinction between causal identification and quantitative estimation of effect size has become central. Does it bear the weight that is put on it? Does it indeed offer a domain in which the POA really can delineate either necessary or sufficient (or both) conditions for a causal inference? How are identification and estimation related?
  3. Are there sufficient conditions for causal inference, of any kind, and if so what kind of sufficiency is it?
  4. What exactly is the role of judgement? Everyone agrees it is needed. But what exactly is it, and how is it to be trained, if formal methods do not take us all the way? Although I am always pushing on the need for judgement, I have great sympathy with the formalising perspective. After all, judgement isn’t magic. If we can do it, there must be something we are doing; and if we could only describe this, we could automate it, refine it, improve it, and so forth. Yet we can’t even say what it is we are doing when we make a causal (or other inductive) inference. How can this be? An old question, obviously, but one that I think is becoming more pressing in the age of technicality.
  5. What is the broader significance of the developments of the last ten years that fall under the POA? Are they part of the development of the science towards what Kuhn would call puzzle-solving? Part of the appeal, if not the motivation, of the new methods seems to be a desire for formalising causal inference. Is this reasonable? Achievable? Healthy or not? Kuhn’s view is that scientists learn by rote, and in order to learn by rote there needs to be something to learn. A methodology that says “think about it and use your judgement” does not lend itself to this. Are these developments partly about getting epidemiology into a “normal science” phase? Of course we don’t have to believe Kuhn was right about science, but I continue to find his lens useful.
  6. What is the social and political dimension to this debate? Will the new methods render difficulty variables like race more tractable? Or will the POA enforce a certain view of them, perhaps by encouraging us to think of race as a genetic trait and not as a complex socio-historical role? Further (and this hasn’t been explicitly asked, but it occurred to me on reading) does the development of a set of methods for measuring effect sizes of interventions in increasingly difficult situations reflect the social and political interests? — certainly not of the methodologist in question, of course; I mean the larger context in which this sort of project is undertaken at Ivy League universities, gains traction, gives rise to intellectual celebrities, and so forth.

The last question in particular is uncomfortable, since it sounds unpleasantly personal. It really isn’t personal; it is not any sort of attack, and indeed it extends to this very debate, and to myself.

The two churches between my Sheraton and Pinnochio’s have signs outside saying “BLACK LIVES MATTER”. A helpful reminder, since I didn’t see many black people on my walk. Most of those I did see appeared to be homeless, so the signs must be quite affirmative for them. The other diners at Pinnochio’s were all white men: two professors having a long conversation about a running conflict with an absent colleague, and three students probably still in their teens, two of whom were in suit and tie. I think they were choristers. I was feeling right at home, until one of the youngsters said to another, “How many suits do you have?”

After a pause, the other said “Six or seven.” The way he said it I could tell he had at least twelve.

I have three, only one of which both is clean and fits me.

Had I not discovered Pinnochio’s ten years ago, perhaps two of them would fit me.

#PSA2016 Symposium: Measurement and Causality in Medical Science

Olaf Dammann, Leah McClimans, Zinhle MnCube, Benjamin Smart and myself hosted a symposium of this title at the Philosophy of Science Association Biennial Conference in Atlanta. Olaf couldn’t make it but was there in spirit. All four papers will be part of an upcoming collection on medicine in medicine, edited by Leah McClimans. Proposal and slides attached.






Letters and authors’ reply on Potential Outcomes paper published (IJE)

Two responses have now been published to the Vandenbroucke-Broadbent-Pearce paper on Potential Outcomes in IJE (Jan 2016).

We have replied to these letters (open access).

As well as this exchange of letters, the IJE has asked for some fuller commentaries on our paper, as well as our response to these commentaries. The exchange is developing in an extremely useful way, in my view, and I look forward to being able to share the next “round”, perhaps later this year.

Our original paper is here.


Paper “Health as a Secondary Property” accepted by BJPS

Subject to a few final adjustments my paper “Health as a Secondary Property” has been accepted by the British Journal of Philosophy of Science.

I’m pleased because this is my first contribution to the debate in the philosophy of medicine about the nature of health. I’m also pleased because the paper is a proving ground for some ideas that I want to develop further in my book on philosophy of medicine. And of course I’m happy because BJPS is a great journal.

It may change in minor ways, but here is the abstract. Obviously I can’t post the full paper here but if you are interested in seeing a pre-print privately, please contact me.


In the literature on health, naturalism and normativism are typically characterised as espousing and rejecting, respectively, the view that health is objective and value-free. This paper points out that there are two distinct dimensions of disagreement, regarding objectivity and value-ladenness, and thus arranges naturalism and normativism as diagonal opposites on a two by two matrix of possible positions. One of the remaining quadrants is occupied by Value-Dependent Realism, holding that health facts are value-laden and objective. The remaining one, holding that they are non-objective but value-free, is unexplored. The paper endorses a view in the latter quadrant, namely the view that health is a secondary property. The paper argues that a secondary property framework provides the resources to respond to the deepest objections to a broadly Boorsean account of natural function, and so preserves the spirit, though not the letter, of that account. Treating health as a secondary property permits a naturalistic explanation of the health concept, specifically, an evolutionary explanation, in terms of the assistance such a concept might have provided to the survival and reproduction of organisms that have it. (This approach is completely distinct from evolutionary and etiological accounts of natural function.) This provides the otherwise missing explanation in Boorse’s account for the fact that function is determined with reference to contribution to goals of survival and reproduction, assessed relative to classes comprising age of sex of species, rather than some other equally natural goals or classes.