Dans les billets précédents de cette sous-série consacrée au livre de Jean Lave « Cognition in Practice – Mind, Mathematics and Culture in Everyday Life » je vous ai présenté la structure, les objectifs et l’introduction du livre.
Dans celui-ci, je vous présente le chapitre intitulé « Missionnaires et cannibales » (à l’intérieur) qui traite principalement du concept de « transfert d’apprentissage »
Ce billet est un peu long et dense, pas toujours simple à lire. Et si vous êtes toujours là, après 19 billets, vous m’aurez certainement déjà pardonné.
« Learning transfer is assumed to be the central mechanism for bringing school-taught knowledge to bear in life after school.
Because transfer is so central, it seems logical to begin an investigation of everyday cognitive activity with a reexamination of this formulationof relations between cognition and the everyday world. »
Jean Lave fait remonter la recherche sur le sujet du transfert d’apprentissage à la critique que faisait Thorndike du concept de « discipline formelle » qui était un argument en faveur de l’enseignement du latin au début des années 1900. Argument que l’on retrouve en faveur de l’enseignement de la géométrie et d’autres branches des mathématiques, dans les années 1980.
Jean Lave explicite la théorie de la connaissance sous-jacente:
« Knowledge is conceived as a set of tools stored in memory, carried around by individuals who take the tools (e.g. « foolproof » arithmetic algorithms) out and use them, the more often and appropriately the better, after which they are stowed away again without change at any time during the process. The metaphor is especially apt given that tools are designed to resist change or destruction through the conditions of their use. »
et présente les deux théories du transfert de l’apprentissage qui en découlent, l’une qui fait référence à un grand nombre d’outils spécialisés et l’autre qui fait référence à un nombre limité d’outils génériques.
« one argues for many special purpose tools, the proper one for each task, while the other argues for a few general purpose tools to be used in the largest number of circumstances »
Les représentants de ces deux théories, selon Jean Lave, sont Thorndike et Judd:
« Thorndike (…) suggested that the more two situations shared specific components, such as « ideas of method and habits of procedure, » the more likely the « spread of improvement » from one situation to the other. Judd (…), a student of Wundt‘s, proposed that learning transfer depended upon generality of understanding: the more general the principle, the more likely the recognition that a newly encountered problem might belong to a class of problems already known. Both Thorndike and Judd reported some successes and numerous fruitless attempts to demonstrate learning transfer in laboratory and school settings. »
Jean Lave présente ensuite les différentes expérimentations étudiées (13 expérimentations issues de 4 recherches) – et dégage le critère d’évaluation le plus commun :
The « ethnographic » exercise which follows is based on four wellknown papers describing some 13 learning transfer experiments. Reed, Ernst and Banerji carried out research on river crossing problems (1974), Hayes and Simon on a version of the tower of Hanoi (1977), Gick and Holyoak (1980) on Duncker‘s « radiation problem » (1945), and Gentner and Gentner on models of simple electrical circuits (1983).
The papers fall into a chronological sequence and the later ones take into account the results of the earlier studies. The experiments, which took place in laboratories, with high school and college students as subjects, consisted of sequences of puzzle-solving tasks. Learning transfer is inferred in several different ways, but the most common criteria are an increase in efficiency or accuracy of performance, or use of a general form of the solution to one problem in solving other problems.
Dans cette section du chapitre, l’auteur passe en revue chaque recherche sur le plan méthodologique et sur les résultats obtenus. C’est une partie arride qui n’intéressera vraiment que les chercheurs et qui ne pourrait être utilement commentée que ces mêmes chercheurs
(1) Reed, Ernst and Banerji set out « to study the role of analogy in transfer between problems with similar problem states » (1974 …).
They began with a formal analysis of the missionary and cannibal problem, a flow diagram showing all permissible moves for transporting pairs of people across a river in such a fashion that cannibals do not outnumber missionaries on either bank. This was paired with a formally isomorphic but slightly more complicated problem, « the jealous husbands, » in which each husband-and-wife pair has a unique identity.
To investigate transfer, defined as significant improvement in performance from one problem to the other, Reed et al. compared solution time, number of moves and number of erroneous moves for each pair of problem-solving attempts, looking for statistically significant improvement.
Cette section du chapitre peut fonctionner comme une synthèse des résultats de ces recherches, et ceux-ci sont bien maigres.
« None of these experiments led to strong evidence of transfer. Hayes and Simon attempted to demonstrate transfer of training and constraints on transfer at the same time. They succeeded in generating partial evidence for both. »
« The other experimenters also summed up their results as only partial demonstrations of transfer (Reed et al. 1974 negatively, and Gick and Holyoak 1980 more tentatively than Gentner and Gentner 1983).
