In the present chapter, we investigate the notion that an action is successful if and only if (iff) it is caused by a true representation. We demonstrate that there indeed exist representations which -even though being false -can systematically lead to successful actions, if specific conditions hold, especially, if there is stochastic noise in the generation of representations and the cost of errors is asymmetrically distributed and the success-relevant feature can only be indirectly assessed via indicator features. Finally, we discuss this observation in relation to illusionary perception and evolutionary epistemology.Keywords: Semantics, success semantics, evolutionary epistemology, teleosemantics, decision theory Final Draft. To appear in: Breyer, T. (2013). Epistemological Foundations of Evolutionary Psychology. New York: Springer.
3/32
When misrepresentation is successfulHow can a representation lead to a successful action? It is widely taken for granted that a representation has to be true in order to be successful (e.g., Shea, 2007;Millikan, 1989; Whyte, 1990;Ramsey & Moore, 1927;Blackburn, 2005). Consider for instance, how could a person successfully sit down on or avoid a chair, unless she has a true visual representation of the chair"s shape and position? Or, to make use of an example of Ramsey in Ramsey and Moore (1927): How can the belief of a chicken, that a certain caterpillar is toxic be useful, unless the caterpillars are actually toxic? It even has been proposed, that "(t)ruth just is the property of a belief that suffices for your getting what you want when you act upon it." (Whyte, 1990, p. 149).Although this notion of truth being a prerequisite for success has a high face validity, we shall argue that it is wrong to suppose that all successful actions require true representations. Success does not require truth in all and every cases. Under certain circumstances, false representations can systematically cause successful actions. The goal of the chapter is to demonstrate what these defined circumstances are.
Success SemanticsThroughout this chapter, we assume a naturalistic theory of semantics, where representations, their content, their truth or falsehood are defined without recurrence to terms which themselves are already intentional. Our argumentation is invariant to what exact version of naturalized semantics is assumed (e.g., Fodor, 1990;Millikan, 1989;Papineau, 1984Papineau, , 2003Dretske, 1981). Now, what does it mean for a representation to be true vs. false or successful vs. unsuccessful?
RepresentationWe understand that observer O has representation r, if and only if (iff) r is a (physical or biological) state or signal within O, which has the property of being about something else, that is to have content. To have content can be considered as a mapping from a set of representations R
4/32(from which r is an element of) to a specific set of states in the world external to the observer, the target domain T.
2. (1) Together with the content mapping, when an observer O has re...