Knowledge and luck do not mix. Our intuitions and definitions of knowledge suggest and require the absence of luck in cases of knowledge. Edmund Gettier’s landmark 1963 paper ‘Is Justified True Belief Knowledge?’ not only prompted a revision in epistemological theorising, but gave us the terms Gettier-examples and the related Gettier-luck. Gettier provided his examples in order to refute the account of knowledge which defines it as justified true belief (JTB). Here is one of the two examples he provided. It is supposed that Smith has strong evidence for the following proposition:
Jones owns a Ford (A)
Smith has another friend, Brown, of whose whereabouts he is totally ignorant. Smith randomly selects the names of three cities and uses them to construct the following three propositions:
- Either Jones owns a Ford, or Brown is in Boston. (B)
- Either Jones owns a Ford, or Brown is in Barcelona. (C)
- Either Jones owns a Ford, or Brown is in Brest-Litovsk. (D)
now, each of B, C and D is entailed by A so Smith comes to accept them. Smith therefore has correctly inferred B, C, and D from a proposition A for which he has strong justification. Hence Smith is justified in having the true beliefs of B, C and D. Now imagine in the scenario that firstly Jones does not own a Ford, but is instead at present driving a rented car. Secondly, by sheer coincidence and unknown to Smith, Barcelona happens to be where Brown is. So even though Smith clearly does not know that C is true, it is true, he believes it and he is justified in believing it.
This example along with the other example in the paper sufficed to show that truth, belief and justification were not sufficient conditions for knowledge. In both of Gettier’s actual examples, the justified true belief came about as the result of entailment from justified false beliefs; in the given example the justified false belief that “Jones owns a Ford”. This led some early responses to Gettier to conclude that the definition of knowledge could be easily adjusted, so that knowledge was justified true belief that depends on no false premises. This “no false premises” solution did not settle the matter however, as more general Gettier-style problems were then constructed or contrived, in which the justified true belief does not result using a chain of reasoning from a justified false belief.
The type of information-theoretic epistemology as expounded by Fred Dretske can easily deal with the above example from Gettier. Smith has not received the information that ‘Jones owns a Ford, or Brown is in Barcelona’, let alone the information that ‘Brown is in Barcelona’.
Other examples of epistemic luck further demonstrate the workings of this information-theoretic epistemology. Consider someone who, upon looking into a field and seeing a sheep-shaped object, forms the true belief that there is a sheep in the field. Unfortunately for this person, however, what they are looking at is in fact not a sheep but a big hairy sheep dog. Nevertheless, their belief is true since there is a sheep in the field, hidden from view behind the dog. This example is different to the ones provided by Gettier, but one reasonable response is that the conclusion there is a sheep in the field over there is based on the false premise that what is being seen is a sheep. So the adjusted definition JTB requiring no false premises, could look after this.
The notion of epistemic luck is not confined to Gettier type luck. Something termed environmental luck can also be a problem. One classic example of it is given by Alvin Goldman. Suppose there is a county in the Midwest with the following peculiar feature. The landscape next to the road leading through that county is peppered with barn-facades: structures that from the road look exactly like barns. Observation from any other viewpoint would immediately reveal these structures to be fakes: devices erected for the purpose of fooling unsuspecting motorists into believing in the presence of barns. Suppose Henry is driving along the road that leads through Barn County. Naturally, he will on numerous occasions form a false belief in the presence of a barn. Since Henry has no reason to suspect that he is the victim of organized deception, these beliefs are justified. Now suppose further that, on one of those occasions when he believes there is a barn over there, he happens to be looking at the one and only real barn in the county. This time, his belief is justified and true. But its truth is the result of luck, and thus his belief is not an instance of knowledge. Whilst Goldman took this to be a case of true belief that falls short of knowledge, it is not clear that this is not a case of knowledge.
Either way many an account of knowledge have fallen prey to invalidation via one or another of the types of luck just discussed. The informational approach to epistemology, as exemplified by Fred Dretske’s work, has seemed robust enough to deal with such cases of epistemic luck. So how does information-theoretic epistemology deal with the two problematic cases discussed above? Starting with the sheep-dog case, which is not a case of knowledge since it is just a matter of luck that the person’s belief is true, information-theoretic epistemology determines rightly that this is not a case of knowledge because their true belief that there is a sheep in the field is not caused by the information that there is a sheep in the field. The conditional probability of a sheep being in the field, given the visual signal of a sheep-shaped, big hairy dog and the person’s prior knowledge, is not 1.
