Abstract
In this paper, we offer an alternative interpretation for the claim that ‘S is justified in believing that φ’. First, we present what seems to be a common way of interpreting this claim: as an attribution of propositional justification. According to this interpretation, being justified is just a matter of having confirming evidence. We present a type of case that does not fit well with the standard concept, where considerations about cognition are made relevant. The concept of cognitive algorithm is presented and explained. Finally, the new reading of ‘S is justified in believing that φ’ is fleshed out. According to this interpretation, being justified in believing that φ is not just a matter of having evidence in favor of φ, but also of having a cognitive algorithm available such that it allows one to form belief in φ on the basis of the relevant evidence.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
We are using the concept of confirmation in a stronger sense than the usual one here. In general, it is assumed that ψ confirms φ when ψ raises the probability of φ, that is, when P(φ | ψ) > P(φ) for some probability function P obeying the standard axioms for the Probability Calculus. We are assuming, however, that ψ confirms φ when φ is more probable conditional on ψ than ~ φ is, that is, when P(φ | ψ & b) > P(~φ | ψ & b), where φ must be a contingent proposition and b is some background evidence. The concept of confirmation is used here as an umbrella term for both inductive and deductive support between propositions.
In Turri’s (2009) classification of the ontological views about epistemic reasons, philosophers defending this view are called ‘statists’. We use the concepts of evidence and reasons interchangeably here.
In order to avoid too much wording sometimes we will use ‘evidence E confirms φ’ as short for ‘the propositional content of E confirms φ’. Also, when we write that ‘S’s evidence confirms φ’ we mean that S’s total evidence confirms φ, so that we need not make reference to a ‘no-defeater’ condition every time.
If, however, bodies of evidence or reasons are propositions, an authentic expression of the fact that S has propositional justification for believing φ would be something like: S is in the possession of a body of evidence E such that, on balance, E confirms φ (where E is a set of propositions). For a recent defense of the view that evidence or reasons are propositions, see Dougherty (2011). We will talk of evidence or reasons as mental states here, but the reader can translate mentalistic claims into propositionalist claims in the following way: whenever it is said that a certain mental state is the evidence one has, we can interpret the claim as saying that the relevant mental state gives one evidence (the proposition that is the content of the relevant doxastic state). John Turri (2009) offers compelling arguments against propositionalism. For an influential propositionalist view about evidence, see Williamson (2000), where evidence is equated with the set of known propositions.
One could claim that the epistemically relevant factor here is the mentioned set of counterfactual truths about A and B itself. Accordingly, one could try to explain propositional justification by means of doxastic justification in counterfactual situations. Goldman (1986), for example, says that we can analyze what he calls ‘ex ante justification’ (propositional justification) in terms of ‘ex post justification’ (doxastic justification). S is taken to be justified in believing that φ in a certain situation when S would come to justifiably believe that φ in that situation (if S were to believe that φ at all). But, presumably, there is a certain property had by A but not by B that plays a crucial role on the explanation why these counterfactuals hold. The truth of the counterfactuals is a consequence of the fact that A has this property and B lacks it.
We are using squared brackets ‘[ ]’ for representing propositions and lowercase sentence letters to name propositions. Also, we use ‘⊃’ for the material conditional.
The points we are about to make are independent of the charge of infinite regress that could be raised against this kind of view. The possibility of infinite regress here could be expressed in the following way: if, in order for S to be justified in believing φ (given her evidence E) S must also believe that E is a reliable indicator of the truth of φ then, apparently, there is no reason to stop reiterating this requirement.
Of course, by saying that Conee and Feldman advocate an approach that is similar to the view just considered, we are not implying that they are trying to ‘save’ a purely confirmational account of justification either.
There is a disanalogy between Smith’s case and Conee and Feldman’s example of the investigator, however. We will get back to this later—the disanalogy is not important now.
Conee and Feldman actually use the concept of scientific evidence, which is to be contrasted with the concept of justifying evidence (2008, 85). In the example of the investigator, f is supposed to be scientific evidence for l but not justifying evidence for l. e is scientific evidence for p when e is publicly available and e reliably indicates the truth of p. We are not sure that Conee and Feldman use the concept of scientific evidence in a consistent way. At first it seems that the concept applies only to objects or events in the world (things that are publicly available), for example, fingerprints in guns and smoke rising from the chimney. But if the evidence that the investigator has is its knowledge of f, then we assume that his evidence is not scientific evidence—for the investigator’s knowledge is not publicly accessible.
It would not help here to point out that this is a rare, unusual, condition.
Pollock and Cruz (1999, 123).
These structures bear some similarity with what some cognitive psychologists call ‘production rules’. However, sometimes cognitive psychologists interpret such production rules as representational items that are part of the cognitive system’s ‘knowledge’ or available information. We are not assuming, though, that cognitive algorithms are represented inside our heads, and we are assuming that one may cognize in accordance with a cognitive algorithm and have no clue as to what algorithm one is instantiating. There may be other interpretations for the concepts of rules and algorithms in the cognitive psychology literature, however, which are more akin to our concept of cognitive algorithm. The reader can find an instructive treatment of the concepts of rules, procedures and algorithms used in cognitive psychology in Thagard (2005, ch. 1, 3). Also, the present purpose is not supposed to assume or imply the view called ‘machine functionalism’ in philosophy of mind (see Clark 2001). We do not intend to explain thought as a process of computation or something like that—our use of the concept of cognitive algorithm is only intended to have a normative role in an epistemological theory of justification (which is not to say, again, that the algorithms themselves are epistemic norms).
Of course, S could also be reasoning in accordance with a ‘special purpose’ algorithm, not in accordance with ‘general purpose’ algorithms. We owe this point to Peter Klein.
It is noteworthy that the propositional forms in TESTIMONY are more fine-grained than the propositional forms in CONJUNCTION-ELIMINATION. Also, TESTIMONY involves terms from natural language. It would be more accurate for us to use the same level of representation for both descriptions, and it may be necessary to include background information available to the agent in the body of algorithms. But this rough characterization is only supposed to give the reader some insight on what we have in mind here.
This algorithm is named after the derivation rule conjunction-elimination from propositional logic, but this is just a matter of convention—the algorithm is not the same thing as that derivation rule.
The relevant algorithm would have the set {B(~φ), B(φ ⊃ ψ)} as input and B(~ψ) as output.
We are not using examples of algorithms with experiences as input because there may be some controversy as to how represent this kind of input. Of course, such controversies may be related to philosophical accounts of perceptual and mnemonic experience. But this is not the place to go into this.
We would not oppose, however, if someone were to point out that this is a kind of reliabilism, or a variation of reliabilism, about having justification. But there are important differences between our approach and standard reliabilist accounts of justification. As developed by Goldman (1986, ch.5), the reliabilist theory of justification is framed on the basis of the notion of a belief-forming process type. However, as we pointed out before, cognitive algorithms are not process types. Also, we are dealing with the state of having justification here—not with the state of justifiably believing. For a version of reliabilism about doxastic justification where conditional probability (instead of ratio of true beliefs) is used to account for the reliability of belief-forming processes, see Comesaña (2009). We thank Alvin Goldman for pressing on this point.
Hájek (2012) contains a precise characterization of the different interpretations of probability.
In general, a theory of confirmation must also solve several puzzles about inductive support, like the Raven Paradox. See Swinburne (1971).
Nothing said here implies that the awareness requirement is not necessary for justification—only that it is not, together with the presence of confirming evidence, sufficient for justification. The algorithmic condition is not supposed to be a substitute for the awareness condition. Also, the conditions at (J) are only claimed to be necessary for justification.
Kelly (2006), section 3, deals with the background information issue and its importance to epistemological judgments about what is justified for one to believe.
So we concur with John Turri’s (2010) contention that the truth of ‘S is propositionally justified in believing φ in virtue of having a certain set of reasons R and S believes that φ on the basis of R’ is not a sufficient condition for S to be doxastically justified in believing that φ. This is because S can form her belief that φ on the basis of R but still do so while instantiating a non-optimal cognitive algorithm. Also, if we are assuming a causal account of the basing relation, the instantiation of an optimal algorithm would exclude ‘deviant causation’ of the belief on the basis of its reasons. So far, however, we make no commitments to the causal account of the basing relation. A more fully developed explanation of doxastic justification and its proper relation to cognitive justification will be dealt with in our future work. For related discussion, see Turri (2011), where the concept of manifestation of a cognitive trait is used to deal with the problem of deviant causation (the notion of a cognitive trait used by Turri bears some similarity with what we call here a ‘cognitive algorithm’).
Adapted from Conee and Feldman (2008, 83).
References
Clark, A. (2001). Mindware. New York: Oxford University Press.
Comesaña, J. (2009). What lottery problem for reliabilism? Pacific Philosophical Quarterly, 90(1), 1–20.
Conee, E., & Feldman, R. (2008). Evidence. In Q. Smith (Ed.), Epistemology: New essays (pp. 83–104). New York: Oxford University Press.
Dougherty, T. (2011). In defense of propositionalism about evidence. In T. Dougherty (Ed.), Evidentialism and its discontents (pp. 223–232). New York: Oxford University Press.
Goldman, A. (1986). Epistemology and cognition. Cambridge: Harvard University Press.
Hájek, A. (2012) ‘Interpretations of probability’, The Stanford encyclopedia of philosophy (winter 2012 edn), ed. Edward N. Zalta: http://plato.stanford.edu/archives/win2012/entries/probability-interpret/
Kelly, T. (2006) ‘Evidence’, The Stanford encyclopedia of philosophy (fall 2008 edn), ed. Edward N. Zalta: http://plato.stanford.edu/archives/fall2008/entries/evidence/
Pollock, J., & Cruz, J. (1999). Contemporary theories of knowledge. Maryland: Rowman & Littlefield.
Swinburne, R. (1971). The paradoxes of confirmation—a survey. American Philosophical Quarterly, 8(4), 318–330.
Thagard, P. (2005). Mind: Introduction to cognitive science (2nd ed.). Cambridge: MIT Press.
Turri, J. (2009). The ontology of epistemic reasons. Noûs, 43, 490–512.
Turri, J. (2010). On the relationship between propositional and doxastic justification. Philosophy and Phenomonological Research, LXXX(2), 312–326.
Turri, J. (2011). Believing for a reason. Erkenntnis, 74, 383–397.
Williamson, T. (2000). Knowledge and its limits. New York: Oxford University Press.
Acknowledgement
We thank Peter Klein, Alvin Goldman, Rodrigo Borges and Kurt Sylvan for valuable suggestions on previous versions of this paper. We also thank Claudio Almeida for giving us space to present this paper in his epistemology seminar at PUCRS (Brazil), where we had a fruitful discussion about the topics discussed here.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Rosa, L. Justification and Cognitive Algorithms. Philosophia 42, 503–515 (2014). https://doi.org/10.1007/s11406-013-9506-6
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11406-013-9506-6