Mostrando entradas con la etiqueta 2007. Mostrar todas las entradas
Mostrando entradas con la etiqueta 2007. Mostrar todas las entradas

15/4/09

A. L. Stinchcombe, The Logic of Social Research. Chicago and London: The University of Chicago Press, 2005.

Arthur L. Stinchcombe is a very distinguished sociologist and is well-known, among other things, for his methodological work. He decided to publish a monograph entirely devoted to methodology only at seventy though. This is thus a mature work, in which Stinchcombe takes stock of a lifelong career as a practitioner sociologist and attentive reader of both philosophy and other social sciences. Yet its intended audience are beginning graduate students (in various disciplines, not only sociology) to whom the lectures in which this book originated were addressed. He offers them “an upside-down version of the ‘unified science’ movement” of his youth. Instead of putting physics first as a yardstick, Stinchcombe issues a plea for a more unified social science whose variety of research methods, well understood, would exhibit the “same intellectual strategy” than any other scientific discipline. Thus one can, perhaps, explain the classical resonance of the title (e.g., Logik der Forschung). Yet, the approach is truly updated.

Having introduced Jon Elster as “probably the leading philosopher of the social sciences” (p. 171), Stinchcombe adopts a very Elsterian piecemeal approach to a big issue, causality, which is explored by means of a series of case studies throughout the book. The cases illustrate at length four methods and seven problems that are covered in it. As to the former, Stinchcombe considers the quantitative, the historical, the ethnographic and the experimental. The problems are: the centrality of distances in study design for causal theories; economy in data collection; the use of data to refine concepts and their measurement; the relevance of context; the empirical research of mechanisms and processes; contrasting theories through hypothesis testing and, finally, the use of data to refine them. Each of these is considered in a separate chapter through the analysis of an array of case studies, where the contribution of the four methods is exhibited and analysed.

The first question then is why causality plays such a central role in the logic of social research. Immediately on page 1, Stinchcombe states that almost all sociological theories are, one way or another, causal. Yet, we will find no philosophical discussion of the concept of causality throughout the book, since Stinchcombe explicitly eschews digression into epistemological topics (e.g., p. 196). Stinchcombe’s approach is methodological instead: there various methods of social research that claim to reach causal conclusions, and the questions are, first, under which conditions they yield these and, second, whether there is any complementarity among them. Stinchcombe’s approach seeks the one in the many: instead of arguing in favor of one particular method as the sole means to assess causality, it is claimed that all four methods can contribute. Obviously, it is necessary to understand causation broadly so that ethnographers and econometricians can cooperate in the pursuit of causal explanations. The author states that “the minimum piece of causal information is two distances”, i.e., some variation in the information we can gather about the units of analysis involved in the causal process. If “a year more of education” is regarded as putative cause of “a three-point increase in a measure of labor market advantage”, the minimum piece of causal information will then be the distance between two observations of education in different people. This approach suggests that indeed all the four methods considered rely on quantitative considerations, namely the measurement of numerical differences in the units of analysis, or lead us to obtain them. As a matter of fact, the chapters exploring the five first problems discussed above are somehow intended as an extended preamble “of a good statistics textbook” (p. 239). At this point, it is necessary to note that the author is implicitly assuming that those social theories that do not produce causal statements do not deserve the consideration of science. I will not question such assumption, but it should have been argued somehow, at least to warn the reader of the consequences of this one-sided causal approach to social research.

Having settled this, Stinchcombe proceeds to explore different research strategies. In chapter three he addresses the problem of costs in the design of a study, stating the general principle that we want “the cheapest right answer”. On this basis, he defends the convenience of oversampling extremes by focusing on those cases “that contain more information on why they are not average”. Thus, they are judged to be more relevant to explore causal distances. Given that these are often initially grasped through vague concepts, in chapters four and five Stinchcombe discusses various uses of data to improve their precision. He explores, first, whether one or more concepts are needed to describe causal distances; second, how to interpret them in terms of causal theories or mechanisms and their contexts. This latter topic is addressed at length in chapter five, where the relevance of boundary conditions for causal analysis is appraised. Information about the context is more or less needed depending on how precisely we can define our units of analysis. When this definition is difficult to attain, narratives providing background information seem to be the main device to improve our understanding of the causal setting through a better grasp of its boundaries. The core chapter of the book, in length at least, seems to be the sixth, about mechanisms, that is, our theoretical understanding of the repetitive processes occurring around the units of analysis which turn causes into effects. Stinchcombe lists and illustrates five types: structural holes (such as social networks), individualism, rational choice (carefully distinguished from the former: there is no token individuality in it), situations and patterns. Chapter seven is about statistical tests, among which Stinchcombe pays particular attention to the role of hierarchical models. The last chapter is, in many ways, “a sermon on the life of research”, the crafting of theories through data as a research program.

