The decoupled representation theory of the evolution of cognition--论文代写范文精选

有一个思想演变的认知,直观的和普遍模式的认知，是环境复杂性的选择,需要某种表征能力。然而,有许多不同的方式可以应用。许多理论喜欢将非认知/认知差异进行对比。这个想法与动物研究认知的普遍观点同样被定义。下面的paper代写范文进行讲述。

Abstract
Sterelny’s Thought in a Hostile World (2003) presents a complex, systematically structured theory of the evolution of cognition centered on a concept of decoupled representation. Taking Godfrey-Smith’s (1996) analysis of the evolution of behavioral flexibility as a framework, the theory describes increasingly complex grades of representation beginning with simple detection and culminating with decoupled representation, said to be belief-like, and it characterizes selection forces that drive evolutionary transformations in these forms of representation. Sterelny’s ultimate explanatory target is the evolution of human agency. This paper develops a detailed analysis of the main cognitive aspects. It is argued that some of the major claims are not correct: decoupled representation as defined doesn’t capture belief-like representation, and, properly understood, decoupled representation turns out to be ubiquitous amongst multicellular animals. However some of the key ideas are right, or along the right lines, and suggestions are made for modifying and expanding the conceptual framework.

Introduction
There is a bewildering range of ideas on the evolution of cognition. An intuitive and widely held schema for the evolution of cognition is that environmental complexity selects for behavioral flexibility, which in turn requires some kind of key representational ability. There are, however, many different ways that this schema can be applied. Many theories attempt to cast the non-cognitive/ cognitive distinction in terms of a contrast between reactive and anticipative creatures. For instance, Dickinson and Balleine (2000) contrast habit machines with capacity for goal-directed action, which they find in rats.

This idea is connected to the widespread view in animal cognition research that cognition is to be defined as, or at least can be diagnosed in terms of, behavior control that can’t explained in terms of basic associative learning mechanisms (Zentall 1999). Another kind of distinction used to frame the evolution of cognition is between domain specific and domain general abilities; a major transition is thought to occur from creatures that only rely on domain-specific abilities to those capable of cross-domain integration. Menzel and Giurfa (2001), Spelke (2003) and Premack (2007) all appeal to this contrast, but whereas Menzel and Giurfa see flexible integration in honeybees, Spelke and 1 Decoupled representation - a critical evaluation Premack think that domain general cognition is uniquely human. Spelke proposes that the first mechanism supporting domain-general cognition is language.

Ideas about specific representational mechanisms are also varied. The most general definition of cognition with positive content sees it as some kind of information processing that intervenes between receipt of the stimulus and the response (Shettleworth 1998; Zentall 1999). Tomasello and Call (1997, p. 11) regard the basis for cognitive flexibility as lying with internal models that can manipulated in advance of interaction with the world. Language is a perennial favorite explanation for human uniqueness, but other proposals include hierarchical ‘mental construction’ abilities (Gibson 2002) and capacity for analogical reasoning (Gentner 2003). Herrmann et al. (2007) even suggest that there are no significant cognitive differences between humans and apes (other than social), and that culture accounts for enhanced human cognition.

Thus, not only is there quite a range of different specific ideas on the evolution of cognition, there are diverse ways of conceptualizing it. In this context Sterelny’s account is notable because it directly and systematically addresses many of the core issues, offering an account of the nature of the selection that drives cognitive complexification, and a taxonomy of complexification culminating in belief-like representation. The account also analyses a transition from drive-based motivation to motivation involving preferences, and in combination these elements provide the core elements for understanding the origins of belief-desire psychology, central to many conceptions of human agency.

Sterelny’s account is therefore an important touchstone for theorizing the evolution of cognition, and what follows is a detailed analysis of the main cognitive aspects.1 The core structure of the theory is outlined in section 2, and section 3 examines in more depth the taxonomy of behavior control forms and the analysis of the selection that drives transformation in behavior control. Sterelny’s account is built around the idea that belieflike representation is associated with a late evolutionary separation of indicative and imperative function; this is questioned, and reasons are given for thinking that complexification of perception-behavior relations should occur much earlier than Sterelny’s account suggests.

Section 4 considers a range of evidence in support of this, including sensori-motor complexification in early neural evolution, and evidence for both an advanced kind of decoupled representation (termed ‘model-based representation’) and preferences in rats, drawing on transitive inference, maze and incentive revaluation experiments. Section 5 integrates this evidence into a revised theoretical picture that clarifies the sources of selection pressure and forms of control involved in the evolution of cognition. With basic versions of the cognitive abilities underlying beliefs and desires present in rats, it is suggested that, rather than being associated with the appearance of new kinds of representational abilities, as proposed by Sterelny and many others, human agency may be an endpoint of a long progressive elaboration and refinement of cognitive abilities supporting flexible goal-directedness.

