Sebastian Schmoranzer

Remarks on scepticism: Principles of exclusion and the relevance of sceptical doubts

I assume that all of you are familiar with Moore’s claim that he knows that there is a hand in front of him. I consider this to be a paradigmatic knowledge claim about material objects. In accordance with Moore I will be using the term "material object" in this paper in a metaphysically realistic sense: An object only is a material object in this sense if it exists in space-time and if it does not have to be perceived at a certain moment in order to exist at that moment.

You all know the dream argument which the sceptic uses to contest Moore’s claim and — since it is paradigmatic — to contest our knowledge about material objects in general:

Now, has the sceptic hereby refuted Moore’s knowledge claim? Since the sceptical argument is valid, we can only examine the plausibility of its premises. I would like to concentrate on the first premise and ask the following question: "Must Moore always be able to exclude to be dreaming a Cartesian dream in order to know that there is a hand in front of him?"

It is a shared opinion in epistemology that the sceptic can only defend the first premise of his argument if he assumes a principle of exclusion to be implied by the concept of knowledge — i.e., if he assumes that knowledge requires the exclusion of error possibilities.

Therefore, the first part of this paper deals with the question which principle of exclusion could be acceptable for the sceptic (II.a) and investigates whether such a principle presupposes a concept of knowledge which is too strong (II.b). The second part deals with whether the sceptic can hold on to the first premise of his argument in spite of contextualist objections against a sceptical principle of exclusion (III).

Three principles of exclusion will be discussed in the following. The first two derive from reflections of Barry Stroud. The third is my own proposal.

This principle is too strong. It follows from this principle that S must know the truth of all propositions implied by p in order to know that p. To illustrate this, let us assume that p implies q. If non-q is true, p is false. If p is false, S does not know that p. So, non-q is an error possibility which is incompatible with S’s knowing that p. Therefore, S must be able to exclude that non-q. If "being able to exclude" means something like "knowing that it is not the case that", S must know that it is not the case that non-q. Hence, S must know that q in order to know that p.

Now, the demand that one must know the truth of all propositions implied by p in order to know that p is certainly too strong. It may be that a proposition p implies infinitely many other propositions or at least a lot of other propositions, so that it is rather counter-intuitive to request that S must know the truth of all those propositions in order to know that p. According to this request S would even have to know the truth of those propositions implied by p no one has ever thought about.

Furthermore, this demand collides with our intuition that one can gain some insights into a certain domain first and extend one’s knowledge in this domain afterwards by examining what follows from those facts one already knows.

There is another principle Stroud suggests:

This principle is too weak. Let us consider the following example: Independently of one another Jack and Peter see their own hand in front of them. Jack is one year younger than Peter and has not taken any philosophy classes at school yet. He has no idea of Cartesian dreams and the like, whereas Peter has just discussed Descartes in his philosophy class. Peter knows Descartes’ dreaming argument. According to the above principle Peter has to be able to exclude the Cartesian dream in order to know that there is a hand in front of him, whereas Jack does not have to exclude this error possibility. So, it follows from the above principle that for those people who - like Jack - have no idea of sceptical hypotheses it is easy to have knowledge about material objects. Yet, the sceptic holds that no one — not even the ignorant Jack — has such knowledge.

I therefore propose the following principle which I call the sceptical principle of exclusion:

Before I discuss whether this principle is too strong, I must admit that the term "we" needs further clarification. Which group does it refer to? How many and what kind of people must know that some error possibility is incompatible with a certain knowledge claim to include this error possibility among those a person must be able to exclude? I have not found a satisfying answer to these questions yet.

None the less, the first premise of the sceptical argument is justified if the following holds:

1. The sceptical principle of exclusion is not too strong and does not get into conflict with our concept of knowledge.
2. If the sceptical principle is acceptable, it is sufficient that Descartes, You and I and many other people know that the dreaming hypothesis is incompatible with Moore’s knowledge claim to make this an error possibility Moore must be able to exclude.

In order to answer this question, I would like to discuss the following questions: A) Does the principle imply that knowledge presupposes certainty? B) Does the principle have the consequence that first order knowledge implies second order knowledge? C) Does our every day practice of knowledge attribution speak in favour of the sceptical principle?

1. Knowledge and certainty

The sceptical principle does not imply that knowledge presupposes certainty. I am not talking about a subjective feeling of certainty but about certainty in the following sense:

A proposition p is certain for a person S if and only if i) S believes that p, ii) p is justified by S and iii) S’s justification of p implies the truth of p.

