Emmanuel Sander

Arguments, More than Confidence, Explain the Good Performance of Reasoning Groups

In many intellective tasks groups consistently outperform individuals. One factor is that the individual(s) with the best answer is able to convince the other group members using sound argumentation. Another factor is that the most confident group member imposes her answer whether it is right or wrong. In Experiments 1 and 2, individual participants were given arguments against their answer in intellective tasks. Demonstrating sound argumentative competence, many participants changed their minds to adopt the correct answer, even though the arguments had no confidence markers, and barely any participants changed their minds to adopt an incorrect answer. Confidence could not explain who changed their mind, as the least confident participants were as likely to change their minds as the most confident. In Experiments 3 (adults) and 4 (10-year-olds), participants solved intellective tasks individually and then in groups, before solving transfer problems individually. Demonstrating again sound argumentative competence, participants adopted the correct answer when it was present in the group, and many succeeded in transferring this understanding to novel problems. Moreover, the group member with the right answer nearly always managed to convince the group even when she was not the most confident. These results show that argument quality can overcome confidence among the factors influencing the discussion of intellective tasks. Explanations for apparent exceptions are discussed.

Analogical transfer as guided by an abstraction process: the case of learning by doing in text editing.

The authors proposed that not only are the first attempts to solve a problem made by analogy but also that progress in learning can be guided by referring to more abstract knowledge, which affords new possibilities. Two experiments investigated this view in a situation of learning how to use a text editor. Experiments 1A to 1C identified the knowledge associated with 3 domains hypothesized as sources of transfer at increasing levels of abstraction (typewriting, writing in general, manipulating objects). Experiment 2 tested whether participants first use their knowledge about typewriting, then about writing in general, and then about manipulating objects. The data showed that the order of acquisition of text editor functions appeared to be strongly related to this hierarchy, supporting the idea that part of learning consists of discovering properties of objects by accessing increasingly general domains.

Arithmetic Word Problem Solving: A Situation Strategy First Framework.

Before instruction, children solve many arithmetic word problems with informal strategies based on the situation described in the problem. A Situation Strategy First framework is introduced that posits that initial representation of the problem activates a situation-based strategy even after instruction: only when it is not efficient for providing the numerical solution is the representation of the problem modified so that the relevant arithmetic knowledge might be used. Three experiments were conducted with Year 3 and Year 4 children. Subtraction, multiplication and division problems were created in two versions involving the same wording but different numerical values. The first version could be mentally solved with a Situation strategy (Si version) and the second with a Mental Arithmetic strategy (MA version). Results show that Si-problems are easier than MA-problems even after instruction, and, when children were asked to report their strategy by writing a number sentence, equations that directly model the situation were predominant for Si-problems but not for MA ones. Implications of the Situation Strategy First framework regarding the relation between conceptual and procedural knowledge and the development of arithmetic knowledge are discussed.

Rôle de la nature des variables dans la résolution de problèmes additifs complexes

Cette recherche teste l’hypothese selon laquelle la nature de la variable mise en jeu dans un probleme arithmetique influence la construction de la representation. Elle est susceptible de contraindre la strategie de resolution en masquant la possibilite de les percevoir toutes, et d’influencer la performance. Pour tester cette hypothese, une experimentation a ete menee aupres d’eleves de CM1-CM2 qui ont eu a resoudre des problemes isomorphes admettant deux procedures de resolution et portant sur les variables Poids, Hauteur et Duree. Les resultats ont montre que la nature de la variable impliquee dans le probleme influencait tant la strategie de resolution que la performance. Les implications de ces resultats pour les apprentissages scolaires sont discutees.

Effets du contexte sur la mise en œuvre de l’inférence de complément

L’inference de complement a un statut particulier parmi les inferences deductives dans la mesure ou, reposant sur la relation parties-tout, elle conditionne la resolution de nombreux problemes arithmetiques a structure additive. Les conditions de mise en oeuvre de cette inference sont etudiees chez des enfants de 10 ans dans une version simplifiee du Mastermind, associee a un programme de reponse destine a placer les participants dans des conditions semblables. Trois situations sont comparees : une athematique, proche de la version classique, une thematique non pertinente, reposant sur une structure narrative sans liens entre les proprietes semantiques de la situation et l’inference de complement, et une thematique pertinente, pour laquelle la relation parties-tout est rendue saillante. Les resultats obtenus montrent que l’inference de complement est rarement mise en oeuvre dans les deux premieres conditions mais frequemment dans la troisieme. Ils suggerent que le caractere realiste d’un enonce n’est facilitateur que lorsque les proprietes semantiques de la situation rendent saillantes une structure adaptee a l’application de l’inference.

The talent so fundamental that it is fuel to our minds

We take for granted our ability to reason using analogies, but the skill is at the core of human cognition, argue Douglas Hofstadter and Emmanuel Sander

Apport pour les apprentissages de l’explicitation des relations d’inclusion de classes

L’acquisition de connaissances sur un domaine depend de la construction de categories structurees sur ce domaine. Dans cette etude, nous faisons l’hypothese que, dans une situation d’apprentissage, l’explicitation de la structuration de connaissances sous forme de reseaux semantiques d’inclusion de classes peut favoriser la categorisation au niveau d’abstraction pertinent pour la tâche et ameliorer la qualite des inferences fondees sur l’heritage de proprietes. Pour tester cette hypothese, des eleves de CM2 ont ete inities au maniement de reseaux semantiques portant sur deux domaines de l’enseignement elementaire : les determinants et les quadrilateres. Leurs resultats ont ete compares a ceux d’eleves ayant etudie les memes contenus de maniere classique : conformement a l’hypothese, le groupe «reseau » a obtenu des taux de reussite significativement plus eleves que le groupe temoin pour chacune des deux disciplines etudiees.

This problem has no solution: when closing one of two doors results in failure to access any.

We investigated what happens when the spontaneous encoding of a problem is incongruent with its solving strategy. We created word problems from which two distinct semantic representations could be abstracted. Only one of these representations was consistent with the solving strategy. We tested whether participants could recode a semantically incongruent representation in order to access another, less salient, solving strategy. In experiment 1, participants had to solve arithmetic problems and to indicate which problems were unsolvable. In experiment 2, participants received solved problems and had to decide whether the solution was appropriate or not. In both experiments, participants had more difficulties acknowledging that problems inducing an incongruent representation could be solved than they had for problems inducing a congruent representation. This was confirmed by response times. These results highlight how semantic aspects can lead even adults to fail or succeed in the solving of arithmetic problems requiring basic mathematical knowledge.

Two different methods to analyze the strategies of problem solving in a natural situation using a simulator: A case study

The PLASTUR project attempted to address a cognitive diagnosis problematic i.e. to understand the problem solving strategies used by operators on a simulator. Carried out by a multidisciplinary team, this research, in its fundamental aspect, aims to confront analyses made almost exclusively from real situations in an ecological context with data obtained by a model in an artificial context. As regards the cognitive modelling of problem solving in natural situations, PLASTUR uses models from different fields of research (cognitive psychology, artificial intelligence) applied to the same set of data. This comparison between the two models highlights the specificities and the complementarities of the two approaches. From a more applied point of view, the objective is to define, with our industrial partner, an interactive learning environment based on the simulator and integrating an intelligible diagnosis module used both by tutors and learners.