Functionalism about inference

ABSTRACT

Inferences are familiar movements of thought, but despite important recent work on the topic, we do not yet have a fully satisfying theory of inference. Here I provide a functionalist theory of inference. I argue that the functionalist framework allows us the flexibility to meet various demands on a theory of inference that have been proposed (such as that it must explain inferential Moorean phenomena and epistemological ‘taking’). While also allowing us to compare, contrast, adapt, and combine features of extant theories of inference into one unified theory. In fleshing out the inference role, I also criticize the common assumption that inference requires rule-following.

KEYWORDS:

• inference
• functionalism
• inference rules
• philosophy of psychology
• epistemology

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The ‘movement of thought’ terminology is from Wright (2004).

2 This is true even if contents arise from transitions on uninterpreted syntax. So this point is not in conflict with any version of inferentialism or conceptual role semantics (see Section 6).

3 Murzi and Steinberger (2013) use the term ‘entertaining’ to cover these attitudes. We might also include partial belief, but I will be neutral about this. See Staffel (2013) for discussion.

4 Invented by Gentzen (1934) and Jaśkowski (1934).

5 Frege (1960) analyzed rejecting a sentence as accepting the sentence's negation. But Frege's analysis is no longer popular – see Priest (2006) and Ripley (2011). I prefer to think of rejection as sui generis, but nothing in my discussion requires this. This option requires a slight generalization of validity – see Smiley (1996).

6 This is noted in Broome (2014b).

7 See also Broome (2014a, 2014b).

8 In Quilty-Dunn and Mandelbaum (2018).

9 Boghossian (2014, 3) gives a similar example, attributing it to Alvin Plantinga.

10 Of course, even psychologically direct examples involve cognitive processes that result from various sub-processes. They are direct only at the psychological level of description.

11 Boghossian (2014, 4) also gives an example like this.

12 Influential versions of functionalism were developed by Putnam (1960, 1967), Fodor (1968, 1987), Armstrong (1968), and Lewis (1966, 1980). Putnam and Fodor think the functional role of mental states come from cognitive psychology. Armstrong and Lewis think they come from folk theory. Arguments against functionalism are in Block (1978), Kripke (1982), Putnam (1988), Searle (1980), and Shoemaker (1982). For applications outside of philosophy see the discussions in Dennett (1991a), Marr (1982), Minsky (1986), and Pinker (1997).

13 See Horgan (2013) and Pautz (2013), for example.

14 See Sellars (1962) and Dennett (1987).

15 See Dennett (1991b).

16 Pace the eliminativism of Churchland (1981).

17 See Squire and Kandel (1999).

18 I have worries about ‘computation’ and ‘representation’ talk in neuroscience and psychology, but I will not insist on them here. See Cao (2018) for some relevant discussion.

19 From Boghossian (2014).

20 See Hlobil (2014) and Wright (2014) for early arguments against taking. See McHugh and Way (2016) for paper-length criticisms. Broome (2014b) accepts taking but does not think we should impose it up front.

21 See Boghossian (2018).

22 See McHugh and Way (2018). See Hlobil (forthcoming) for criticism.

23 See the discussion in Boghossian (2018).

24 Boghossian (2014, 10–11).

25 See Martin (1994), Johnston (1992), and Smith (1977) for some of the ways.

26 See my (2020a) response to Kripke (1982).

27 In Neta (2013). His account includes a linguistic analysis of constructions like ‘so’ or ‘therefore’ as picking out some contextually salient doxastic relationship.

28 See Hlobil (2014).

29 For details, see Moore (1982).

30 For details on this and other similar cases, see Burgess (2005).

31 Boghossian (2014, 3).

32 Boghossian recognizes this. In his (2014) he characterizes inferences as ‘System 1.5’ reasoning. Contrasting with Systems 1 and 2 of Kahneman (2011). In his (2018), he distinguishes between inference 2.0, 1.75, and 1.5.

33 This helps to explain why being primed into a belief is not an inference, since presumably the only control we could exert would be by avoiding primes or having been more attentive at an earlier time. Also, of course, it is extremely unlikely that a primed belief would satisfy counterfactual, but perhaps it could via some kind of stable confabulation.

34 Boghossian does not use the ‘procedural’ terminology, but his (Email Rule) is procedural in my sense.

35 For Broome's rule-based approach, see his (2014b).

36 See Harman (1986). See $4.\mathrm{VIII}$ of my (2020b) for a more nuanced account of the normative role of inference rules.

37 In Kripke (1982).

38 In Boghossian (2014).

39 Sylvan (2016) offers a similar but more general approach that appeals to Sosa's (2015) notion of a ‘competence’. He argues that this has some advantages over Boghossian's view, but notes that it shares the cost of taking a psychological notion as a primitive.

40 See my (2020a) for a full response to Kripkenstein's anti-naturalism about rule-following in general, and and section 2.VIII of my (2020b) for a dispositional analysis of inference rule-following.

41 Valaris (2017) has argued against a inference-rule-based approaches to reasoning in favor of a semantically rich approach.

42 See Goodman (1955). Personally I do not think the grue puzzle makes broadly formal approaches to inductive logic impossible, but I will not insist on this here since it is controversial.

43 Norton (2003) explicitly makes the case that induction is not rule-governed in the way that deduction is.

44 See Wason (1966, 1968) and Wason and Shapiro (1971) for the Wason selection task. See Cosmides and Tooby (1992) for the evolutionary case against general rules.

45 See Davies, Fetzer, and Foster (1995), for example.

46 See van Bentham (2008).

47 Warren (2017) argues for this.

48 Compare Brandom's (1994) treatment of material inference.

49 As Boghossian (2014) notes (pages 17–18), the rule-following approach to inference puts pressure on particularism about reasons. Dancy (2004) defended particularism in morality.

50 In his (2018).

51 For Quilty-Dunn and Mandelbaum (2018) they are ‘rich’ inferential transitions, in contrast to the more fundamental ‘basic’ inferential transitions.

52 In Malmgren (2018).

53 See Boghossian (2014, 2018), Broome (2013, 2014a, 2014b), Hlobil (2014), and Wright (2014).

54 See Kornblith (2012) and Quilty-Dunn and Mandelbaum (2018).

55 This might be a bit too strong – bad habits are, in a sense, the thinker's responsibility. The point here is that unconscious processes are not the thinker's responsibility in the same way that inferences are.

56 Neither Kornblith (2012) nor Quilty-Dunn and Mandelbaum (2018) substantively disagree with this, from what I can tell.

57 Thanks to Rosa Cao, Daniel Waxman, Jack Woods, and several anonymous referees. I first finished a paper draft giving an account of inference in July 2014. This paper began as a major revision of that one. The revision was finished in February 2017. This paper took a long time to publish, earning numerous split verdicts, so over the last (nearly) four years I have slightly refined the theory and (I hope) greatly improved the writing. Shorter accounts of my functionalist theory of inference can be found in chapter 2 of my (2020b) and in my (2021). About this paper's first two sentences: I do not have and have never had a dog named ‘Chrysippus’ and I am actually very fast.