Resumen de Privacidade, variáveis não aparentes e a investigação do comportamento de inferir

João Lucas Bernardy Cardoso, Lorismario Ernesto Simonassi, Paula Teixeira Andrade Sousa

• português

  Este estudo teve como objetivo discutir o comportamento de inferir por meio de um modelo experimental e relacioná-lo com diferentes concepções de privacidade. Skinner, em uma concepção behaviorista radical, propôs a análise de estímulos e respostas privadas como variáveis comportamentais. Entre os críticos da concepção de Skinner da privacidade estão os defensores da visão molar/multiescalar, que argumentam que o conceito de privacidade é prejudicial ao desenvolvimento de uma ciência do comportamento. Contudo, propõe-se que eventos não aparentes continuam sendo relevantes, mesmo na visão molar/multiescalar, dada a natureza histórica do comportamento. Conduziu-se dois experimentos que comparam a probabilidade de inferências corretas sobre o comportamento de outros participantes com o número de estímulos discriminativos acessados (Experimento 1) e com uma história de treino de relações formas-palavras em comum (Experimento 2). Participantes do Grupo Base escreveram na presença de conjuntos de palavras, enquanto que os participantes do Grupo Inferência, que não tiveram acesso às palavras escritas por aqueles participantes, inferiram sobre as respostas do Grupo Base. Tanto o acesso aos estímulos discriminativos quanto a história em comum, aumentaram a probabilidade de inferências corretas. Argumenta-se que estudos que investigam a privacidade entre dois participantes podem ser uma solução para estudar o comportamento de inferir.

• English

  The present study aims to discuss the inferring behavior while presenting an experimental model. We discuss that proposal while describing different conceptions about the issue of privacy. When Skinner presented Radical Behaviorism, he analyzed the psychological terms appealing to the concept of privacy. Throughout his works Skinner described an analysis of both private stimuli and private responses as behavioral variables. More than fifty years later there’s little, if any consensus about Skinner’s assumptions. Among the critics of Skinner’s approach are those who hold a molar/multiscale view, who argue that privacy is a harmful concept to the development of a natural science of behavior. Instead, the molar/multiscale view defines behavior as a time extended phenomenon. According to that view, behaviorists appeal to privacy when they try to analyze behavior as a discrete variable. In that context, while accepting the argument of the jeopardies of the concept of privacy, we propose that inferred events are still relevant, given the historical and extended nature of the behavioral phenomena. We conducted two experiments. In both experiments, participant called “Inference Group” tried to make inferences about another participant’s verbal behavior. Experiment 1 compared the probability of right inferences with the number of accessed variables controlling a verbal response. Base participants wrote five words in the presence of five sets of eight stimuli presented in a computer screen, and the Inference Group tried to infer those written verbal responses. The Inference Group accessed the same five sets of eight stimuli, but cumulatively and in a series. For every and each new stimulus at the screen, observers tried to make a new inference. Experiment 2 compared the probability of correct inferences with a common history of operant conditioning. Base participants were exposed to two match-to- -sample procedures. These procedures used verbs or nouns as sample stimuli and geometrical forms as comparisons. After the training sessions, base participants wrote words in the presence of the same geometrical figures used as comparison stimuli during the training. The Inference Group was exposed to a similar procedure, but they have tried to make inferences about base participants verbal behavior in three different moments: at the beginning of the experiment, after the training of verbs, and after the training of nouns. The results of both experiments suggest that the independent variables affected the probability of occurrence of correct inferences. Even that these studies involve well described behavioral relations, namely stimulus control and differential reinforcement, we argue that studies that arrange privacy between at least two participants could be a solution for the investigation of inferring behavior. We also accept the molar/multiscale argument that in these procedures nothing is private in principle, on the other hand the inferring behavior is typical among observers when the relevant variables for the explanation of behavior are not available. This analysis contributes to a naturalistic view of epistemology, offering an operational view of the inference, and approaching philosophical questions with the methods of a natural science of behavior.