This synopsis conveys just a vague impression of the topics covered in the book. The reason is that most of the concepts listed above are not introduced and discussed analytically, but rather exemplified at length through ample transcriptions or summaries of articles and book chapters, very often authored by Stinchcombe himself. He wants to show the four methods discussed at work, and it is shown indeed that there is a complementary role for each in many of the cases presented. The virtue of this casuistic approach seems to be in its utility in class: the examples are detailed enough to allow a discussion of the concepts and methods they illustrate. The interest of the cases selected is obviously guaranteed by the author’s expertise. Yet, it is a reflection of his own (necessarily partial) interests and, as such, several topics are underrepresented or simply omitted (e.g., experiments or simulations). The author’s warning regarding the rhetoric of the book should be taken into account: it is the oral style of a lecture (p. xi) and, given the number of cases discussed, it is easy to get lost in the details. It seems better to use it as a trigger for debate on each particular topic in class than as a conventional textbook.

A different issue is whether this selection of cases sustains the thesis that Stinchcombe is defending, namely the complementarity of the four methods in the search for causal accounts of social phenomena. In my view, the proclaimed methodological unity of the social sciences seems a bit vague, probably because the very concept of unity was not intended to capture the family resemblances that Stinchcombe’s discussions suggest. The stricter we are in our definition of causal explanation, the less unity there will probably be among the cases discussed in the book. However, it was not aimed at convincing philosophers of any matter of principle, but to make students more open to the variety of available methods and the convenience of using them. As such, it deserves all praise. In this respect, I only miss that the author has not summarized the debate presumably elicited by many of the studies compiled, just as a sign of the professional reception that was granted to the attempts to integrate various methods. This would have been useful for students and for philosophers alike as an index of the real unity that may exist if not in the social sciences, among the social scientists at least.

{November 2007}
{Philosophy of the Social Sciences 38.2 (2008), 296-298}

14/4/09

Serena Olsaretti, ed., Preferences and Well-Being [Royal Institute of Philosophy Supplement: 59], Cambridge, Cambridge University Press, 2006.

In 2004 a conference took place in Cambridge sponsored by the Royal Institute of Philosophy on preferences and well-being. Drawing on the papers presented therein, Serena Olsaretti has prepared with great care a volume that, according to her, is structured around three different sets of questions. Namely, the formulation of normative and descriptive accounts of preference-formation; whether preferences conform with requirements of rationality and what reasons can support them; and finally the normative significance of those preferences that do not meet such requirements, in particular for policy-making purposes. Five papers deal with the first topic, and there are three more for each of the remaining two. So far for the unity of the collection. Its most interesting aspect lies, as usual, in the divergencies. Let me try then another classification.

First, there seems to be quite a divide regarding the theoretical approach to preferences. Whereas the first four papers (by Arneson, Rosati, Brännmark and Qizilbash) apply pure conceptual analysis almost without positive asides, the rest of them stay more or less close to Rational Choice Theory (RCT) in their discussion of preferences. The first set of papers provide a good sample of an ongoing disciplinary debate among moral philosophers about the human good and whether this should be defined in terms of preferences satisfaction or rather by a list of objectively valuable goods –or something hybrid. Central to this debate is the proper formation of preferences: under which conditions our desires will be able to match our conception of well-being. Depending of our conception of the latter different issues will gain or loose prominence. By way of example let us just mention a few ones discussed in this set of papers: information as to the alternatives, motivational force, parental guidance, authorship as to one’s own life, etc. Though informative and interesting, given the formation of my own preferences, I find quite problematic the assumption that these four papers more or less take for granted: that the empirical processes of desire formation are somehow congruent with their normative discussion.

A good measure of the difficulties with this assumption is the contrast between this array of papers and a second one in which the discussion turns around RCT, exploring its conceptual foundations as to the concept of preference. First of all, Hausman and Pettit take issue with one or another aspect of our common understanding of RCT. The former addresses a default principle implicit in game theory, that individuals prefer a comprehensive outcome (in Sen’s terms: the outcome as seen from the path through a game in extensive form that yields it) to the same extent that they prefer its actual result (dissociated from that path). When this principle collapses, consequentialism fails. Pettit argues for a more complex idea of preference based in deliberation. In this account, RCT appears as dealing with a rather restricted case (self-interested tastes, as exhibited by our species and many others).