But the news in this ethnographic excursion is how little transfer there is, rather than how much »
Et établit aussi le lien avec un concept celui de la résolution de problème (problem solving) un des concepts clés.
« Each of these characteristics of problem solving appears to be an equally apt characterization of assumptions about problem solving in school settings as well. Part of the network of relations between schooling and the study of cognition is located in their mutual influence on the choice and constitution of experimental tasks and situations, and especially in assumptions about problems and problem solving.
Transfer theory presupposes that problem-solving activities are always a quest for truth or the « right answer » to a given problem. The rationale in terms of scientific thinking both reflects and contributes to this view, which is crucial not only to theories about how people solve problems, but also to methodological formulations of research strategy. »
Pour Jean Lave, la relation entre les modèles normatifs et les pratiques quotidiennes est une question centrale (à laquelle elle reviendra plus loin) et pose d’emblée deux difficultés, qu’elle entend esquisser:
« Relations between normative models and everyday practice are a central issue to which we shall return in other chapters. But two preliminary indications of the difficulties with this cornerstone of cognitive research may be sketched here, one empirical, the other theoretical:
1. The use of normative models as the reference point for interpreting subjects’ activities stands in contrast with a recent example (…) of a sustained attempt by an informant to figure out how an unfamiliar heat exchange mechanism worked (…). They report that an informant generated one model of the exchanger from which he inferred the answer to an initial question. In attempting to answer another question he recognized flaws in the first model and invented a second. This too was flawed. The informant subsequently moved between these models, in a process that advanced his understanding of the heat transfer mechanism and improved the models at the same time. The two models, and their limitations and contradictions, served as devices for generating useful experience, though not in normative terms.
2. The second challenge to the practice of deriving experimental tasks from normative models is that they foster a static, objectified conceptualization of processes of reasoning, a transformation that occurs between their initial formulation and their incorporation into experimental procedures. It is hard to avoid the conclusion that while skills, mental maps and analogies may be conceived of as processes by the experimenters, they are treated for experimental purposes as objects. Thus, they are sometimes referred to in these papers as cognitive « tools. » Subjects are on the whole expected to apply or map relations from one problem to the next.
All of this underscores the static quality of transfer in experimental practice: it is treated as a process of taking a given item and applying itsomewhere else.
In sum, there is no impatience, no hint in this work, that the meager evidence for transfer garnered from a very substantial body of work
might indicate that the concept is seriously misconceived. This research genre involves school-like assumptions about the nature of problem solving and its high priority in ongoing activity. It relies on normative models of good thinking as justification, source and standard for experimental tasks and performances. This in turn has unintended consequences for the conceptualization of cognitive processes in termsthat appear to pose a serious dilemma for the field. »
Dans cette section du chapitre, Jean Lave aborde deux nouveaux concept clés (contexte et motivation) et commente leur prise en considération dans la recherche sur le transfert d’apprentissage.Jean Lave examine de manière méthodique les variations du sens donné aux mots contexte, situation … parfois à l’intérieur d’une même recherche
« Whether talking about transfer across social situations (Simon 1980) or semantic domains (Gick and Holyoak 1980), all of the experimentalists operationalize the concept of setting or situation in essentially identical ways: problem content is the only « context » germain to problem solving activity. But though operationalized similarly across papers, the terms « context » and « situation » are given multiple meanings at different points within them.
I shall argue in chapter 4 that the merging of meanings of « context » grows from a particular specification of relations among knowledge, culture and socially organized groups in a cognitivist worldview.
There are other difficulties with the conceptualization of context in these experiments and more broadly in the experimental study of cognition. Transfer is characterized as occurring across unrelated, or analogically related, or remotely related situations, but never across settings complexly interrelated in activity, personnel, time, space, or their furnishings. There is an implicit assumption that each context of activity exists in virtual isolation from all others. The abstract quality of this vision is called into question by observational studies in which much of what transpires across settings interpenetrates in various ways – for example, meal planning, grocery shopping, managing money, dieting, cooking and sitting down to meals. Why the implicit assumption of isolation between contexts? It is both constructed in, and reflects, the prime institutionalized exemplar of transfer, the transportation of knowledge from school to other settings. Functional psychological theory treats school as the decontextualized (and hence privileged and powerful) site of learning that is intended for distant and future use If indeed it were possible to acquire knowledge « out of context » the school (non)context of learning would necessarily stand in an arbitrary and unspecifiable relation with all contexts-of-application. Such logic contributes to the characterization of learning transfer as occurring across unrelated situations.
