مقاله انگلیسی رایگان در مورد مدل سازی فرایند تصمیم گیری برای حرکت مستقیم (الزویر)

Human beings perceive and behave based on uncertain information and sometimes make unreasonable judgments when viewed in objective terms. Modeling of the process used for instantly making some kind of decision—even if unreasonable—from uncertain information for which conditions can suddenly change is important not only for understanding the human cognitive process but also for constructing intelligent systems that can adapt to actual environments. Mach (Mach, 1980) and Helmholtz (Helmholtz, 1925) (Helmholts, 1962) suggested that stochastic inference provides a basis for the human cognitive process based on uncertain information. Additionally, advances in artificial intelligence research have driven research of cognition and behavior based on such stochastic inference (Knill, 1998) (Shimozaki, 2003) (Murray, 2003) (Saunders and Knill, 2004). Given that a causal relation between two events, such as “if p then q” is true, the contraposition “if not q then not p” is true, but the converse “if q then p” or the converse of the contrapositive “if not p then not q” is not necessarily true. It is ∗ Corresponding author at: 537 Kami-Hongo Matsudo, Chiba, 271-0064, Japan. E-mail address: youichi.horii.jx@hitachi.com (Y. Horry). known, however, that humans will sometimes perceive the converse to be true and make unreasonable judgments as a result. This lies in the tendency to make any asymmetry in cause-and-effect relations symmetric and is therefore called “symmetry bias,” which has come to be studied in detail through standardized experiments (Hattori, 2003) (Wasserman et al., 1990) (Anderson and Sheu, 1995) (Hattori and Oaksford, 2007). In these experiments, subjects were asked to evaluate the extent to which the casual relation of presented events could be relied upon and attempts were made to model the processes involved. For example, the contingency model of Jenkins (Jenkins and Ward, 1965) and the probabilistic contrast model of Cheng (Cheng and Novick, 1992) have been proposed. In addition, Hattori discovered that the causal inference parameter (P(p|q)P(q|p))1/2 that deals symmetrically with “if p then q” and “if q then p” had a high correlation with this sense of reliability (Hattori and Oaksford, 2007). Gigerenzer showed by experiment that human thinking is performed along the lines of Bayesian inference, and, that it can be influenced by converting expressions for rate of event occurrence into “frequency” or “probability” (Gigerenzer and Hoffrage, 1995). Meanwhile, Knill described how humans and other living organisms perform stochastic inference based on uncertain information obtained from the real world and reported that many experimentally observed examples could be explained by a Bayesian perceptual system (Knill and Pouget, 2004). In contrast to reacting suddenly to particular input, this system integrates information that propagates over space and time. He explained that this property holds not only for inputted information but also for uncertainty in the results of behavior. In addition, Manktelow (Manktelow, 2012) classified and analyzed how humans perceive a variety of uncertain events such as coin-toss games, probability that a weather report is accurate, existence of the Loch Ness Monster, etc. It was described here that Bayesian inference, which is used to update a person’s hypothesis after observing an event as posterior probability, is the basis for human intuitive cognition. There are also many studies that children’s developmental process is in line with Bayesian principles (Bonawits et al., 2012), (Goodman et al., 2011), (Gopnik and Wellman, 2012), (Griffiths et al., 2012). Pellicano, meanwhile, described how the perceptual process of autistic patients follows Bayesian inference and that autistic people tend to dislike unconventional stimuli that do not agree with experience (Pellicano and Burr, 2012). This result suggests the ability and limitations of the Bayesian inference.