No description available.

No description available.

This dissertation consists of three research papers on cheap talk game and satisficing behaviour. The first chapter examines the potential for communication via cheap talk between an expert and a decision maker whose type (preferences) is uncertain. The expert privately observes states for each type of the decision maker and wants to persuade the decision maker to choose an action in his favour by informing her of the states. The decision maker privately observes her type and chooses an action. An optimal action for the decision maker depends upon both her type and type-specific states. In equilibrium the expert can always inform the decision maker in the form of comparative statements and the decision maker also can partially reveal her type to the expert or public. The second and third chapters build a dynamic model of satisficing behaviour in which an agent’s “expected” payoff is explicitly introduced, where this expectation is adaptively formed. If the agent receives a payoff above her satisficing level she continues with the current action, updating her valuation of the action. If she receives a payoff below her satisficing level and her valuation falls below her satisficing level she updates both her action and satisficing level. In the second chapter, we find that in the long run, all players satisfice. In individual decision problems, satisficing behaviour results in cautious, maximin choice and in normal form games like the Prisoner’s Dilemma and Stag Hunt, they in the long run play either cooperative or defective outcomes conditional on past plays. In coordination games like the Battle of the Sexes, Choosing Sides and Common Interest, they in the long run coordinate on Pareto optimal outcomes. In the third chapter, we find that satisficing players in the long run play subgame dominant paths, which is a refinement of subgame perfection, and identify conditions with which they ‘always cooperate’ or ‘fairly coordinate’ in repeated Prisoner’s Dilemma and Battle of the Sexes games, respectively, and truthfully communicate in sender-receiver games. Proofs and simulations are provided in appendices.

To United Minds, large-scale surveys are an important offering to clients, not least the public opinion poll Väljarbarometern. A risk associated with surveys is satisficing – sub-optimal response behaviour impairing the possibility of correctly describing the sampled population through its results. The purpose of this study is to – through the use of multivariate outlier detection methods - identify those observations assumed to be non-representative of the population. The possibility of categorizing responses generated through satisficing as outliers is investigated. With regards to the character of the Väljarbarometern dataset, three existing algorithms are adapted to detect these outliers. Also, a number of randomly generated observations are added to the data, by all algorithms correctly labelled as outliers. The resulting anomaly scores generated by each algorithm are compared, concluding the Otey algorithm as the most effective for the purpose, above all since it takes into account correlation between variables. A plausible cut-off value for outliers and separation between non-representative and representative outliers are discussed. The resulting recommendation is to handle observations labelled as outliers through respondent follow-up or if not possible, through downweighting, inversely proportional to the anomaly scores.

Notre recherche porte sur l’évaluation du « satisficing organisationnel » du système de pilotage des entreprises à vocation sociale (EVS). Pour cela, nous mobilisons une méthodologie qualitative centrée sur deux études cas enchâssés, dont une en longitunale. Il s’agit de Sos Maintien à Domicile et l’Eveil, deux entreprises associatives dans le champs institutionnel du médico-social. Nous mettons en évidence trois niveaux de résultats. Tout d’abord, nous évaluons, à travers les cartes cognitives, le système de représentations partagées permettant d’évaluer le sentiment d’efficacité organisationnelle. Ensuite , à travers les comptes de surplus de productivité global, nous mesurons la performance économique et sociale. La convergence des deux premiers niveaux de résultats permet de statuer sur l’état positif ou négatif du satisficing organisationnel. Enfin, l’analyse des cas et leur confrontation avec le modèle conceptuel et la revue de littérature, nous permet de définir des variables essentielles dont le système de pilotage de l’EVS, représenté par le pouvoir managérial ,peut s’en servir dans les différentes modes de régulation du projet. / Our research focuses on the evaluation of the "organizational satisficing" of social purpose business (EVS). For this, we mobilize a qualitative methodology centered on two embedded case studies, including one in longitunale way. "Sos Maintien à Domicile" and "Eveil" are both two associative enterprises in the institutional medico-social care field. We highlight three levels of results. First, we evaluate, through the cognitive maps, the system of shared representations to assess the sense of organizational effectiveness. Then, through the global productivity surplus accounts, we measure the economic and social performance. The convergence of the first two levels of results allows to rule on the positive or negative state of organizational satisficing. Finally, the analysis of cases and confrontation with the conceptual model and the literature review, allows us to identify key variables with the EVS control system, represented by the managerial authority, can be used in various project control modes.

Many environments in which an agent can use reinforcement learning techniques to learn profitable strategies are affected by other learning agents. These situations can be modeled as general-sum games. When playing repeated general-sum games with other learning agents, the goal of a self-interested learning agent is to maximize its own payoffs over time. Traditional reinforcement learning algorithms learn myopic strategies in these games. As a result, they learn strategies that produce undesirable results in many games. In this dissertation, we develop and analyze algorithms that learn non-myopic strategies when playing many important infinitely repeated general-sum games. We show that, in many of these games, these algorithms outperform existing multiagent learning algorithms. We derive performance guarantees for these algorithms (for certain learning parameters) and show that these guarantees become stronger and apply to larger classes of games as more information is observed and used by the agents. We establish these results through empirical studies and mathematical proofs.

No description available.

