Study Guide

Week Three: What is Knowledge?

Materials Assigned for the Week
The Central Point of This Week's Material
Other Concepts and Points you are Expected To Master This Week
Miscellaneous Comments and Clarifications

I. Materials Assigned for the Week

Reading:

Lecture 12: What is Knowledge?
Lecture 14: The Reliability Theory of Knowledge
Lecture 2: Abduction (again) pp. 25-35
Readings: Plato, Knowledge is something more than true belief, pp. 204-211

Exercises:

Review Questions: pp. 161, 185
Problems for Further Thought: pp. 161, 185-186

II. The Central Point of this Week’s Material

Most philosophical questions in this course will take the form: "Can we have knowledge of so and so?" (Of the existence of God, of the nature of an objective world, of minds in addition to bodies, of freewill and responsibility.) All of these sorts of questions require a definition of the concept of knowledge. Such a definition must put what knowledge is high enough to be worth acquiring but not so high as to be impossible in principle to attain.

Sober discusses the three main definitions philosophers give:

Knowledge = justified true belief
Knowledge = belief immune from any possibility of error.
Knowledge = belief based on evidence which in the present circumstances is reliable.

III. Other Concepts and Arguments You Are Expected to Master This Week

The difference between knowledge-how, object knowledge and propositional knowledge (or knowledge-that)
The difference between an empirical generalisation and a definition (if you can discriminate, even roughly, between a definition and a theory, so much the better)
The general form which any definition must have - see the "X if and only if Y" business in the box on pg. 155
The difference between necessary and sufficient conditions
The difference between believing something is true and that something being true (a development of the distinction in Week One of "subjective" versus "objective")
The meaning - or more accurately the standard range of meanings - of "justification"
What a counter-example is (and why Gettier’s Smith case, Russell’s clock case and Sober’s lottery case are supposed to count as counter-examples, and counter-examples to what)
What scepticism is - specifically, scepticism about the objective world
That the word "scepticism" has two legitimate spellings: "scepticism" and "skepticism" - both the British and the American spellings are perfectly acceptable
The difference between highly reliable evidence for some proposition and infallible evidence for it
The difference between a priori knowledge and a posteriori knowledge - you may need to look back to the Box on pg. 83 and to pp. 89-90 to get a fuller grasp
The difference between introspection (of a mental idea in my own private mind) and inspection (of a physical thing or property in the public objective world) - but much more of this in Week Seven
What the Reliability Theory of Knowledge is - is it another definition of the necessary and sufficient conditions for knowledge, or a true-blue theory this time?
The difference between impossibility and necessity
The three different types of necessity: logical necessity, nomological necessity and circumstantial necessity - don’t be put off by the long words, you should be able to state the differences without them
The corresponding three different types of impossibility: logically impossible, nomologically impossible and circumstantially impossible - if you can do it for necessity, you can automatically do it for impossibility (why?)
The difference between being relative to so and so and being absolute (not relative to any other thing)
The difference between being relative and being arbitrary - see Box pg. 184 for this common but very dangerous confusion

IV. Miscellaneous comments and clarifications

IV.A Three competing definitions of knowledge

There can be good and bad definitions of anything. "A bachelor is an unmarried male" is a pretty good definition (but are we really happy to say monks are bachelors?). "A bachelor is a man who does pretty much whatever he likes" is a pretty bad definition (it’s mostly true maybe but hardly distinguishes bachelors from others). Likewise, there can be good and bad definitions of knowledge. The difficulty is to find a good one and to show it’s good. Sober discusses two and gives some preliminary assessments of them.

First definition:

For any individual S and any proposition p, S knows that p if and only if:

S believes that p

p is true

S is justified in believing that p

This he calls the "JTB Theory" (pg. 157) - knowledge is Justified True Belief. I’d call it not a "theory" but a "definition". However, "theory" is all right if he means simply the theory that there do exist propositions which satisfy those conditions.

