Information, Veridicality and Inferential Knowledge

Copyright The Board of Trustees of the University of Illinois 2017 Preprint (not copyedited or formatted) Please use the version published in American Philosophical Quarterly when citing or quoting Information, Veridicality and Inferential Knowledge Nir Fresco, Patrick McGivern and Aditya Ghose Abstract. Is information always true? According to some authors, including Dretske, Grice, Barwise, and recently, Floridi, who has defended the Veridicality Thesis, the answer is positive. For, on Floridi’s view, there is an intimate relation between information and knowledge, which is always true. It is argued in this article that information used in inferential knowledge can, nevertheless, be false, thereby showing that the Veridicality Thesis is false. Introduction Claims to the effect that cognitive agents, brains, neurones, and, in general, all living things process information are common. Yet, there is no consensus on what this information is and whether there is any hope of a unified theory of information. There are many competing theories trying to characterize information. Notably, the mathematical theory of communication (and other quantitative theories of information, such as Kolmogorov complexity) is concerned with the amount of information in a message or a signal. However, it disregards the qualitative features of information: “What makes something information?” Recently, Luciano Floridi has developed an influential theory of strongly semantic information that can supposedly serve as a foundation for information-based epistemology (2011). The idea that epistemology should be based on a theory of information is not new (cf. Dretske 1981; Grandy 1987; Barwise and Seligman 1997). On Floridi’s theory, information is semantic content--understood as meaningful wellformed data. But like others before him (e.g., Dretske 1981; Grice 1989; Barwise and Copyright The Board of Trustees of the University of Illinois 2017 Preprint (not copyedited or formatted) Please use the version published in American Philosophical Quarterly when citing or quoting Seligman 1997), Floridi insists that information must be true. The debate about the veridicality of information is still ongoing and goes beyond technical details in philosophy of information. Recent defenses of the veridicality of information are found, for example, in Sebastian Sequoiah-Grayson (2007) on logical grounds validating Floridi’s theory, in Fred Adams (2010) on epistemic grounds, and in John Mingers (2013) who emphasizes the intimate relationship between information and meaning. A defense of information being truth-like, rather than being “maximally” true, has been offered by Martin Frické (1997) and more recently by Gustavo Cevolani (2014). Floridi adopts the Veridicality Thesis (VT)--that semantic content qualifies as semantic information only if it is true (2011, p. 183) for two main reasons. One reason is to provide “a robust and intuitive link between semantic information and knowledge” (Floridi 2011, p. 184), since propositional knowledge (or knowledgethat) also requires truth. Another reason is that VT is (supposedly) required so as to avoid “some paradoxes about the alleged informativeness of contradictions” (Floridi 2010a, p. 407). The second reason has motivated Floridi’s criticism of the classical theory of semantic information (Bar-Hillel and Carnap 1953). Yehoshua Bar-Hillel and Rudolf Carnap’s theory is the first systematic true alternative to the mathematical theory of communication insofar as it focuses on qualitative, semantic features of information. Nevertheless, these two theories are formally similar, inasmuch as they are based on an inverse relationship between information and probability. The latter is based on the relationship between information and the probability of a message being selected from a set of messages (Shannon and Weaver 1949). The former is based on the idea that information eliminates possible worlds. The underlying principle was coined by Copyright The Board of Trustees of the University of Illinois 2017 Preprint (not copyedited or formatted) Please use the version published in American Philosophical Quarterly when citing or quoting Jon Barwise and Jerry Seligman as the Inverse Relationship Principle (IRP): the less likely a message is, the more informational it is (1997). Yet, IRP leads to what Floridi has dubbed (“with a little hyperbole”) the Bar-Hillel/Carnap Paradox: namely that contradictions are maximally informative (Floridi 2011, p. 109). Adopting VT avoids this paradox, since contradictions are necessary falsehoods. However, it seems clear that there are many instances where semantic content used in reasoning and cognitive processing is false and informational nonetheless. In the right circumstances, false semantic content can lead an agent to acquire true beliefs and, arguably, can even lead to (some form of) knowledge. Several authors have already rejected the idea that information must be truthful. Christopher Fox, for example, has attacked this idea starting with an analysis of the common use of information in ordinary language and “extend[ing] the results of the analysis into conclusions about knowledge” (1983, p. 188). James Fetzer has argued that accepting VT has the counterintuitive consequence that a given sentence is informational whilst its negation is not, “where no one knows which is information and which is not!” (2004, p. 226). Andrea Scarantino and Gualtiero Piccinini have argued that if one accepts VT, false non-natural information (in the spirit of Gricean nonnatural meaning) cannot qualify as “genuine” information, and hence, the view that cognition involves information processing can at best be incomplete (2010, p. 324) We are free to define “information” in a variety of different ways, and thus any dispute over VT might look merely terminological. However, as stated by Floridi, “whether false information is a genuine type of information has important repercussions on any philosophy and pragmatics of communication” (2015). The general gist of our arguments is that if we pay attention to the ways in which we use semantic content in inferential knowledge, scientific knowledge, learning, and belief- Copyright The Board of Trustees of the University of Illinois 2017 Preprint (not copyedited or formatted) Please use the version published in American Philosophical Quarterly when citing or quoting formation, then whether or not that content is true is often irrelevant to how it is used. We could draw a line between genuine information and non-information based on the veridicality of the content. Yet, this distinction would, arguably, be entirely arbitrary, and disregard important common features in the way semantic content is used. We confine the discussion to symbolic, semantic information, since it is symbolic constructs, such as sentences, propositions, utterances and statements, that are either true or false as a function of the relations they bear to states of affairs. We begin (section 2) by discussing VT and some key features of strongly semantic information so as to clarify the focus of our critique. Section 3 is the central part of this article, and it contains our arguments against VT. In section 4, we respond to two possible objections to our arguments. Section 5 concludes the article.