Daniel Andler

We seem to suffer from a case of cognitive dissonance. On the one hand, we seem to have almost unanimously rejected as hopeless or incoherent the aim of a unified science. On the other, we passionately debate about the prospects of research programs which, if successful, would considerably enhance the prospects of unification: from particle physics to cognitive neuroscience, from evolutionary theory to logical modeling or dynamic systems, a common motivation seems to be the quest for unity1. The purpose of this paper is to relieve the dissonance. I will defend a moderate form of unity, one which is compatible with the diversity and open-endedness of science, for which I can think of no better name than federalism, as it combines plurality and the construction of a common epistemic area. This view is not original: Otto Neurath himself espoused it, albeit in a context which is in certain respects quite unlike ours2.

In Hubert Dreyfus’s critique of artificial intelligence1, considerable importance is given to the matter of context –used here as a blanket term covering an immense and possibly heterogeneous phenomenon, which includes situation, background, circumstances, occasion and possibly more. Perhaps the best way to point to context in this most general sense is to proceed dialectically, and take as a first approximation context to be whatever is revealed as an obstacle whenever one attempts to account for mental dynamics on the formal model of a combinatorial game over families of interchangeable, fixed, self-standing – context-free as it were– elements. Dreyfus has argued persuasively that the human mind always ‘operates’, if that is the right word, within a setting which permeates whatever it is it operates on. Dreyfus’s thought on these matters was shaped during the years when artificial intelligence (AI) was at the height of early hopes and initial successes. This was also a period during which what was to become known as ‘cognitive science’ was little more than a ray of light falling on psychology..

Fifty years before the present volume appeared, artificial intelligence (AI) and cognitive science (Cogsci) emerged from a couple of small-scale academic encounters on the East Coast of the United States. Wedded together like Siamese twins, these nascent research programs appeared to rest on some general assumptions regarding the human mind, and closely connected methodological principles, which set them at such a distance from phenomenology that no contact between the two approaches seemed conceivable. Soon however contact was made, in the form of a head-on critique of the AI/Cogsci project mostly inspired by arguments from phenomenology. For a while, it seemed like nothing would come of it: AI/Cogsci bloomed while the small troop of critical phenomenologists kept objecting. Then AI and Cogsci went their separate ways. AI underwent a deep transformation and all but surrendered to the phenomenological critique. Cogsci meanwhile pursued the initial program with a far richer collection of problems, concepts and methods, and was for a long time quite unconcerned by suggestions and objections from phenomenology. The last decade and half has seen a remarkable reversal: on the one hand, a few cognitive scientists have been actively pursuing the goal of reconciliating Cogsci, whether empirically or foundationally, with some of the insights procured by phenomenology; on the other, many cognitive scientists and philosophers of mind who think of themselves as, respectively, mainstream and analytic, and have no or little acquaintance with, and often little sympathy for, phenomenology, have been actively pursuing research programs geared toward some of the key issues identified by phenomenological critics of early AI/Cogsci. It might seem then as if those critics were now vindicated. But while these new directions are undoubtedly promising, it is not yet clear that phenomenology and Cogsci can be truly reconciled. Some suspect that Cogsci must distort beyond recognition those phenomenological themes it means to weave into its fabric, while phenomenology may be losing touch with its roots by tuning onto the logic of Cogsci which is, after all, an empirical science. To the present writer, it is far too early for anything like a verdict, as the task of clarifying the issues and gaining a much deeper understanding of the issues on both sides has barely begun. But whatever emerges from this exploration will probably have deep consequences for both Cogsci and philosophy..

Scholars from various philosophical schools of thought, including cultural relativism, hermeneutics, and postmodernism, have recently critiqued rationalism in light of new developments in the cognitive sciences. Each of these new developments set into motion new inquiries in each school philosophical school of thought. Now, in Facets of Rationality, a distinguished team of scholars examines these new inquiries and bring rationality back into the mainstream of the social sciences. The unique feature of this book lies in its multidisciplinary exploration of rational concerns and in discovering the integral meaning of rationality as viewed by the perspectives of different disciplines. As such, it will of considerable interest to those involved with the study and teaching of philosophy, theoretical psychology, cognitive science, political theory, and linguistics.

Conventional wisdom holds two seemingly opposed beliefs. One is that communities are often much better than individuals at dealing with certain situations or solving certain problems. The other is that crowds are usually, and some say always, at best as intelligent as their least intelligent members and at worst even less. Consistency would seem to be easily re-established by distinguishing between advanced, sophisticated social organizations which afford the supporting communities a high level of collective performance, and primitive, mob-like structures which pull the group towards the lower end of the achievement scale. But this reconciliation meets with some objections. The most familiar ones concern the mixed record of elaborate social systems, which are said to occasionally or even, according to some accounts, systematically produce wrong decisions, poor assessments, disastrous plans, counterproductive measures, etc. A more recent set of objections rests on cases where ‘crowds’, i,e, groups not organized in a sophisticated way, produce good results, in fact, results which better those of most, or even all, members of the group. Such cases are collected in James Surowiecki’s book The Wisdom of Crowds, which argues more generally in favor of an ‘order out of chaos’ view of collective thinking: whether sophisticated or simple, social organizations for the production of knowledge or problem-solving can benefit from the absence of certain individualistc constraints which are traditionally thought to foster excellence in cognitive tasks. This flavor of paradox is enhanced by Surowiecki’s choice of phrase: at the surface level, ‘wisdom of crowds’ conflicts with the well-entrenched cliché of the folly1 of crowds; but at a deeper level, Surowiecki seems to appeal to one frequent connotation of ‘wisdom’ which is precisely its paradoxical character. Whether Surowiecki actually intended to exploit this..