![Sample our Behavioral Sciences journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/110801/TOC-Subject-Sample-2021-Behavioral-Sciences.jpg"})
Abstract
In the last decade several prominent critics have charged that invocation of representations is not only not essential for cognitive science, but should be avoided. These claims have been followed by counterarguments demonstrating that the notion certainly is important in explanations of cognitive phenomena. Analyzing some important contributions to the debate, Anthony Chemero has argued that representationalists still need to explain the significance of the notion once there is an available formal account of a system and has, accordingly, challenged representationalists to provide such an explanation. This paper's first part explains why the representationalist should take an interest in Chemero's challenge. It discusses William Bechtel's account of the representational structure of the Watt Governor, which, among other things, was motivating Chemero to question the relevance of a representational account once a dynamical one is available. The second part contains the answer to Chemero's challenge. It is motivated by the thought that only a representational account of the Watt Governor with a comparable level of detail could possibly add explanatory value to a dynamical account. However, accepting the account also means that it becomes difficult to understand dynamical and representational accounts as rivals. Instead, it would be more adequate to speak of a dynamical account of the representational structure.
Notes
[1] van Gelder mentions Haugeland (Citation1991) as the inspiration for this account of representations.
[2] Data structures in computers have usually been interpreted as representations in virtue of having syntax.
[3] This account of how the governor works is a paraphrase of van Gelder (Citation1995, pp. 356–357).
[4] See his (van Gelder, Citation1995, p. 352). Van Gelder advances three other arguments against a representational account of the Watt Governor but they seem to trade on a somewhat blurred understanding on what should be taken to define representations. See Bechtel (Citation1998, pp. 303–304), for a refutation of these arguments.
[5] Compare Dennett's (Citation1998, p. 637) comment on van Gelder.
[6] This is a recurring theme in Millikan (Citation1993b, Citation1984).
[7] Misrepresentation can occur in a system as long as it is “reproductively warranted,” that is, the representation has a success rate allowing proliferation of the system and it has no better means to perform this function. How much misrepresentation this allows for is context-dependent. However, misrepresentation always makes room for improvement and substantial misrepresentation makes room for substantial improvement. Compare to this the fact that the Watt Governor as a type was generally considered to be an optimal solution.
[8] I have here presented Bechtel as offering two suggestions. In fact, he does not distinguish very explicitly between them. The second suggestion discussed presently is presented as a possible reply to someone disputing the first claim.
[9] Similar views seem to be held by so-called anti-representationalists. Rodney Brooks does not reject representation–talk simpliciter, but prefers to criticize explicit or detailed representations (Brooks, Citation1991, pp. 406–407). Varela et al. also acknowledge that the notion has use within cognitive science (Varela et al., Citation1991, p. 134).
[10] The form of representation which is applied in various discourses in cognitive science sometimes seems to be based on observed correlation or covariance, and do not operate with the possibility that the vehicle or signal could be misrepresenting on most occasions. The response to Chemero's challenge elaborated below should also be of interest to these discourses. The response does not specifically trade on the possibility of misrepresentation, but makes a more general appeal to pay more attention to the consumer when setting up the covariance relation. The point is that observed correlation, even if combined with the knowledge that information about the content might be useful for the system, is not enough to establish a representational relation that the system can use.
[11] Compare van Gelder (Citation1998, p. 622), and Chemero (Citation2000, pp. 630–631).
[12] Compare Dennett's thought experiment in Dennett (Citation1995, chapter 14).
[13] Compare Millikan (Citation1993a, p. 103).