Bilişsel Bilim
Permanent URI for this community
Browse
Browsing Bilişsel Bilim by Author "Bilgiç, Taner,"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item ACT-R based memory models of iterated prisoner’s dilemma(Thesis (M.A.)-Bogazici University. Institute for Graduate Studies in Social Sciences, 2009., 2009.) Çetinkaya, Ayşegül.; Mungan, Esra.; Bilgiç, Taner,Iterated Prisoner’s Dilemma game is an important tool for studying cooperation in social, biological and artificial environments. Various behavioral and neuroscientific experiments point to complex decision making and memory processes for human subjects. This thesis proposes four distinct memory models of Iterated Prisoner’s Dilemma game that are built upon ACT-R cognitive architecture. This work aims to overcome the shortcomings of a previous ACT-R based memory model by Lebiere et al. (2000), by providing extensive exploration of the parameter space and analysis of simulation results for all data points. Moreover, in contrast to previus work, this study introduces distinct declarative memory modules for each player. Third, model behavior is analyzed for the cases where it plays the game not only against itself, but against basic condional and unconditional strategies as well. Finally, by implementation of three new memory models for Iterated Prisoner’s Dilemma, this study intends to attain cooperation against teaching strategies. In decision making process, all memory models evaluate expected payoffs of possible moves according to the most likely outcome making that move. First model records game history in terms of frequency and recency of possible outcomes. Second memory model records outcome patterns that are experienced in the course of the game. Third model has a two step decision process where expected payoff is calculated according to both types of information about game history. Forth model employs an association mechanism between goal and declarative modules which enable the model to record outcome history in relation to contextual information that is kept in goal module. After parameter setting, simulations are conducted for the cases where each model plays iterated game with itself and with basic game strategies. According to simulation results, all models were successful in exploiting and defending against unconditional strategies. Against teaching strategies, although they presented learning behavior, all models except third model have failed to attain cooperative equilibrium. First, second and forth models have adapted their behavior to exploit learning Pavlovian strategy and forgiving teaching strategies. All models exhibited learning behavior against basic strategies. For the cases where each model plays the iterated game against itself, all models have successfully attained cooperation in a significant portion of the games. Apart from second model, all models exhibited a learning pattern consistent with human subjects. Moreover, similar to human subjects, simulated agents can be classified into teaching and learning groups according to their behavioral patterns.Item Gigerenzer’s eclectic normativism(Thesis (M.A.)-Bogazici University. Institute for Graduate Studies in Social Sciences, 2009., 2009.) Selim, Aslı.; Kılınç, Berna.; Bilgiç, Taner,This thesis examines Gigerenzer’s criticism of classical rationality and evaluates the adequacy of the ecological rationality view that he offers in its place. Classical rationality assumes that normative standards are determined by formal logic, probability theory, and decision theory. Several studies have demonstrated that people usually fail to conform to the norms of classical rationality and concluded that people are subject to various cognitive biases and fallacies. Gigerenzer rejects this view, claiming that classical rationality is not suitable for the study of human reasoning. First, I analyze Gigerenzer's criticism of the cognitive fallacy studies and the normative benchmarks of classical rationality. I argue that rational norms need not be descriptively correct and that formal logic, probability theory and decision theory should be retained as the normative benchmarks of rationality. Secondly, I discuss Gigerenzer’s ecological rationality view, in which it is assumed that instead of formal logic, probability theory, and decision theory, psychologically plausible heuristics can be used for describing human reasoning and prescribing rational norms. I argue that the heuristics that have been proven to be effective and are suitable for prescription are not psychologically plausible and the ones that are psychologically plausible do not perform well consistently enough to be suitable for prescriptive purposes. I conclude that the study of psychologically plausible heuristics should be confined to the description of human behaviour and that the heuristics that are suitable for prescription need not be psychologically plausible.