Philip Dawid, William Twining, and Mimi Vasilaki, eds., Evidence, Inference and Enquiry

Philip Dawid, William Twining, and Mimi Vasilaki, eds., Evidence, Inference and Enquiry, Oxford-N.York, OUP/British Academy, 2011, 504 pages, ISBN 978-0-19-726484-3

This edited volume compiles seventeen papers developed in the framework of the Evidence programme, an interdisciplinary venture funded by the Leverhulme Trust and the ESRC between 2004 and 2007. Two of the editors, the statistician Philip Dawid and the jurist William Twining, led this project at University College London, in parallel to another research project directed by Mary Morgan at the London School of Economics (“How Well Do Facts Travel?”: the proceedings have also been edited in 2011 by Cambridge University Press). During those three years both UCL and LSE hosted many evidence-related events, bringing together scholars from many different fields, in what may have been –as Dawid suggest in his introduction (pp. 5-6)– the greatest achievement of the project. A survey paper by Jason Davies provides a good summary of those meetings. As to the accomplishment of the over-arching goal stated in the programme title (“Towards an Integrated Science of Evidence”), both Dawid and Twining are understandably modest, but they certainly had an intellectual agenda aiming at a unified treatment of all sorts of evidence. They found their inspiration (p.4) in the interdisciplinary theory of evidence developed by David Schum, an information scientist from George Mason University. One major flaw of this collection, in my view, is that it does not provide a précis of Schum’s approach, despite the explicit acknowledgement of its limited impact across disciplines (p. 93)

We just get a presentation of Schum’s bi-dimensional classification of evidence, since apparently it was the topic that most controversies elicited in the project. For Schum (p. 19), the three main properties of evidence are relevance, credibility and inferential force. For Schum information becomes evidence only when it bears upon a hypothesis (relevance), either directly or indirectly. As to credibility, it should be ascertained depending on the type of evidence we are dealing with (the main division being between tangible and testimonial). According to Schum, relevance and credibility would be the two main dimensions for classifying all kinds of evidence, since the inferential force –however we measure it– depends on a previous assessment of the former. Hence, whatever the hypothesis under consideration we can judge whether it is, for instance, testimonial and directly relevant for it. The classification would be purely formal (or “substance-blind” as Schum puts it). Many of the participants in the project considered controversial the possibility of one such general classification of evidence (framed in a scientific fashion), or so we gather from references in various papers in the collection, but, sadly, nobody takes issue directly against it.

Another contribution of Schum and Twining to the project is the recovery of the theory of evidence of John Henry Wigmore (1863-1943), an American legal scholar who defended the use of informal logic in law –the analysis of evidence in this tradition is brilliantly summarized in Twining’s paper. Wigmore represented legal arguments in form of networks, making explicit how evidence contributed to their cogency. In their respective papers, Terence J. Anderson and Peter Tillers, both legal scholars, use also Wigmore charts in order to analyze the role of generalizations in evidential arguments –the former in a case-based fashion and the latter in a more speculative manner. Schum and Dawid (together with A. Hepler) present a systematic comparison between Wigmore’s networks and Bayesian nets, applying both to analyze the evidential grounds of the conviction of Sacco and Vanzetti. I would have expected here some sort of attempt at integrating both approaches, but the authors just make explicit the differences, mentioning in a final footnote the possibility of combining both in a more formal manner. The contentious point might have been the resistance to quantification implicit in, at least part of, the Wigmorean tradition (p.86). Flexible formal approaches such as the evidential logic presented in John Fox’s paper might have provided a middle way, but Fox engaged neither Schum nor Wigmore and, unfortunately, his own paper was scarcely cross-referenced in the volume.

As I see it, David Lagnado pushes the farthest what seems to me the core intuition in the UCL project: evidence features in uncertain reasoning through piecemeal networks, exemplified at best in legal reasoning. In “Thinking about Evidence”, Lagnado, a psychologist, presents and defends the following theses. Bayesian networks can successfully model distinctive aspects of legal reasoning. They capture as well the qualitative networks underlying our cognitive processes (even if we cannot handle properly the probabilities in uncertain knowledge). In particular, they allow us to understand how we deal with fragments of such networks in evidential reasoning and how we adjust them to changes. Lagnado examines alternative approaches and presents two experiments on legal reasoning illustrating the viability of his own claims. 

The problem with this volume is that, interesting as some of them are, the remaining eight papers are a collection of independent contributions without explicit ties with any of the above mentioned topics. The narrative counterpoint within the programme is here represented by J. Russell and T. Greenhalgh, whose case study in discourse analysis shows how evidence is quantitatively framed in a health policy unit –without any explicit exchange with their fellow researchers. Drawing on a recent historiographical controversy, J. Davies ponders the pros and cons of analyzing historical evidence about ancient religions in terms of either ritual or belief. Again in a quantitative note and on legal topics, Tony Gardner-Medwin presents a short but very interesting plea for an explicit incorporation of our levels of uncertainty into our judgments (from students’ answers in exams to legal evidence in trials).

We also find here three papers by philosophers of science. N. Cartwright presents another installment of her theory of evidence for evidence-based policy (co-authored here with J. Stegenga), providing a very accessible presentation of her project. H. Chang and G. Fisher present their case for a contextual appraisal of evidence through an analysis of the ravens paradox. A. Wylie defuses the generalized skepticism about archeological evidence defended by several theorists in the field, applying ideas from Glymour and Hacking to show how different evidential sources in archaeology support each other in a non-circular manner.

From perhaps a too parochial perspective, I would complain about the quality of the two remaining papers in the collection. David Colquhoun, a biostatistician, contributes a paper “in praise of randomization” aimed, by all appearances, mainly against John Worrall. It turns out to be just a nice collection of illustrations of the virtues of this allocation procedure without an explicit engagement with the arguments of Worrall or any other objection against it (e.g., in Bayesian approaches such Howson and Urbach’s). In “What Would a Scientific Economics Look Like?”, the epidemiologist Michael Joffe shows his skepticism about the scientific status of mainstream economic theory through a comparison with biology. An enterprise perfectly legitimate as such, but with a very odd collection of references as intellectual ground: the methodologists he engages with (e.g., Friedman) are not anymore representative of the theoretical approaches Joffe is targetting. 

This collection shows that there is much more in the discussion of evidence, particularly in legal theory, than philosophers of science usually take into account – though, in point of fact, Larry Laudan, among others, took notice of it a while ago. The crucial point is to what extent the forms of argument in Law (in particular, trials) are comparable to the argumentative patterns in other fields, in particular science. As Twining reminds us though (p. 92), contested trials are just a step within the legal process and most of the cases discussed in this compilation are too narrowly focused on it. As a paradigm for an “integrated science of evidence”, I am afraid that this volume will leave many readers unimpressed. And this is partly due to the format chosen for this collection. As the editors warn, the three years of this project were not enough to form a coherent view of its goals among its members, as it shows in the final collection of papers. And I do not see the point today of publishing them all in a single volume as if there was some added value in binding them all together. A special issue in a mainstream journal plus a set of independent publications in specialized outlets, properly cross-referenced and all linked in the project’s webpage, might have been enough for most interested readers. 

{July  2012}
{British Journal for Philosophy of Science 64.3 (2013), 665-668}

No hay comentarios:

Publicar un comentario