O objetivo desse estudo foi testar as hipóteses de simetria e assimetria entre reforçamento e punição utilizando análise oriunda da Economia Comportamental. Foram utilizados cinco ratos Sprague Dawley, treinados em uma cadeia comportamental mantida por economia de tokens, em uma caixa tendo como operanda duas rodas e uma barra: respostas de girar a roda produziam tokens (LEDs) e as de pressionar a barra trocavam os tokens por sacarose. O elo de produção de tokens consistia em um esquema concorrente, sendo um oferecendo reforçamento positivo (adição de tokens) e o outro um esquema misto de reforçamento positivo e punição negativa (adição ou remoção de tokens, respectivamente). A variável independente foi a exigência de determinado número de respostas de girar a roda para liberação de 1ml de sacarose (preço unitário), sendo a variável dependente a alocação de respostas entre os dois esquemas concorrentes. Todos os sujeitos estabeleceram distribuições estáveis de respostas (analisados como pacotes de respostas) entre os dois esquemas, os quais variaram em função do preço unitário. Os resultados confirmaram que os LEDs tiveram função de estímulo discriminativo, mas não foram claros sobre a sua função punitiva quando removidos contingente à resposta. Consequentemente, os dados obtidos não permitiram que se concluísse sobre as hipóteses testadas. Os resultados foram então analisados por três modelos explicativos do comportamento de escolha: melioração, maximização e satisficing. O modelo de satisficing foi o que produziu a melhor explicação das escolhas molares de todos os sujeitos, sob todos os preços unitários. O modelo de melioração ofereceu explicações adequadas para três sujeitos, especialmente quando o preço unitário era baixo, enquanto o de maximização foi adequado na condição de preços unitários altos, mas apenas para dois sujeitos / The objective of this study was to test the hypotheses of symmetry and asymmetry between reinforcement and punishment using an analysis that stems from behavioral economics. Five Sprague Dawley rats were used, submitted to a behavioral chain maintained by a token economy. The operant chamber had two response wheels and a lever: wheelspinning responses produced tokens (LEDs) and lever-presses exchanged tokens for sucrose. The token-production link was a concurrent condition: a positive reinforcement (token production) schedule, and a mixed schedule with a positive reinforcement and a negative punishment component (token production and removal, respectively). The independent variable was unit price, that is, the number of wheel-spins required to produce 1 cc of sucrose; the dependent variable was response allocation between the two concurrent schedules. All subjects established stable response distributions (considered as response packages) between the two schedules, which varied according to unit prices. The results confirmed that the LEDs acted as discriminative stimulus, but were not clear as to their punitive function when removed, response-contingently. Therefore, the results did not support any conclusion about the tested hypotheses. They were then analyzed based on three choice models: melioration, maximization, and satisficing. The satisficing model produced the most comprehensive explanation of molar choices among all subjects and unit prices. The melioration model produced adequate explanations for three subjects, mostly when unit prices were low, while maximization was adequate in the condition when unit prices were high, but only for two subjects

O objetivo desse estudo foi testar as hipóteses de simetria e assimetria entre reforçamento e punição utilizando análise oriunda da Economia Comportamental. Foram utilizados cinco ratos Sprague Dawley, treinados em uma cadeia comportamental mantida por economia de tokens, em uma caixa tendo como operanda duas rodas e uma barra: respostas de girar a roda produziam tokens (LEDs) e as de pressionar a barra trocavam os tokens por sacarose. O elo de produção de tokens consistia em um esquema concorrente, sendo um oferecendo reforçamento positivo (adição de tokens) e o outro um esquema misto de reforçamento positivo e punição negativa (adição ou remoção de tokens, respectivamente). A variável independente foi a exigência de determinado número de respostas de girar a roda para liberação de 1ml de sacarose (preço unitário), sendo a variável dependente a alocação de respostas entre os dois esquemas concorrentes. Todos os sujeitos estabeleceram distribuições estáveis de respostas (analisados como pacotes de respostas) entre os dois esquemas, os quais variaram em função do preço unitário. Os resultados confirmaram que os LEDs tiveram função de estímulo discriminativo, mas não foram claros sobre a sua função punitiva quando removidos contingente à resposta. Consequentemente, os dados obtidos não permitiram que se concluísse sobre as hipóteses testadas. Os resultados foram então analisados por três modelos explicativos do comportamento de escolha: melioração, maximização e satisficing. O modelo de satisficing foi o que produziu a melhor explicação das escolhas molares de todos os sujeitos, sob todos os preços unitários. O modelo de melioração ofereceu explicações adequadas para três sujeitos, especialmente quando o preço unitário era baixo, enquanto o de maximização foi adequado na condição de preços unitários altos, mas apenas para dois sujeitos / The objective of this study was to test the hypotheses of symmetry and asymmetry between reinforcement and punishment using an analysis that stems from behavioral economics. Five Sprague Dawley rats were used, submitted to a behavioral chain maintained by a token economy. The operant chamber had two response wheels and a lever: wheelspinning responses produced tokens (LEDs) and lever-presses exchanged tokens for sucrose. The token-production link was a concurrent condition: a positive reinforcement (token production) schedule, and a mixed schedule with a positive reinforcement and a negative punishment component (token production and removal, respectively). The independent variable was unit price, that is, the number of wheel-spins required to produce 1 cc of sucrose; the dependent variable was response allocation between the two concurrent schedules. All subjects established stable response distributions (considered as response packages) between the two schedules, which varied according to unit prices. The results confirmed that the LEDs acted as discriminative stimulus, but were not clear as to their punitive function when removed, response-contingently. Therefore, the results did not support any conclusion about the tested hypotheses. They were then analyzed based on three choice models: melioration, maximization, and satisficing. The satisficing model produced the most comprehensive explanation of molar choices among all subjects and unit prices. The melioration model produced adequate explanations for three subjects, mostly when unit prices were low, while maximization was adequate in the condition when unit prices were high, but only for two subjects

Thèse numérisée par la Direction des bibliothèques de l'Université de Montréal.