The core structure of the theory
In thumbnail sketch Sterelny’s theory is as follows. The account begins with GodfreySmith’s (1996) analysis of selection for flexible control of behavior, which says that flexible response to environmental variation is adaptive when the benefits of detecting the variation and responding to it outweigh the costs. Sterelny’s theory describes a succession of increasingly complex forms of sensorimotor control, beginning with a detection system baseline involving specific adaptive response to a specific environmental signal (p. 14)2 . He defines a detection agent as an organism equipped only with detection systems (p. 14). The next grade of complexity is robust tracking, which involves tracking important features of the environment using multiple cues (p. 17). The culminating grade, decoupled representations, are “...internal states that track aspects of our world, but which do not have the function of controlling particular behaviors.” (p. 29).

The selection pressure driving this transformation is translucency, defined as the condition where the functionally relevant features of the environment map in complex ways onto the cues the organism can detect (p. 21). These perceptuo-motor forms appear progressively, with decoupled representation being associated with social cognition in great apes.

This empirical distribution lends plausibility to the claim that decoupled representation is the basis for beliefs in humans. In addition to these stages of cognitive complexification Sterelny gives an account of the evolution of motivation describing two stages of motivational complexification: drives and preferences. Drives are a non-representational form of motivation control, whereas preferences represent goals for action and can be formed and modified through learning (pp. 92-95). Sterelny argues that drive-based motivation will be inadequate when the animal’s behavioral repertoire include many options and it must access a wide range of ecological resources (pp. 92-94). He claims that this transition has occurred in the human lineage but is incomplete. I now explicate the account in more detail, schematizing the main elements and formulating the account in terms of models in order to make the overall structure clearer and to provide points of reference for comparison.

The Environmental Complexity
Thesis Sterelny’s account begins with Godfrey-Smith’s (1996) account of the evolution of behavioral control, which Sterelny describes as a framework (p. 11). Godfrey-Smith conceptualizes the idea that cognition evolves in response to environmental complexity as the environmental complexity thesis (ECT), according to which the function of cognition is to enable the agent to deal with environmental complexity. To make this more precise he develops an analysis of the circumstances under which environmental complexity will select for flexible behavior (1996, ch. 7). This analysis poses the problem in terms of the two possibilities (1996, p. 207) depicted in figure 1. In the baseline state the animal’s sensory discriminations don’t distinguish between distinct environmental conditions, and the animal’s behavior must be adaptive across these hidden variations. Effectively the animal employs a fixed behavioral strategy in varying conditions. In the alternative state the animal discriminates the variations in environmental state and produces behaviors matched to each condition. The question is what conditions will favor 1b over 1a.

The evolution of beliefs and desires After outlining the Environmental Complexity Thesis Sterelny elaborates an account of the evolution of complex cognition, with the ultimate aim of understanding human agency (p. 4). He addresses what are standardly seen as the two fundamental components of human agency: beliefs and desires. In part his account examines the status of folk psychology, but this aspect is largely bypassed here since it makes little difference to the core structure of the theory. Sterelny uses theoretical definitions of beliefs and desires, and places the burden of justification for these definitions on the overall success of the account (p. ix). Questions will be raised about these definitions below, but shifting the strategy of justification to the theoretical and empirical analysis of cognitive architecture is a reasonable move.

Formal structure of the account Figure 2 depicts Sterelny’s account of the stages of complexification leading to decoupled representation. This will be called representational/control complexification here because it straddles both representation and behavior control. Robust tracking increases the sensory pathways between a key environmental condition and behavior, whilst decoupled representation increases the number of behaviors to which a given sensory discrimination contributes.4 As Sterelny notes, on the face of it robust tracking and decoupled representation are independent kinds of complexification (e.g., p. 36). Thus, in principle decoupled representation could be based on detection, as depicted in figure 3a. In practice Sterelny thinks this situation is unlikely (pp. 31-32), and that the typical case will be as depicted in figure 3b, with decoupled representation based on robust tracking. This gives something of a stage-structure to the three forms, with robust tracking being a precondition for decoupled representation. Thus, the account can be interpreted as proposing the evolutionary model 2a → 2b → 3b.

Sterelny argues that the selection force driving this complexification is informational translucency. Translucency is defined in contrast with informational transparency: an environment is informationally transparent for an animal when there is a simple and reliable correspondence between sensory cues and the functional properties of the environment that determine success (p. 20). That is, 2a will be adaptive. It is translucent when the functionally relevant features of the environment map in complex ways onto the cues the organism can detect (p. 21). In this condition 2a is no longer adaptive, and selection will drive a population towards 2b and 2c/3b.(paper代写)

51Due网站原创范文除特殊说明外一切图文著作权归51Due所有；未经51Due官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯著作权现象，51Due保留一切法律追诉权。
更多paper代写范文欢迎访问我们主页 www.51due.com 当然有paper代写需求可以和我们24小时在线客服 QQ:800020041 联系交流。-X(paper代写)