This programme of research investigates the effect of belief bias in syllogistic reasoning. Belief bias is conventionally characterised as a non-logical tendency to accept or reject deductive inferences on the basis of belief rather than logical argument. However, some theorists have argued that the effect is weak compared with that of logic and that it arises from misinterpretation of the premises or failure to accept the logical task. Despite the adoption of controls recommended in the recent literature, Experiments 1 to 3 found consistently strong belief bias effects on the syllogistic evaluation task. However, there were equally strong effects of logic and an interaction between the two factors. Verbal protocol analysis revealed some possible misinterpretation of premises. More strikingly, however, it suggested the presence of three different modes of reasoning which were forward, backward or conclusion based and associated respectively with increasing levels of belief bias. Belief bias was not observed in Experiments 4 and 5 which employed similar problem content on the syllogistic construction task. However, in view of findings recently published by other researchers, it appears that more salient beliefs are needed to produce the effect on this type of task. Experiments 6 to 9 investigated the cause of the logic times belief interaction observed here and in earlier published studies: in essence, the effect of belief is stronger on invalid than valid problems. This could be due to misinterpretation of the logical concept of necessity, but extended instruction on logical interpretation failed to eliminate the effect. The findings were more consistent with a selective scrutiny model of belief bias which claims that arguments supporting unbelievable conclusions are more thoroughly analysed than those supporting believable conclusions. This model is discussed with reference to contemporary theories and findings in the psychology of reasoning.

La représentation des connaissances et le raisonnement sur le Web sémantique se sont récemment concentrés, pour des raisons pratiques, sur le sous-ensemble de la logique du premier ordre appelé règles existentielles. Dans cette thèse, nous étudions le raisonnement avec des règles existentielles en présence d'informations contradictoires et introduisons un raisonnement existentiel défaisible. Nous proposons trois résultats principaux: Premièrement, nous montrons que les techniques de raisonnement défaisibles classiques doivent être revisitées pour les règles existentielles et étudions leurs défis théoriques et de mise en œuvre. Deuxièmement, nous fournissons une nouvelle structure combinatoire qui permet de capturer diverses variantes du raisonnement défaisable et étudions son expressivité et sa polyvalence. Troisièmement, nous évaluons notre travail par rapport à l'état de l'art dans le traitement des incohérences et des inconsistances dans les règles existentielles et étudions l'intérêt humain de telles techniques de raisonnement. / Knowledge representation and reasoning on the Semantic Web has recently focused, due to practical rationale, on the subset of first order logic called existential rules. In this thesis we investigate reasoning with existential rules in presence of conflicting information and introduce defeasible existential rule reasoning. We provide three main salient results as follows. First we show that classical defeasible reasoning techniques need to be revisited for existential rules and study their theoretical and implementation related challenges. Second, we provide a new combinatorial structure that allows for diverse variants of defeasible reasoning to be captured together and study its expressivity and versatility. Third we evaluate our work with respect to the state of the art in inconsistency handling in existential rules and investigate the human appeal of such reasoning techniques.

