The project: aims and objectives
The Arts and Humanities Research Board (AHRB)-funded research project 'Causality: metaphysics and methods' was launched in October 2002. Its principal aim is to investigate the applicability and prospects for the development of methods of causal inference formulated in recent philosophy of science. It pays special attention to the details of how these methods work in concrete contexts, primarily in the areas of econometrics, medicine and biology.
The research builds on that of Judea Pearl, who provides semantics for different types of connection, and on recent work by David Lewis and his students. Pearl's results are fairly tightly tied to determinism. Lewis's work relies at base on possible-worlds semantics and hence does not easily connect to methodology. Our research aims to push this starting work beyond determinism without sacrificing the close fit to methodology.
Cartwright, Nancy - Department of Philosophy, Logic, and Scientific Method, LSE and Department of Philosophy, University of California San Diego
Reiss, Julian - Department of Philosophy, Logic, and Scientific Method, LSE
Sober, Elliott - Department of Philosophy, Logic, and Scientific Method, LSE and University of Wisconsin, Madison
More information: Elliot Sober's Home Page
Worrall, John - Department of Philosophy, Logic, and Scientific Method, LSE
General developments October 2002-October 2003
At the end of the second year of the project, we have come a good deal closer to our aims as stated in the scheme of research.
Seven new papers have been published in our series Causality: Metaphysics and methods technical reports. A number of contributions to academic journals in philosophy and in the sciences have been published and others have been submitted for review. We continued our seminar series with roughly biweekly meetings during term, held one special workshop and organised an annual conference at the beginning of July with local, national and international speakers.
In the seminar series we maintained our practice of having a mixed programme of discussions of published research, presentations by group members and formal talks from outside speakers.
In the economics arm of the project, we had a discussion of James Heckman's econometrics (our work on Heckman is reported in two research papers: CTR 11/03 Uncertainty in Econometrics: Evaluating Policy Counterfactuals (PDF) and CTR 12/03 Practice Ahead of Theory: Instrumental Variables, Natural Experiments and Inductivism in Econometrics (PDF)) and a talk by the eminent econometrician David Hendry (Oxford). A version of Hendry's talk is forthcoming in the discussion paper series.
In the area of biomedical science, we organised a public lecture by Sir Michael Marmot (University College London (UCL)) on finding out about causal relations between socio-economic inequalities and various health conditions. We held a special workshop with Amartya Sen (Cambridge), Lord Richard Layard (LSE) and Marmot on reasoning about health and happiness. Basudeb Chaudhury (Caen), a project associate, reported research on a related topic conducted in France. Nik Rose (LSE) gave a paper on biological psychiatry and Donald Gillies one on causal mechanisms in medicine.
In a third area, which we call general philosophical and methodological background, we had discussion of new work on counterfactual accounts of causation. Jon Williamson (King's College London (KCL)), a project associate, gave a talk on Judea Pearl's work. Dan Hausman (Wisconsin-Madison) presented his paper 'Probability causality and practical causal generalizations' to the group. Tim Crane (UCL) presented his work on physicalism. Finally, Robert Northcott (LSE), also a project associate, gave a paper on defining causal strength, which has been published as CTR 14/03 Defining Causal Strength (PDF).
The conference took place on 2 July 2003. Julian Reiss spoke on Counterfactual Reasoning in Econometrics, Elliott Sober on Epistemological Holism and the Duhem-Quine Thesis: learning from the analysis of variance, Samir Okasha (University of York) on Multi-level Selection, Price's Equation and Causality (PDF) and Malcolm Pike (University of Southern California) on The Causes of Breast and Ovarian Cancer: exploiting this knowledge for prevention. About 40 scholars from LSE and elsewhere participated and discussions were lively. Okasha's paper is now published in the discussion paper series, Reiss's and Pike's are in the production process.
Specific aims and developments
In our view, we have made good progress in all areas of research.
With respect to the metaphysics of causation, Nancy Cartwright reports a novel view in CTR 07/03 Causation: One Word; Many Things (PDF), where she claims that the notion of cause piggy-backs on 'thick' causal notions such as compress, attract and smother and that the thick causal notions are not reducible to the concept of cause plus X, whatever that X may be. In CTR 09/03 From Causation to Explanation and Back (PDF), she gives a historical overview of the relation between causation and explanation, and in CTR 10/03 What Makes a Capacity a Disposition? (PDF) she relates the concepts of capacity and disposition.
In general, we moved slightly away from the methods designed to give answers to all causal questions (such as the Bayes' nets methods, counterfactual and invariance accounts) to applications in the special sciences. Research undertaken in the past year is reported under each heading separately.
Causality in biology
We were lucky to have Samir Okasha speaking at our conference. In his paper, he examines two alternative statistical approaches to modelling multi-level selection (the covariance approach and the contextual approach) and argues that they have different implications for what constitutes group selection and what does not. He shows that the contextual approach is philosophically preferable and discusses the implications on the frame-shifting debate and additivity. His paper has been published in our discussion paper series (CTR 13/03 Multi-Level Selection, Price's Equation and Causality (PDF)) and is forthcoming in the British Journal for the Philosophy of Science.
In a different context, Nikolas Rose discussed approaches to causal inference in biological psychiatry in a presentation for our seminar series. In particular, he was interested in how new styles of reasoning in the emergent discipline constitute new classifications, and in what counts as evidence for causal claims and for the existence of objects.