Second definition:

For any individual S and any proposition p, S knows that p if and only if:

S believes that p

p is true

it is logically impossible for S to be mistaken about the truth of p.

Sober does not give this a separate name. Let me offer one: the "TIB Theory" - knowledge is True Infallible Belief. That is, knowledge is true belief for which the justifying evidence is not merely highly reliable but perfectly reliable (pg. 159). Knowledge is true belief "totally immune from the possibility of error" (pg. 160). Clearly, this is simply a proposed "definition" of knowledge as it stands; it becomes a "theory" of knowledge with the added claim that there exist propositions which do meet such high standards.

Third definition:

For any individual S and any proposition p, S knows that p if and only if:

S believes that p

p is true

in the specific circumstances which S occupies, S wouldn’t believe that p unless p were true.
(Alternatively: relative to the circumstances S occupies here and now, it is circumstantially impossible - not logically impossible nor nomologically impossible notice, but circumstantially impossible - that S believe p and p be false)

This third definition occurs in a chapter of its own (Lecture 14). Sober calls it the "Reliability Theory of Knowledge" or the "RTK Theory" (pp. 178-179). Basically it is the definition: knowledge is True Circumstantially Reliable Belief. The emphasis here is as much on the "circumstantially" as on the "reliable". In one set of circumstances, the evidence might not be reliable at all - Sober gives the example where a weather van on top of a barn has been frozen by rust so that the direction it now points to is tells nothing useful about which way the wind is now blowing; a belief based on the position of the weather van in those circumstances has no claim to count as knowledge. In another set of circumstances, of course, the very same evidence might be completely reliable - when the weather van is well greased, and designed properly, and there is enough wind to overcome any friction, then the position it points to is quite sufficient to indicate which way the wind blows; a belief based on the position of the weather vane in those circumstances counts perfectly well as knowledge.

Take notes! Write down the three definitions. Make sure you understand why they are not the same.

As with all definitions (above), it’s no good just giving them. They still have to be tested and assessed as definitions. That is, it still has to be determined whether they are particularly good definitions or at least better ones than their competitors, incisive, to the point, useful to the purposes to which they will be put, neither too strong nor too weak. In epistemology, this assessment procedure takes a typical turn.

Step one: Usually philosophers find reasons for marking down the "knowledge is justified true belief" definition [or theory], on the grounds that there are cases which satisfy the necessary and sufficient conditions specified in the definition, but which we would still be uneasy saying amounted to having knowledge about some matter. This is the role of counter-examples, such as the Gettier cases, the lottery paradox and others. In short, this first definition seems to set things too low.
Step two: Presumably we need a definition of knowledge which raises the necessary and sufficient conditions higher. (And thus makes them into conditions which the previous counter-examples fail to satisfy.) Philosophers typically look to the "knowledge is true belief beyond the possibility of being mistaken about" for those higher standards. [Descartes’ Foundationalism of Weeks Four and Five is such an attempt.] But this has the unfortunate consequence of seeming to make it impossible to attain knowledge even of things we would want to say we had perfectly good knowledge of - that you are reading a white sheet of paper right now, for instance. That is, on this "stronger" definition, scepticism of the objective world looks a strong runner - after all, the senses never provide infallible evidence, it is always logically possible for us to be mistaken in our beliefs about the external world, sense-experience is far from immune to the very possibility of error. Such facts seem to mark down the second definition as raising the conditions for knowledge too high to make any knowledge attainable.
Step three: Something stronger than the JTB Theory and weaker than the JIB Theory is necessary. That is, philosophers must pay much closer attention to that innocent-looking word "justified". The fight is rejoined there. Many times, with many variations, for many subject areas. It becomes clear after many such attempts that what counts as rationally justified belief for so and so may not usefully count as rationally justified belief for such and such. That is, the demands are different in different cases, and what counts as satisfying those demands does too. Hence we need to examine each case on its own merits. Which philosophers then do.
Step four: Sober offers the RTK Theory as a definition of knowledge tailored precisely to take into account this emphasis on the particular circumstances of the situation. This is why Sober spends so much time emphasising the difference between "logical" necessity/impossibility and "nomological" necessity/impossibility and "circumstantial" necessity/impossibility. "Logical" impossibility is what we have already met in the definition of validity in Week Two for instance. "Nomological" impossibility is a long word for saying something easy: that so and so would violate an empirically discovered law (Greek nomos) of nature. "Circumstantial" impossibility is the key one here. Tying my shoe laces is sometimes circumstantially possible for me to do - first thing in the morning, after I’ve had a cup of coffee. But the same thing is sometimes circumstantially impossible - when I’m carry three bags of groceries and it is pouring down rain in the dark. The point is that whether I can or can’t tie my shoelaces is a matter which depends entirely on the specific situation in which I find myself. So too for whether I do or don’t know a certain proposition, says the RTK definition of knowledge. Sometimes my specific situation is such that I can justify and know some matter; sometimes my situation is such that I can’t. It all depends on which situation my present situation is.