Next is the case of Barney in ‘barn facade county’, which is not a case of knowledge since it is a matter of [environmental epistemic] luck that his belief is true. According to information-theoretic epistemology it can once again be argued that Barney’s true belief that there is a barn in front of him is not based on the information that there is a barn in front of him, since in this context the conditional probability of an actual barn being in front of him given the visual signal of a barn and his prior knowledge is less than 1. For example, if there were 9 barn facades in the county plus 1 real one, then this conditional probability would be 1/10. Alternatively, if Barney already knew which were the 9 barn facades, then his true belief that there is a barn in front of him would count as knowledge. This type of example is subject to the Relevant Alternatives Theory, where the fake barns are relevant alternatives. (Dretske discusses Relevant Alternatives Theory in Knowledge and the Flow of Information. A quick rundown on Relevant Alternatives Theory can be found here: http://www.jimpryor.net/teaching/courses/epist/notes/dretske.html. For a discussion of the information-theoretic epistemology take on a similar example, see [Doyle, Anthony. `Is Knowledge Information-Produced Belief? A Defense of Dretske Against Some Critics’, The Southern Journal of Philosophy XXIII (1985), 33-46.])
There are a range of purported counterexamples and objections to Dretske’s account of knowledge. The usual strategy is to provide an example where intuition would say that knowledge is not involved though Dretske’s account would say that knowledge is present. The objections generally fail because the cases they give would actually not be classed as knowledge given a genuine and correct Dretskean analysis. For a collection of these examples accompanied by defences see [Doyle, Anthony. `Is Knowledge Information-Produced Belief? A Defense of Dretske Against Some Critics’, The Southern Journal of Philosophy XXIII (1985), 33-46.].
Whilst Dretske’s account can deal with a variety of problematic cases, there is one case which merits greater consideration. The particularly cogent counterexample, found in [Maloney, J. Christopher. `Drestske on Knowledge and Information’, Analysis 43 (1983), 25-28], suggests the need for some notion of justification to be introduced into Dretske’s account. Suppose that Bob lives in a world where scientists have not yet discovered that water is H2O. Bob is an avid reader of science fiction books though, and one of the new books he is reading contains the speculative idea that water is H2O. Based on his reading, Bob forms the true belief that water is H2O. It now starts to rain, and Bob notices that a drop of water has just fallen on his hand. Since a drop of water has just fallen on his hand, and water is H2O, it follows that Bob has received the information that H2O has just fallen on his hand. But though he forms the true belief that H2O has just fallen on his hand due to the information that H2O has just fallen on his hand, it would be wrong to say that he has the knowledge that H2O has just fallen on his hand. The issue here is that Bob’s true belief about the identity of water with H2O, a belief integral to the formation of this new true belief, is just mere true belief.
A more general point is to be extracted from this example. Say there are two pieces of information, A and B, associated with a signal s, such that A is information if and only if B is information. Someone receives the signal and though they only know that A is associated with s, they somehow come to form a belief about B. But though the signal that contains B as information causes the belief about B this is not really a case of knowledge. Here are some possible ways to approach this apparent issue, discussed in reference to the Bob example above.
- The true belief in the case of Bob was not completely information caused. It was not just because a raindrop fell on his hand that Bob formed his belief. The belief was also due to the non-information that water is H2O, which is not information in this case because it was the result of a lucky, or at least semi-educated guess.
- Further qualification to the definition of information and the informational value of a signal, as hinted at in the section ‘Modifying Dretske’s Definition of Information’. Even though the signal of a raindrop falling on Bob’s hand contains the information that water is falling on his hand and the information that H2O is falling on his hand, it is only the information that water is falling on his hand which is of any informational value, because Bob only knows of the information association between raindrop and water. If we are going to speak of a signal’s information causing belief, it must be because that information has informational value to the agent in this sense. If someone who does not know that fire causes smoke sees some smoke and for some reason takes a guess that there is fire, then they are not actually using the information that there is fire because they do not know of the connection between fire and smoke.