Causality in economics
Counterfactuals are at the heart of economic policy making. In order to select the best of an array of possible courses of action, policy makers need to know 'what would happen if' any of the alternatives were implemented. Unfortunately, there does not exist a satisfactory systematic methodology of how counterfactual claims are to be evaluated.
Julian Reiss and Nancy Cartwright try to fill in this gap. They discuss three approaches to the problem found in the econometrics literature (LeRoy, Heckman and Pearl), point out their respective deficiencies and generalise these results into a sketch of a methodology for evaluating policy counterfactuals. Their paper is published as CTR 11/03 Uncertainty in Econometrics: Evaluating Policy Counterfactuals (PDF) and will come out in a volume on truth and accountability in economic policy, edited by the National Bank of Austria.
It is part of standard methodological wisdom that large-scale experiments are impossible in economics for a variety of reasons. Julian Reiss argues in CTR 12/03 Practice Ahead of Theory: Instrumental Variables, Natural Experiments and Inductivism in Econometrics (PDF) that under certain conditions the instrumental variables technique in econometrics can mimic experiments and hence allow inductive causal inference. He also discusses the more general implications for economic methodology. The paper has been submitted to the British Journal for the Philosophy of Science.
Causality in medicine
This subject has enjoyed a large amount of attention during the past year, in particular in two areas. One area is a good example of how interdisciplinary research, characteristic of our project, can lead to fruitful results.
An important topic for health policy is how socio-political factors can affect the probability of suffering from certain medical conditions, for example how poverty can increase the likelihood of obesity, which in turn may cause diabetes, which itself increases the risk of heart failure. In some recent analyses, econometric methods of the kind we are studying in the economics branch of the project are used to establish causal relations.
The initial foundation for that research was laid in the public lecture by Michael Marmot that we organised (Life and Death on the Social Gradient). A conversation after the lecture led to a special workshop held in March in which Amartya Sen and Lord Layard discussed with Marmot the reasoning of health and happiness. Our research group also examined recent work on the topic in seminars.
Finally, a project associate, Basudeb Chaudhuri (Caen), started a systematic applied research programme at his home university on Social Inequalities in the Screening and Treatment of Cancer and reported first findings to our group in a seminar meeting (a paper reporting some of the results is also forthcoming in the series).
The other area concerns randomised clinical trials (RCTs), which form an important evidential foundation in the programme of evidence-based medicine.
Building on his previous research, John Worrall extended his criticism of 'randomisation as sine-qua-non dogma' to arguments from the literature on probabilistic causality. He reported his new results to the group in a meeting, and in another, we discussed various Bayesian arguments on the topic.
We were fortunate to have Malcolm Pike (University of Southern California) to speak at our conference and report on his research on breast and ovarian cancer, which largely proceeds without large-scale RCTs.
We will continue to have regular research group meetings in the third and final year.
For the Michaelmas term we already have a good line-up. Jon Williamson will present a new paper on 'Epistemic Causality'. Damien Fennell (LSE, a project associate) will present his research on Herbert Simon's work on causality in econometrics. Georg Theiner (Indiana University Bloomington) will present work on causality and dispositions. We are in touch with Trudy Dehue, a psychologist from the University of Groningen who has worked on randomised trials and may give a paper during this term. Gregor Betz (Free University, Berlin) visits the CPNSS for two terms, and we hope to persuade him to present his research on the limits of forecasting in econometrics.
Nancy Cartwright will become a 'participant-observer' in Marmot's group, looking at the causal relations that underpin health status correlations. She will also visit and work with the group of Princeton economist Angus Deaton who studies the same issues using economic methods.
Julian Reiss has in the past analysed a number of methods of causal inference in econometrics, which all suffer from the deficiency that either a large amount of background knowledge is required in order to make inferences possible or that they are applicable only to narrowly defined socio-economic systems or both. He plans to extend the research to methods that apply when these conditions are not fulfilled. A second branch of his research concerns singular causal conference. His work will build on the classic account from Max Weber as well as Adolf Grünbaum's views (the latter has been worked on by Nancy Cartwright, as reported in CTR 08/03 How Can We Know What Made the Ratman Sick? Singular Causes and Population Probabilities. An Essay in Honour of Adolf Grunbaum (PDF)).
Chaudhuri, Basudeb - Department of Economics and Groupe d'Economie Mathematique et Microeconomie Appliquee (GEMMA), University of Caen
Dash, Michael - St Thomas Hospital, London
Epstein, Philip - Department of Economic History, LSE
Gillies, Donald - Department of Philosophy, King's College, London
Mosini, Valeria - Centre for Philosophy of Natural and Social Sciences (CPNSS) and Department of Chemistry, University of Rome
Papineau, David - Department of Philosophy, King's College, London
Williamson, Jon - Department of Philosophy, King's College, London
Associated research students
Asgari-Targhi, Marzieh - Department of Philosophy, University of Sussex
Fennell, Damien - Department of Philosophy, Logic, and Scientific Method, LSE
Larski, Stanislav - Department of Philosophy, Logic, and Scientific Method, LSE
McKay, Phyllis - Department of Philosophy, King's College, London
Saaristo, Antti - Department of Philosophy, Logic, and Scientific Method, LSE
Stentenbach, Michael - Department of Philosophy, Logic, and Scientific Method, LSE
Virdi, Arhat - Department of Philosophy, Logic, and Scientific Method, LSE
Causality: metaphysics and methods technical reports
(discussion paper series)