It’s plain that the worry behind all these steps has been how to fill out the third condition on "justification". Descartes seized the bull by the horns. What the Reliability Theory of Knowledge is obviously doing is somehow backing off from the super-strong demands that only Descartes’ foundational items (and what can be deduced from them) should be allowed to count as properly known.

What was so striking about the super-strong demand of ITB? That items of knowledge be absolutely immune to doubt. That they be such that it is logically impossible to be mistaken about them no matter what the circumstances. It is this "no matter what the circumstances" which now strikes the RTK as overly phoney and metaphysical. After all, everything else of interest to us is quite firmly situated within our quite specific circumstances. Why not knowledge too then? The sceptic’s doubt is simply way over the top. So what if it is logically possible that there exists an Evil Genius? So what that, if such a being were doing his best to deceive me, my present belief that I am pounding away on an off-white keyboard would - in those weird circumstances - be mistaken? None of this stops me from plodding ahead, or getting RSI, or being tired, or ... anything else, does it? All of which suggests that what I really need to attend to is not some weird hypothetical situation, but my day-to-day real-life situation. Given my present circumstances here and now, am I using enough care, is there some real-life possibility to my eyes and senses getting it wrong? In the present case, for example, the lighting is okay, I’m not all that tired, these are large physical objects I’m talking about, I’m using my normal computer glasses not untested contact lenses, I haven’t just had an operation on my nerves, etc. So I’m not basing my conclusion on any untried equipment; moreover, the situation I am in is well within the parameters in which the equipment I am using is generally reliable; why balk at allowing my conclusion to be properly justified then? - for all that were my situation otherwise in some strange way I might, in that situation, be wrong.

So RTK is making two fundamental changes:

It is making reliability relative rather than absolute. Foundationalism wants the evidence or justification of some proposition p to be reliable no matter what, reliable in all hypothetical situations, reliable in all possible worlds. (Nothing less could guarantee absolute immunity from the very possibility of error.)
It is making reliability relative to my present specific circumstances. Foundationalism doesn’t give a hoot what my present circumstances were in determining whether some proposition was or wasn’t indubitable, foundational, known. In his odd way too, the Gettier and lottery cases are indifferent to my present circumstances too - indeed, the Gettier counter-examples to JTB were not counter-examples to my having knowledge in my present circumstances at all; they only showed that I wouldn’t have knowledge were my circumstances bizarrely different than they actually are (look back at pg. 158).

IV.B The Plato Reading

The reading from Plato is an excerpt from a late dialogue called the Theaetetus. Plato, through his mouthpiece Socrates, reads as if he is simply taking delight in showing that no consistent or useful definition of knowledge will be forthcoming from any of the participants. This is because Plato has a hidden agenda. Knowledge for him is not true belief with a certain kind of backing, as in both the JTB and TIB definitions. Rather, knowledge is direct acquaintance with a special sort of object, a perfect "Form". So any definition which fails to capture that aspect is wildly following the wrong trail according to Plato’s own theory. The Theaetetus is a dialogue of deliberate failure: all the proposed definitions of knowledge do leave out Plato’s theory of Forms; so they are all doomed; so if we want a definition of knowledge that does work - and who doesn’t? - well, we had better bring back attention to the Forms after all. Now, Forms are pretty arcane stuff. But we can still learn a lot about the general endeavour of defining knowledge from this skilful failure.