In other words, if an intendedly positive theory (RCT) only empirically meaningful under such constraints, we may wonder why do we expect better of an abstract examination of the concept of preference, such as the one attempted in the first set of papers. Piller and Broome illustrate a more parsimonious approach to conceptual analysis in continuity with the idea of preference exhibited in RCT. Piller explores the desirability of having a desire and whether we have any reasons to justify it. Broome argues instead that we should reason over our preferences in terms or their content rather than on any second-order requirements on their desirability.

Yet, RCT can be equally contested as an approach to preferences on a purely empirical basis. Here is a third divide, represented in this volume by the papers of Sugden, on the one hand, and Sunstein and Thaler, on the other. They all take issue with the experimental failures of RCT, though with a different aim. Sugden proposes a model of unconsidered (neither coherent nor stable) preferences trying to capture the normative value of satisfying them as they are. Sunstein and Thaler defend a libertarian paternalism, in which the empirical failures of individual rationality would justify the framing of public choices in a way that would paternalistically favour the interests of the agents, despite their incapability to grasp it at first (for instance, opting in or out of insurance schemes). Finally, Voorhoeve draws also on preference change to contest those conceptions of welfare based on preference satisfaction.

Therefore, given that this wonderful conference brought together all these approaches, I would have expected a more explicit discussion of these theoretical divides. Whether RCT provides a better (or worse) framework for the normative discussion of preferences than pure conceptual analysis. Whether we should try to improve the formation of our preferences through RCT, given its experimental failures. And whether we can attribute any normative significance at all to these failures. But these are just my preferences, not a list of objectively valuable questions. Yet, the list of papers compiled is valuable enough to be widely read and discussed.

Edith Dudley Sylla, ed., Jacob Bernoulli, The Art of Conjecturing, together with Letter to a Friend on Sets in Court Tennis, Baltimore, The Johns Hopkins University Press, 2005.

La Historia de la probabilidad no es una disciplina de creación reciente (en 1865, apenas dos siglos después de su “nacimiento”, Isaac Todhunter publicaba el primer tratado sobre su evolución), pero su explosión conceptual tiene apenas treinta años. En 1975, Ian Hacking publicaba La emergencia de la probabilidad, al que seguirían, por un lado, los trabajos de Stephen Stigler (1986) y Anders Hald (1986, 1998) y, por otro, los de Ted Porter (1986, 1995). Sus autores son todos especialistas de primer nivel y ante sus discrepancias no cabe acusación de ignorantia elenchi: sus desacuerdos se originan principalmente en sus respectivos enfoques. A Hacking le preocupa la genealogía de las distintas concepciones filosóficas de la probabilidad, mientras que Stigler y Hald se ocupan ante todo de la construcción de su aparato matemático; la de Porter es, ante todo, una Historia social ¿Cómo escoger entonces entre estos enfoques? Buena parte de la respuesta se encuentra en la edición de los clásicos sobre los que se sustenta el análisis. Pues aunque todos los autores citados manejan ediciones originales y a menudo también material de archivo, queda aún mucho trabajo filológico por realizar y cabe esperar que nos proporcione evidencias más ajustadas para juzgar las distintas interpretaciones hoy disponibles sobre los clásicos.

Un magnífico ejemplo de este trabajo nos lo proporciona aquí la edición que nos propone Edith Sylla de la obra fundacional de Jacob Bernoulli, su Ars conjectandi (1713), la primera versión completa a una lengua moderna –el original latino podía manejarse desde 1975 en sus Obras Completas. Sylla es muy explícita respecto a su propósito crítico: su trabajo supone enmiendas puntuales o generales de la interpretación de Bernoulli servidas por Hacking y sus continuadores. (e.g., p. x, nn. 7-8; p. xvii, n. 27). Es más, nos exige revisar la propia recepción de la obra a lo largo de los tres últimos siglos. Al conocerse sólo parcialmente (principalmente, la parte IV, donde se encuentra la demostración de su ley de los grandes números) muchos confundieron las tesis de Bernoulli con las de sus expositores (principalmente, De Moivre), oscureciendo, en particular, su concepción de la probabilidad.

Sylla no efectúa enmiendas al texto latino, establecido ya en las Obras Completas, sino que se concentra en cómo trasladarlo a un inglés que recoja su sentido. Para ello, combina su propio análisis conceptual con el vocabulario de numerosos textos de la época. Sylla no contrasta su versión con las traducciones francesas, alemanas o italianas (todas parciales) que recoge en su bibliografía –en la que omite, por cierto, la castellana de Andrés Rivadulla: Llull 30 (1993)–, pero, dejando aparte el interés polémico de la comparación, sus opciones quedan sobradamente justificadas. Quizá el principal empeño de Sylla sea mostrar que el léxico de Bernoulli debe interpretarse desde una concepción normativa de la probabilidad articulada sobre el valor esperado de una apuesta como medida de su equidad (y no del grado de aleatoriedad del proceso). Esta lectura es bien conocida desde los trabajos de Daston (1988) y Franklin (2001), pero casa mal con la imagen de Bernoulli como simple precursor de las leyes de los grandes números. Pero, si no empleó un concepto de probabilidad asimilable al nuestro, ¿qué se probaba entonces el teorema fundamental de Bernoulli?