Even if one is able to exclude all error possibilities known to us, the justification for this can be fallible. By this, I mean that p can be false although one has justified p and excluded all error possibilities we know. Let us consider the following example: A biologist points to an ant on an anthill and claims to know that it is a female worker. The ant actually is a female worker. Let us assume that the error possibilities we know are i) that it is a male ant without wings and ii) that the sun shines in such a way on the anthill that one cannot see whether the ant has wings or not. The biologist excludes the second error possibility by creating shadow with his cap. He excludes the first error possibility by the following reasoning: The ratio of male ants to female workers is 1:100. The ratio of male ants without wings to male ants with wings is 1:100. The probability of looking at a male ant without wings is 1:10.000. Thus, the biologist has very good reasons to exclude that the ant is a male ant without wings. In our example the biologist has excluded all known error possibilities, satisfies the sceptical principle of exclusion and knows that the ant is a female worker.

However, despite the justification he gives it is possible that the ant is not a female worker but a male ant without wings. The probability of this is admittedly very low (1:10.000). But the biologist’s justification does not imply that it is a female worker. Therefore, the biologist has knowledge but not certainty.

2. First order and second order knowledge

It does not follow from the sceptical principle of exclusion that first order knowledge implies second order knowledge.

First order knowledge is, for example, knowledge concerning ants in anthills. Second order knowledge is, for example, knowledge about knowledge concerning ants in anthills. If our biologist wants to know that the ant is a female worker he must be able to exclude all error possibilities we know regarding the ant’s actually being a female worker. Yet, in order to know that he knows that it is a female worker more is demanded. The biologist must furthermore know that he is able to exclude all error possibilities we know regarding the identification of ants in anthills. There is a difference between actually being able to exclude all known error possibilities regarding a certain knowledge claim and knowing that one is able to do so. In the first case less error possibilities have to be excluded than in the second case. This is the reason why one can have first order knowledge without having second order knowledge even if one satisfies the sceptical principle of exclusion regarding first order knowledge.

To show this, let us return to the example of the biologist. We assume that the error possibilities we know regarding the identification of the ant’s sex are the two I mentioned above: the sun shines in a certain way on the anthill and the ant is a male without wings. The biologist can exclude both these possibilities and the ant actually is a female worker. Now, let us further assume that the biologist has not read all the scientific articles about how to avoid errors of identification concerning female ants, that he would admit this fact and is known to be lazy as far as the theoretical preparations of his scientific work are concerned. The scientific articles the biologist has not taken into account, however, do not mention any further error possibilities concerning the identification of female ants.

Under such conditions the biologist knows that the ant in front of him is a female worker. The fact that he has not read all the scientific articles has no consequence for his first order knowledge.

But what about his second order knowledge? In order to know that he knows that it is a female worker he must know that he has excluded all known error possibilities. But then he must know that he has taken notice of all known error possibilities. It would be strange to say that someone knows to have excluded all known error possibilities without knowing what the error possibilities are.

When does someone know that he has taken notice of all known error possibilities? He must at least i) know all of them and ii) be justified in his claim to know them all. In the case of the biologist the first condition is satisfied. He actually knows all known error possibilities. Yet, the second condition is not satisfied. In order to be justified in his claim to know all error possibilities the biologist must be able to exclude that he has failed to take several important articles into account since they could have mentioned further important error possibilities.

But this is something the biologist cannot because i) he has actually failed to take the articles into account, ii) he would admit this and iii) he has no good arguments to deny his failure since he is known to be reluctant as far as research work in libraries is concerned.

We may conclude that the biologist does not know that he has taken notice of all known error possibilities. Therefore he does not know that he can exclude all known error possibilities. This in turn implies that he does not know that he knows that the ant is a female worker.

The "important-articles" error possibility is one which puts into doubt the biologist’s second order knowledge. But it does not infect the biologist’s first order knowledge. The biologist knows without knowing that he knows.

3. Knowledge and Context

There might be another reason why the sceptical principle is too strong. Is it really necessary to be able to exclude all error possibilities we know of or is it sufficient that one can exclude all relevant error possibilities?

Examining our practice of knowledge attributions in ordinary, life Austin remarks in "Other Minds" that not every doubt has to be excluded but only those which are themselves justified. And only those doubts are justified which are relevant relative to the present intents and purposes in a given context.

According to Austin, only context-relevant doubts have to be excluded. With this idea he is one of the fathers of epistemological contextualism or of the so-called "Relevant-Alternatives"-Approach. Supporters of this approach defend a contextualist principle of exclusion:

I will give two examples which speak in favour of the contextualist principle of exclusion and also convey the impression that in a lot of contexts the sceptical doubt is not relevant at all.