Neste trabalho, investigou-se a utilização de algoritmos evolutivos para identificação de procedimentos de raciocínio utilizados por humanos na construção de soluções para uma classe de problemas cuja principal característica é a utilização de raciocínio dedutivo. Para isso, utilizou-se uma abordagem que explora os diferentes níveis de complexidade do problema, partindo da análise das estratégias apresentadas por jogadores humanos. Foram realizados diversos ensaios que evoluíram primeiramente, para um modelo de solução puramente combinatória guiada por um algoritmo genético e independente do jogador humano, até atingir um modelo que procura identificar um procedimento de solução que guarde semelhanças com os procedimentos apresentados pelos jogadores humanos. Como resultado, apresentou-se um algoritmo denominado Classificador Genético um sistema de operadores guiado por um algoritmo genético capaz de identificar procedimentos de raciocínio para solução de combinações específicas do problema proposto. Os ensaios permitiram ainda identificar conjuntos de operadores que se combinados corretamente, através de um mecanismo que simule a tomada de decisão do jogador humano, são capazes de aumentar o potencial de identificação de soluções do algoritmo proposto. O estudo também revelou a importância dos mecanismos de memória no processo de solução do problema e as dificuldades em manipular operadores gerais com métodos puramente evolutivos. Foi possível ainda identificar de que modo jogadores humanos tratam os fatores relacionados à diversidade de possíveis encaminhamentos no processo decisório, que afetam a solução do problema proposto. / In this work we investigated the use of evolutionary algorithms to identify logic procedures used by humans in the construction of solutions of a class of problems in which the main characteristic is the use of deductive reasoning. In order to do that it was used an approach that explores the problems different levels of complexity, starting from the strategies analysis presented by human players. Several experiments were carried out where at first moment used a model of solution that is strictly combinatorial guided by a genetic algorithm and independent of the human player that evolved to a model that tries to identify a solution procedure that maintains the similarities with the procedures presented by human players. As a result, we presented an algorithm denominated Genetic Classifier - a system of rules guided by a genetic algorithm - able to identify reasoning procedures for solution of specific combinations of the proposed problem. Moreover, the experiments allowed identifying clusters of rules that if combined correctly, through a mechanism that simulates the decision making performed by a human player, are capable of increasing the potential to identify the solutions of the proposed algorithm. The study also revealed the importance of the memorys mechanism in the process of solving the proposed problem and the difficulties to manipulate general rules with regular evolutionary methodologies. It was also possible to identify the way human players deal with the factors related to the diversity of possible directions in the decision process.

Neste trabalho, investigou-se a utilização de algoritmos evolutivos para identificação de procedimentos de raciocínio utilizados por humanos na construção de soluções para uma classe de problemas cuja principal característica é a utilização de raciocínio dedutivo. Para isso, utilizou-se uma abordagem que explora os diferentes níveis de complexidade do problema, partindo da análise das estratégias apresentadas por jogadores humanos. Foram realizados diversos ensaios que evoluíram primeiramente, para um modelo de solução puramente combinatória guiada por um algoritmo genético e independente do jogador humano, até atingir um modelo que procura identificar um procedimento de solução que guarde semelhanças com os procedimentos apresentados pelos jogadores humanos. Como resultado, apresentou-se um algoritmo denominado Classificador Genético um sistema de operadores guiado por um algoritmo genético capaz de identificar procedimentos de raciocínio para solução de combinações específicas do problema proposto. Os ensaios permitiram ainda identificar conjuntos de operadores que se combinados corretamente, através de um mecanismo que simule a tomada de decisão do jogador humano, são capazes de aumentar o potencial de identificação de soluções do algoritmo proposto. O estudo também revelou a importância dos mecanismos de memória no processo de solução do problema e as dificuldades em manipular operadores gerais com métodos puramente evolutivos. Foi possível ainda identificar de que modo jogadores humanos tratam os fatores relacionados à diversidade de possíveis encaminhamentos no processo decisório, que afetam a solução do problema proposto. / In this work we investigated the use of evolutionary algorithms to identify logic procedures used by humans in the construction of solutions of a class of problems in which the main characteristic is the use of deductive reasoning. In order to do that it was used an approach that explores the problems different levels of complexity, starting from the strategies analysis presented by human players. Several experiments were carried out where at first moment used a model of solution that is strictly combinatorial guided by a genetic algorithm and independent of the human player that evolved to a model that tries to identify a solution procedure that maintains the similarities with the procedures presented by human players. As a result, we presented an algorithm denominated Genetic Classifier - a system of rules guided by a genetic algorithm - able to identify reasoning procedures for solution of specific combinations of the proposed problem. Moreover, the experiments allowed identifying clusters of rules that if combined correctly, through a mechanism that simulates the decision making performed by a human player, are capable of increasing the potential to identify the solutions of the proposed algorithm. The study also revealed the importance of the memorys mechanism in the process of solving the proposed problem and the difficulties to manipulate general rules with regular evolutionary methodologies. It was also possible to identify the way human players deal with the factors related to the diversity of possible directions in the decision process.