Several definitions are proposed in turn:

Knowledge = True belief
Knowledge = True belief plus first hand witness
Knowledge = True belief an "account" or "formula" or "description" (all translations of the Greek "true belief plus a logos")
Knowledge = True belief about X specifically plus a distinguishing mark of X specifically
Knowledge = True belief about X specifically plus knowledge of a distinguishing mark [notice the trick here!]
Knowledge = True belief plus an image of the thing in words
Knowledge = True belief plus a catalogue of the parts of a thing together with a specification of how those parts fit together
Knowledge = True belief plus a mark by which that thing can be distinguished from all other things.

Each of these is rejected.

Be reassured that Plato is not engaging in mere word play here. For all his hidden agenda, Plato is sincere that it is a serious, and difficult, business saying exactly what more knowledge requires than true belief. The search for a coherent and useful concept of knowledge is much harder than it looks at first. (As is the search for coherent and useful concepts of most philosophical ideas.) Sober gives you a hands-on exercise with this difficulty, in looking at how delicately we have to examine

Knowledge = true belief plus a justification.

It’s the same philosophical difficulty, of trying to get the requirements not too strong and not too weak, with Plato’s

Knowledge = true belief plus an account.

The Reading is another hands-on exercise. Conducted by a past master of hands-on exercises. Use it as such.

IV.C A problem about “reliability”

Here is a key quote from Sober’s Lecture 14:

“The Reliability Theory of Knowledge says that an individual knows a proposition if the individual is related to the proposition the way a reliable thermometer is related to the temperature it measures. A reliable thermometer wouldn’t say n degrees Fahrenheit unless the temperature were n degrees Fahrenheit. An individual knows that there is a page in front of her precisely when she wouldn’t have believed there is a page in front of her unless there were one.” (Sober, pg. 178)

Think! Does this cheat?

Quite possibly “yes”. There are actually two pieces of business involved in this quote, and it’s not at all as clear as it should be that both can actually be pulled off.

A definition of what a reliable thermometer (or whatever) is. Let us go for a definition such as:
A thermometer is reliable if and only if the following condition holds: When the thermometer says that the temperature is n degrees Fahrenheit, then the temperature must be n degrees Fahrenheit.
That is, reliability in a thermometer is a certain fit between how the environment goes and how the measuring instrument reports it goes.
A determination by some user that some specific thermometer X is indeed a reliable one rather than an unreliable one. After all, it need not be the case that all thermometers are reliable. Or even that most are reliable in most circumstances. (Actually we know they aren’t: mercury thermometers are completely unusable above the boiling temperature of mercury.)

So even given the definition in 1. we always have the second issue, to decide whether some instrument at hand satisfies that definition. Some thermometer may well satisfy the definition, but it is always an separate question whether anybody knows or doesn’t know of some thermometer that it does. (Compare Sober, pg. 177)

The worrisome question is this: For all the palaver about “the reliability of evidence is reliability only relative to specific circumstances”, doesn’t someone who accepts that as a definition of “reliability” (task 1.) still have to decide whether some particular piece of evidence actually does satisfy that definition and so is reliably relative to those circumstances (task 2.)? And how does the RTK propose to accomplish that task?

Exercise!

Is the worry a pseudo one? Is an answer to be found not by looking to the RTK - a definition of knowledge and not a procedure for finding some - but to, say, abductive arguments which we already know are nicely relativised-to-circumstances?

Contents
Previous Week: Argument
Next Week: Finding Some Foundations of Knowledge