Example 1: At the market

I am at the market. The merchant tells me that he has fresh cucumbers. I can confront him with the doubt whether he is sure that his cucumbers are really fresh, or whether he is sure that they are not courgettes. However, if I put his claim into doubt by asking whether he is sure that he is not dreaming a Cartesian dream I make myself ridiculous. The dreaming-hypothesis is an irrelevant error possibility and does not have to be excluded in this situation.

Example 2: At a scientific congress

Let us suppose that at a congress about Roman History two historians have opposite views concerning the question whether Caesar was in Britain or not. They can put their respective positions into doubt by asking whether the other’s historical sources are reliable. Telling the other that he may be dreaming a Cartesian dream only discredits one’s own scientific reputation but not the opponent’s knowledge claim.

The examples seem to support the following argument by which the contextualist can refute the sceptic’s premise that Moore must always be able to exclude the dreaming-hypothesis in order to know that there is a hand in front of him:

The sceptic could attack the first premise and try to defend his principle of exclusion against the contextualist principle. I do not want to discuss this very complex line of reasoning here.

Instead, I would like to concentrate on the second premise. Is there really a context C in which the sceptical doubt is irrelevant as far as Moore’s knowledge claim is concerned?

As we have seen, the sceptical doubt is irrelevant in contexts of ordinary life as well as in scientific contexts. When confronted with such examples the sceptic should give an explanation which satisfies the following demands:

Firstly, the explanation must explain why the sceptical doubt is of no relevance in a lot of contexts.

Secondly, the sceptic must present an argument which shows that the sceptical doubt is relevant as soon as someone makes the same knowledge claim as Moore.

In order to give the requested explanation the sceptic can first of all refer to the idea of different scopes of the knowledge operator. The sceptical doubt is not relevant in many contexts since the scope of the knowledge operator is restricted in those contexts. At a congress about Roman History a historian claims to know that Caesar was in Britain. According to the sceptic the position of the historian in the given context has to be analysed as follows — taking implicit assumptions of the historian into account. ("c" denotes Caesar, "B" stands for "... was in Britain", "M" stands for "... is a material object" and "K" symbolises the knowledge operator.) :

Mc & K (Bc)

The sceptical doubt is of no relevance because the question whether Caesar is a material object or not is not the issue. The historian assumes that Caesar is a material object. But he just presupposes it. The metaphysical aspect of his position (that Caesar is a material object) is not the issue since metaphysical questions play no role in such scientific contexts as a congress about Roman History. What the opponents are discussing is whether Caesar was in Britain or not. Respectively, only this part of the historian’s position lies within the scope of the knowledge operator.

However, when, in a philosophical paper, Moore makes the claim that he knows that there is a hand in front of him after having entered into an intricate analysis of what kind of objects hands are, he defends the following position — I am again taking implicit assumptions into account. ("H" stands for "... is a hand", "M" stands for "... is a material object", "E" is the existential quantifier and "K" symbolises the knowledge operator. The universe of discourse is the set of all objects.):

K (Ex (Mx & Hx)

By making his assertion in an explicitly philosophical context in which one discusses the philosophical theories of idealism, realism and scepticism, the assumption that the hand he perceives is a material object lies within the scope of the knowledge operator.

By expanding the scope of the operator, philosophical questions concerning our knowledge of the ontological status of an object come into focus. In such a situation the sceptical doubt becomes relevant — according to the sceptic. Thus, as soon as someone holds Moore’s above formalised position the sceptical doubt becomes relevant.

Why is this so? At this point the sceptic refers to the idea of a minimal standard of justification. For a person S to know that p in any context, S must satisfy a minimal standard of justification. A Martian S might justify his true belief that p with the following justification: "The God of the Martians is great and p. Therefore p". Maybe this kind of justification is sufficient in the community of Martians. In this case S has "Martian-knowledge". But S surely does not have the kind of propositional attitude we call knowledge. S has "Martian-knowledge" but not "Human-knowledge" since S does not even satisfy a minimal standard requested to have "Human-knowledge".

According to this idea, for Moore’s assertion to be true in any context he has to satisfy a minimal standard of justification. And this minimal standard requires the exclusion of the dreaming-hypothesis, says the sceptic. Is this reasoning plausible? I think it is.

