Reliabilism, continued

June 3, 2012

I’ve been saying that Brandom contrasts his own position w/r/t perceptual knowledge with a ‘reliabilist’ one. What does ‘reliabilism’ consist in?

All the theories of perceptual knowledge in contention in this theoretical space agree that for something to be knowledge it must be:

a) a belief
b) true.

All the theories in contention furthermore agree that not all true beliefs are knowledge: an additional criterion or criteria are required in order to distinguish knowledge from the larger class of true beliefs.

The classic additional criterion to distinguish knowledge from true belief is justification: on this account, something is knowledge if it is justified true belief. But, notoriously, it is difficult to give an account of what such justification can consist in that gives results compatible with our intuitions as to what does and does not count as knowledge.

Raymond Gettier’s classic article ‘Is Justified True Belief Knowledge?’ poses a series of dilemmas for the straightforward ‘justified true belief’ position –these dilemmas have been extended and added to by subsequent philosophers. [An example of a relatively simple ‘Gettier problem’: let’s say I put ten coins in my pocket this morning; later in the day, five coins fall out of my pocket; still later in the day, a kindly sprite places five more coins in my pocket; I now believe, correctly, that there are ten coins in my pocket, and I am justified in that belief, because I remember putting ten coins in my pocket earlier, but can I really be said to know that there are ten coins in my pocket, given the actual reasons why there are ten coins there, which are not at all the ones I imagine? These sorts of examples are always enormously silly and tedious; I’ll try to keep them to a minimum in the remainder of the discussion; I promise we’re headed places that have more bearing on things that people actually sometimes care about.] Much subsequent work within analytic philosophy has focused on ways in which the ‘justified’ criterion can be made more complicated and sophisticated to take account of such ‘Gettier’ problems.

One obvious issue with the ‘justified’ criterion, is whether the possessor of knowledge needs themselves to be conscious or aware of the justification for their true beliefs, in order for those beliefs to be justified. If we decide that the believer must themselves be aware of the justification for their belief, we seem (at least potentially) to be vulnerable to an infinite regress of the ‘KKp’ type. That is, if we must know the basis for our knowledge, then it seems we must also know the basis for that knowledge of our knowledge’s basis, and so on. This looks to be a vicious infinite regress – which in turn implies that at some point we need to bottom out w/r/t knowledge (or even belief concerning basis of knowledge) – another version of the ‘regulism’ problem I have already discussed on the blog.

If we assume, then, that we do not necessarily need to know why our knowledge is justified, in order for it to be justified, we seem to open the nature of justification to being something extra-psychical or -meaningful – something outside the ambit of consciousness or of social practice. A series of post-Gettier accounts of knowledge aim to cash out the ‘justification’ criterion in terms of the causal mechanisms that connect knowledge claims and the objects of those knowledge claims. For example, Alvin Goldman’s 1967 ‘A Causal Theory of Knowing’ argued that the missing justification-criterion was an appropriate causal connection between the object of the knowledge-claim and the knowledge-claim itself (potentially regardless of how the knowledge-claimer themselves understand that causal chain).

There are various problems with a straightforward causal theory of knowledge. One of them is that we can imagine or construct scenarios that involve exactly the same causal mechanism connecting a knowledge-claim and the object of knowledge, where it seems that some of these scenarios genuinely produce knowledge, but others do not.

An example here, much beloved of Brandom, is the ‘Barn Façade County’ thought experiment. [I think the origin of this is Goldman’s 1976 ‘Discrimination and Perceptual Knowledge’, though it’s obviously been extended since that paper, and I’m not sufficiently abreast of the literature to be able to track the thought experiment’s progress.] In this thought experiment a perceiver (call him, following Goldman, Henry) perceives what appears to him to be a barn. Unbeknownst to Henry, however, he is presently in Barn Façade County – a district the favourite pastime of whose inhabitants is constructing incredibly realistic facades of barn-fronts. [Sorry again about the moronic analytic philosophy thought experiments.] Here it seems unreasonable to say that Henry knows that there is barn in front of him. Nevertheless, it also seems unreasonable to say that in most other circumstances Henry is unable to know about the existence of a barn, because it is always possible that he be in Barn Façade County or some similarly idiosyncratically deceptive environment.