Further, when « tool » is used as a metaphor for knowledge-in-use across settings, there is assumed to be no interaction between tool and situation, but only an application of a tool on different occasions. Since situations are not assumed to impinge on the tool itself, a theory of learning transfer does not require an account of situations, much less of relations among them. Knowledge acquisition may be considered (and organized in schools and experiments on cognition) as if the social context of activity had no critical effects on knowledge-in-use. Given this view, it is difficult to see how researchers could do anything but assume away complex relations among social situations, along with the situations themselves. It might be added that experiments, which after all are social situations, reflect and are produced by this theoretical position, for they have no multiple, well-formed, actor-generated relations with other situations and activities in the lives of their subjects. »
En passant, l’on notera qu’un des concepts clés des « communautés de pratique »  (définie par le triptyque: Communauté, Pratique(s) et « Domaine de connaissance » par Etienne Wenger et Jean Lave , est sérieusement démoli par cette dernière quand elle écrit:
« There is little in this genre of research to hold its analyses to events in space and time, partly because, in the functionalist conception, knowledge is necessarily abstracted from experience. This belief is reflected in the common concept of « knowledge domains, » a term that appears to locate knowledge-in-use in time and space without in fact doing so. Here, too, ambiguity may be the advantage which gives the concept a secure place in discussions of cognition; it provides a pseudo-space, an illusion that knowledge is the context of problem solving. But the effect on cognitive research of « locating » problems in « knowledge domains » has been to separate the study of problem solving from analysis of the situations in which it occurs. »
« I think. I suspect that « knowledge domain » is in fact a name for a conventionally acknowledged claim by a social group (e.g. a profession or academic discipline) heavily invested in maintaining its boundaries. Control of a body of knowledge plays a major strategic role in such enterprises. If this view is correct, a « knowledge domain » is a socially constructed exoticum, that is, it lies at the intersection of the myth of decontextualized understanding and professional/academic ~specialization and thus is probably not the general concept sought by cognitive psychology.
Further, and also worrisome, the « knowledge domain » as a putative context of cognitive activity is an inert category of analysis. It has no interactive, generative or action-motivating properties. »
Avant d’arriver au constat, qu’il manque une dimension dans l’exploration expérimentale de la résolution de problème et du transfert d’apprentissage: une théorie de la motivation.
« A major factor missing from experimental investigations of problem solving and transfer is an account of what motivates people to recognize and undertake to solve problems when not required to do so. The question need not arise when subjects have tacitly agreed to comply with an experimenter’s requests, problem solving is the major ongoing activity, and problems are furnished by problem givers. But in everyday activity the presence or absence of problem solving is often not controlled by others, nor is it determined by some general eagerness or reluctance to solve problems. To analyze problem solving in everyday activity, in short, we shall need a theory of motivation. For whether to have a problem or not, and the specifIcation of what constitutes the problem, are commonly choices made by problem solvers.
There are, then, two consistent, well-structured lacunae in this work: serious shortcomings, I would argue. One concerns the absent social situation, the other a silence about what motivates problem solving and the transfer of knowledge from one setting to another. Both will be considered further in later chapters.
Les conclusions de Jean Lave:
– les recherches sur les transferts d’apprentissage séparent la cognition du monde social
– Toutes ces recherches dissocient la cognition de son contexte, …
– ce qui explique l’absence d’une théorisation d’expérimentations en tant que situations sociales et de la cognition en tant qu’activité située
– elles posent également des questions sur la nature des problèmes et sur l’activité de résolution de problème
Learning-transfer research clearly falls in the functionalist tradition sketched in chapter 1. Its central characteristics include the separation of cognition from the social world, the separation of form and content implied in the practice of investigating isomorphic problem solving, and a strictly cognitive explanation for continuity in activity across situations.
All of these dissociate cognition from its contexts, and help to account for the absence of theorizing about experiments as social
situations and cognition as socially situated activity
This appraisal raises questions about the nature of problems and problem solving, and I would like to propose an initial conclusion:
Problems of the closed, « truth or consequences » variety are a specialized cultural product, and indeed, a distorted representation of activity in everyday life, in both senses of the term -that is, they are neither common nor do they capture a good likeness of the dilemmas addressed in everyday activity.
But this advances speculation beyond the current state of the project. Observations about the cultural particularities of transfer research cannot provide positive confirmation of variation in arithmetic practice across situations. This is required as a next step in exploring the situated character of everyday cognition. In the following chapters both empirical evidence and further analysis of cognitive theory provide a stronger rationale for moving into the experienced, lived-in world as the site and source of further investigations of cognitive activity. »
 Community of Practice (définition et présentation sur le site d’Etienne Wenger)