Results of psychological experiments have shown that humans make assumptions, which are not necessarily valid, that they are influenced by their background knowledge and that they reason non-monotonically. These observations show that classical logic does not seem to be adequate for modeling human reasoning. Instead of assuming that humans do not reason logically at all, we take the view that humans do not reason classical logically. Our goal is to model episodes of human reasoning and for this purpose we investigate the so-called Weak Completion Semantics. The Weak Completion Semantics is a Logic Programming approach and considers the least model of the weak completion of logic programs under the three-valued Łukasiewicz logic. As the Weak Completion Semantics is relatively new and has not yet been extensively investigated, we first motivate why this approach is interesting for modeling human reasoning. After that, we show the formal correspondence to the already established Stable Model Semantics and Well-founded Semantics. Next, we present an extension with an additional context operator, that allows us to express negation as failure. Finally, we propose a contextual abductive reasoning approach, in which the context of observations is relevant. Some properties do not hold anymore under this extension. Besides discussing the well-known psychological experiments Byrne’s suppression task and Wason’s selection task, we investigate an experiment in spatial reasoning, an experiment in syllogistic reasoning and an experiment that examines the belief-bias effect. We show that the results of these experiments can be adequately modeled under the Weak Completion Semantics. A result which stands out here, is the outcome of modeling the syllogistic reasoning experiment, as we have a higher prediction match with the participants’ answers than any of twelve current cognitive theories. We present an abstract evaluation system for conditionals and discuss well-known examples from the literature. We show that in this system, conditionals can be evaluated in various ways and we put up the hypothesis that humans use a particular evaluation strategy, namely that they prefer abduction to revision. We also discuss how relevance plays a role in the evaluation process of conditionals. For this purpose we propose a semantic definition of relevance and justify why this is preferable to a exclusively syntactic definition. Finally, we show that our system is more general than another system, which has recently been presented in the literature. Altogether, this thesis shows one possible path on bridging the gap between Cognitive Science and Computational Logic. We investigated findings from psychological experiments and modeled their results within one formal approach, the Weak Completion Semantics. Furthermore, we proposed a general evaluation system for conditionals, for which we suggest a specific evaluation strategy. Yet, the outcome cannot be seen as the ultimate solution but delivers a starting point for new open questions in both areas.

Results of psychological experiments have shown that humans make assumptions, which are not necessarily valid, that they are influenced by their background knowledge and that they reason non-monotonically. These observations show that classical logic does not seem to be adequate for modeling human reasoning. Instead of assuming that humans do not reason logically at all, we take the view that humans do not reason classical logically. Our goal is to model episodes of human reasoning and for this purpose we investigate the so-called Weak Completion Semantics. The Weak Completion Semantics is a Logic Programming approach and considers the least model of the weak completion of logic programs under the three-valued Łukasiewicz logic. As the Weak Completion Semantics is relatively new and has not yet been extensively investigated, we first motivate why this approach is interesting for modeling human reasoning. After that, we show the formal correspondence to the already established Stable Model Semantics and Well-founded Semantics. Next, we present an extension with an additional context operator, that allows us to express negation as failure. Finally, we propose a contextual abductive reasoning approach, in which the context of observations is relevant. Some properties do not hold anymore under this extension. Besides discussing the well-known psychological experiments Byrne’s suppression task and Wason’s selection task, we investigate an experiment in spatial reasoning, an experiment in syllogistic reasoning and an experiment that examines the belief-bias effect. We show that the results of these experiments can be adequately modeled under the Weak Completion Semantics. A result which stands out here, is the outcome of modeling the syllogistic reasoning experiment, as we have a higher prediction match with the participants’ answers than any of twelve current cognitive theories. We present an abstract evaluation system for conditionals and discuss well-known examples from the literature. We show that in this system, conditionals can be evaluated in various ways and we put up the hypothesis that humans use a particular evaluation strategy, namely that they prefer abduction to revision. We also discuss how relevance plays a role in the evaluation process of conditionals. For this purpose we propose a semantic definition of relevance and justify why this is preferable to a exclusively syntactic definition. Finally, we show that our system is more general than another system, which has recently been presented in the literature. Altogether, this thesis shows one possible path on bridging the gap between Cognitive Science and Computational Logic. We investigated findings from psychological experiments and modeled their results within one formal approach, the Weak Completion Semantics. Furthermore, we proposed a general evaluation system for conditionals, for which we suggest a specific evaluation strategy. Yet, the outcome cannot be seen as the ultimate solution but delivers a starting point for new open questions in both areas.