The problem here is a version of a general skeptical problem. It is always possible that we be mistaken about any given claim – but we don’t want to conclude, because of this, that we can never have knowledge of anything.

The epistemological theory – a version of the causal theory of knowing – that many philosophers adopt in order to give a non KKp-vulnerable account of the basis of justification, is that a person knows P if they believe P, P is true, and the person is generally a reliable reporter of P. This is ‘reliabilism’.

Brandom does not dispute the claims in the previous paragraph. But he disputes that they are enough to give us an account of justification.

Returning again to the Barn Façade County example, Henry can be a generally reliable reporter of the presence of barns (in almost all counties), and an extremely unreliable reporter of the presence of barns within Barn Façade County. Whether or not Henry is a reliable reporter of the presence of barns is relative to the environment in a way that does not impact on the causal connection between the object of Henry’s knowledge and his knowledge claims.

Reliablist theories of knowledge try to take this problem into account. As Goldman puts it in his 1976 paper:

A person knows that p, I suggest, only if the actual state of affairs in which p is true is distinguishable or discriminable by him from a relevant possible state of affairs in which p is false… The qualifier ‘relevant’ plays an important role in my view. If knowledge required the elimination of all logically possible alternatives, there would be no knowledge (at least of contingent truths). If only relevant alternatives need to be precluded, however, the scope of knowledge could be substantial. (p. 773-4)

Superficially, this class of positions appears to give us a definition of justification which is wholly separable from the social practice of taking-as-justified. On the reliabilist account, a person is justified in believing p if they reliably believe p when p, and not-p when not-p, in relevant scenarios. Whether or not the person exhibits this reliable responsive relationship to an objective fact about the world seems itself to be an objective fact about the world – it is not dependent on (narrowly understood) social acts of taking-as (other than the belief that p or not-p itself, and whatever mechanisms led to that belief-formation).

And in a sense this is true. But the devil is in the detail – or, more precisely, in the concept of ‘relevance’. Relevance cannot be objectively defined. Whether or not a scenario is ‘relevant’ is not a bare fact about the physical make-up of the world. To take a scenario to be relevant is to take it to be relevant. This means that the reliablist account must be folded back into an account of social (or at least normative) acts of taking-as.

As Brandom puts it, in relation to the ‘Barn Façade County’ example:

Goldman’s idea is that reliability is an objective affair, determined by the objective probability of a correct judgment, given one’s circumstances. But such probabilities vary with the specification of those circumstances. Given a reference class of relevantly similar cases, frequencies of success define objective probabilities. The question remains how a privileged reference class is to be determined. What is the correct reference class with respect to which to assess such probabilities?

Focusing on the relativity of reliability to decisions about where to draw these boundaries makes it evident that the question “Reliable or not?” is underdetermined in exactly the same way that the question “Regular or not?” is underdetermined. There are always some regularities that are being instantiated, and (in the case where the claim one is making is true) there are always some reference classes with respect to which one is reliable. Using these naturalistic notions to stand in for genuinely normative assessments works only relative to some way of privileging regularities or reference classes. (MIE p. 211)

This privileging of reference classes – which provide the framework within which naturalistic judgements of reliability can be made – is itself a normative practice; which means, if Brandom’s apparatus is correct, a social practice. Causal and reliabilist theories of knowledge, therefore throw us back onto social processes.

It is worth noting, in passing, that this provides another example of the (virtuous) circularity of Brandom’s own apparatus. Brandom too, of course, relies upon the concept of reliability to play a foundational explanatory role: Brandom’s entire apparatus is built up out of reliable differential responsive dispositions. Yet Brandom is clear that what constitutes reliability can only be judged from within a normative (social) framework that provides reference classes from within which reliability can be assessed. RDRDs provide the foundation for Brandom’s apparatus, therefore – but to discuss RDRDs at all requires a social framework generated by those RDRDs, in their complex interaction. This is true for many reasons – but the necessity of normatively determined reference classes for the analysis of reliability is one such reason.


One Response to “Reliabilism, continued”

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: