Epistemic Value and the Swamping Problem

In Plato's Meno, Socrates asks why knowledge should be prized above true opinion. Someone who knows the road to Larissa and someone who merely has a correct belief about it will both get there. If both arrive, what does knowledge add?

Socrates' answer is that true opinions are like the statues of Daedalus, which were said to run away unless tied down. A true belief without knowledge may be dislodged by the next clever argument. Knowledge is true belief tied down by an account of the reason why, and is therefore more stable.

The question lay mostly dormant for two thousand years. It returned at the end of the twentieth century with a sharper edge, as an objection to a particular theory of knowledge. If reliabilism is right, it seemed, knowledge could not be worth any more than true belief at all.

Kinds of value

The distinction matters because the question "why is knowledge more valuable?" can be heard in two ways. Is it more useful? Or is it better in itself? The Meno reading is about usefulness: knowledge gets you to Larissa more dependably. The contemporary puzzle is about the other kind.

Veritism

Veritism is the natural companion of reliabilism. If a reliable process is good because it tends to produce truths, then its epistemic value lies in the truths it produces. Justification is valuable because it tracks truth. Rationality is valuable because rational methods tend to get things right. The one thing valued for its own sake is truth itself.

This is elegant. It explains why we care about evidence, careful reasoning and good instruments: because they lead to truth. It explains why we criticise wishful thinking: because it does not. And it gives epistemology a single aim in the way consequentialist ethics gives morality one.

The swamping problem

Zagzebski's version, from 2003, uses a cup of coffee.

A good espresso machine is valuable because it tends to make good espresso. It is reliable. But suppose you have a cup of excellent espresso in front of you. Does it taste better, or is it more valuable, for having come from a reliable machine rather than a temperamental one that happened to produce a good cup? It does not. Once the espresso exists, the reliability of its source adds nothing to it. The value of the machine is exhausted by the value of the cups it makes, and a particular good cup is just as good whatever made it.

The argument does not deny that reliable processes are valuable. They are, since they produce more truths. The claim is narrower: the reliability of the source cannot make a given true belief more valuable than it would otherwise be. And since we think knowledge is more valuable than mere true belief, the reliabilist has a problem.

The argument generalises. It threatens any theory on which knowledge is true belief plus a property whose value is instrumental to truth. That is why it is sometimes put as a problem for veritism rather than for reliabilism in particular.

Responses

The conditional probability solution

Alvin Goldman and Erik Olsson replied in 2009 with what they called the conditional probability solution. A true belief produced by a reliable process makes it more probable that future beliefs of the same kind will be true. If you know the road to Larissa, you are likely to know the road next time too, and the time after. The reliably formed belief is a sign of a reliable faculty, and that faculty will keep producing truths.

The coffee analogy does not capture this, because a good cup of espresso says nothing about the next cup. But a reliably formed true belief does say something about the future. It raises the probability that your next belief will also be true.

Critics reply that this explains why it is useful to have reliable faculties, and why it is good to have a history of reliably formed beliefs, but not why a given item of knowledge is better than the corresponding true belief considered on its own. It is also unclear that the solution handles cases in which no future beliefs of the same type will be formed.

Credit and achievement

A more influential response comes from virtue epistemology. Ernest Sosa, John Greco and Zagzebski herself argued that knowledge is a kind of achievement: true belief that is creditable to the believer's intellectual virtue. Achievements have final value. We admire the archer who hits the target through skill more than the one who hits it by a lucky gust of wind, even if the arrow ends up in the same place.

If knowledge is an achievement, its value does not come entirely from the truth of the belief. It comes from the fact that the truth was reached through the believer's competence. That is a relational property of the belief, and relational properties can carry final value. The swamping argument's third premise fails, because the virtue that produces knowledge is not valuable only as a means to truth.

The credit view faces an objection from testimony. When you ask a stranger for directions and believe them, you seem to know where the station is, but little of the credit for getting it right belongs to you. It belongs to the stranger. Jennifer Lackey pressed this objection, and virtue epistemologists have responded by weakening the credit requirement to allow shared or partial credit.

Abandoning the premise

A third response, associated with Kvanvig, is to accept the swamping argument's conclusion and relocate the value. Kvanvig argued that knowledge is not, in fact, distinctively valuable over true belief plus justification, and that the thing we really prize is understanding: grasping how various truths hang together and why things are as they are. Understanding cannot be reduced to knowledge of a set of propositions, and it has a value that knowledge lacks.

This is an unusual move, since it treats the value problem as showing that epistemologists have focused on the wrong target. But it has become influential, and the study of understanding is now a significant subfield.

The Meno question again

Socrates' answer was about stability. That answer was set aside in the recent literature, but it has made a comeback. Timothy Williamson argued that knowledge is more robust than true belief: someone who knows is less likely to lose their belief in the face of misleading evidence than someone who merely believes truly. This is a practical advantage, and it does not face the swamping problem, since robustness is a feature the knower has that the lucky believer lacks, not a feature of the source that affects only the belief's truth.

This returns the discussion to the instrumental reading, and it might be that the Meno question has two answers. Knowledge is more useful because it is more stable. Whether it is also finally more valuable depends on whether one accepts the credit view.

Next

Probability as a constraint on rational credence, the Dutch book argument for that constraint, and the rules for updating credences in response to evidence.