Para averiguarlo, el lector puede acudir al amplísimo estudio preliminar y comentario (de unas 150 pp., sobre las 430 de la obra) con el que se acompaña su traducción. Entre las perspectivas anteriores, Sylla opta por minimizar el comentario matemático, contentándose con elucidar los distintos pasos del análisis bernoulliano, y evita digresiones filosóficas sobre el sentido de la probabilidad para concentrarse en la genealogía intelectual de la obra desde su contexto social. Sylla reconstruye así la gestación del Ars Conjectandi, usando con destreza un amplio repertorio de fuentes ya editadas y algún material de archivo. Un mérito no menor aquí es el de apoyarse en una buena colección de citas de textos latinos que acompaña de su traducción inglesa. Quedan así dilucidadas cuestiones tan diversas como las circunstancias de la impresión original del texto, debida a su hijo antes que a su sobrino Nicolas (autor del prefacio), o la superposición en él de resultados acumulados por Bernoulli durante más de dos décadas.

A diferencia de muchos de sus coetáneos, preocupados casi exclusivamente por los juegos de azar, Bernoulli, con Leibniz, supo apreciar el uso social que podía darse al cálculo de probabilidades para aplicarlo a cuestiones cívicas, morales y económicas (según reza en el título de la IV Parte). Bernoulli no llegó a realizar su empeño, pero queda por explicar su origen (que, apunta Sylla, podría estar en su pertenencia a una familia de comerciantes) y, sobre todo, su influencia en la posteridad, que aquí queda inexplorada. Una de las posibilidades que abre esta edición es, justamente, la de reconsiderar la contribución de Bernoulli a los orígenes de la matemática social que florecería en el XVIII. A mi juicio, la principal contribución exegética de Sylla es la de mostrarnos todo lo que le separa de esta tradición ilustrada: el autor del Ars conjectandi sería antes un teólogo que un científico social.

Es sabido que Bernoulli cursó estudios de teología en Basilea con la Reforma ya consolidada, como suele recordarse cada vez que es necesario explicar algunas alusiones teológicas, principalmente el párrafo al comienzo de la parte IV (Bernoulli, 1713, 210-11) sobre la omnisciencia y omnipotencia divina. Como la propia Sylla explicó ya en dos magníficos artículos anteriores a su edición, la predeterminación de cualquier acontecimiento pasado, presente o futuro sirve como garante de que su número es fijo y puede ser por tanto aproximado empíricamente a través de la ley de los grandes números, tal como Bernoulli originalmente la concibió. La expansión binomial con la que construyó su demostración representaría los distintos modos en que se puede dar un acontecimiento en el mundo, como si todos ellos estuviesen ya dados intemporalmente. Nosotros seríamos incapaces de percibirlos así, pero Dios podría. No obstante, el teorema nos serviría para justificar que somos capaces de obtener buenas aproximaciones a esas razones intemporales entre acontecimientos sobre la base de frecuencias empíricas.

La aparente paradoja de esta posición es que, por un lado, el cálculo de probabilidades se nos presenta como instrumento para mejorar nuestras decisiones pero, por otro, se afirma que estas ya están previstas en el plan divino de la creación. ¿Cuál es entonces el valor moral de esta matemática moral? En su introducción, Sylla nos presenta la reconstrucción más completa hoy disponible de las convicciones teológicas de Bernoulli, en particular en lo que respecta a la cuestión de nuestra libertad. Apoyándose en trascripciones de los archivos, información sobre el credo de la época, referencias indirectas en la correspondencia y otros escritos, Sylla nos presenta a Bernoulli como un partidario de la libre elección. La principal dificultad que, a mi juicio, plantea esta reconstrucción es que Bernoulli estaba lo suficientemente familiarizado con las disputas teológicas de su época como para saber que semejante libertad era difícil de encajar doctrinalmente con la afirmación de la omnisciencia y la omnipotencia divina, y mucho menos en un medio calvinista. La propia Sylla nos informa de que existe aún material de archivo relevante para dilucidarlo, de modo que sólo podemos agradecerle este primer paso exegético que sirve, además, como preámbulo a una magnífica edición.

{Febrero 2007}
{Dynamis 27 (2007): 387-389}