By claiming to know that the perceived object is a material object in the metaphysically realistic sense indicated at the beginning of this paper, Moore defends a philosophical position and claims to have the competence to distinguish material objects from perfect illusions of material objects on the basis of the perceptual information provided by the senses. So, Moore claims that his perception is a reliable instrument as far as the question of the ontological status of the perceived hand is concerned. This claim is in need of justification. The more you claim the more you have to justify. Moore must at least be able to exclude that his senses are completely unreliable relative to the question concerning the ontological status of his hand. Therefore, he must be able to exclude that he is dreaming a Cartesian dream since this is a case in which his senses are completely unreliable relative to the question of the ontological status of Moore’s hand.

Contextualists will still not be satisfied with this explanation. Are there not context in which someone explicitly claims to know that something is a material objects and in which the sceptical doubt is none the less irrelevant? Imagine two historians discussing whether Caesar was a real person or just a fiction of historiography. One historian claims to know that Caesar was not a ‘mere’ fiction but a material object. In this context i) the sceptical doubt is irrelevant and ii) the scope of the knowledge operator includes the claim that Caesar was a material object.

The sceptical reply to this example goes like this: Either the historian uses the term "material object" in a metaphysically innocent sense or he uses it in a metaphysically realistic sense.

By using the term in the first sense, the historian just wants to make plain that Caesar is the same kind of object as the chairs and tables around him and that he is not the same kind of object as Pegasus, Nessie or Santa Claus are. Using the term this way would not commit the historian to a philosophical position. In this case the sceptical doubt is irrelevant. But then the historian is not making the same kind of knowledge claim Moore makes. Moore uses the term "material object" in the metaphysically realistic sense I have been using in this paper. An object is a material object in this sense only if it exists in space-time and does not have to be perceived in order to exist. To see the difference between Moore’s claim and the historian’s position let us formalise both of them. ("M*" stands for "... is a material object" used in a metaphysically innocent way, "M" stands for "... is a material object" in a metaphysically realistic sense):

Moore’s claim: K (Ex (Mx & Hx))

The historian’s claim: K (M*c)

But maybe the historian also uses the term "material object" in a metaphysically realistic sense. Then he is making the same kind of knowledge claim Moore makes. But then the historian does not only defend a historical position but also a philosophical position. In this case the sceptical doubt becomes relevant for the same reasons as in Moore’s case.

Let us sum up the sceptical argument: In a lot of contexts the sceptical doubt is irrelevant either because the claim that something is a material object does not lie in the scope of the knowledge operator or because the term "material object" is used in a metaphysically innocent way. But as soon as one extends the scope of the knowledge operator and uses the term "material object" in a metaphysically realistic sense the sceptical doubt becomes relevant and is part of the minimal standard of justification one has to fulfil in order to have such knowledge.

In other words: By making the kind of knowledge claim Moore makes, one defends a philosophical position and creates a philosophical context in which the sceptical doubt becomes relevant. Therefore, there is no context in which Moore can make his knowledge claim and in which the sceptical doubt is irrelevant.

Relative to Moore’s assertion the sceptic has refuted the contextualist argument without refuting the contextualist principle of exclusion. But has he not thereby won a Pyrrhic victory? Interestingly enough, if someone holds positions which would be analysed in the following way the sceptical doubt is irrelevant:

Ex (Mx & K (Hx))

Mc & K (Bc)

K (M*c)

Only if someone holds a position which can be analysed as follows does the sceptical doubt become relevant:

K (Ex (Mx & Hx))

We can see that the sceptical doubt puts less into question than we would have expected before. Scientific knowledge and the knowledge of ordinary life remain untouched by scepticism as long as the scope of the knowledge operator is restricted or the term "material object" is used in a metaphysically innocent sense.

None the less, the sceptical argument has disturbing consequences. If the sceptic is right we can never explicitly claim to know that the object we perceive is a material object in a metaphysically realistic sense. As soon as we extend the scope of the knowledge operator and use a metaphysically loaded term we create a philosophical context in which the sceptical doubt becomes relevant.

The above reflections have shown two facts. Firstly, the sceptic can give an argument and rely on a principle of exclusion without presupposing that knowledge implies certainty or that first order knowledge implies second order knowledge. It remains problematic, however, whether the sceptical principle of exclusion is still too strong when one considers our practice of knowledge attributions which, at first sight, favours a contextualist principle of exclusion.

But, secondly, even if one prefers the contextualist principle the sceptic is able to defend his argument against Moore. He can insist on the relevance of his doubt relative to Moore’s knowledge claim. Interestingly enough, this sceptical defence has the consequence that less is put into doubt than we would have expected. Nevertheless, we are still confronted with the following problem: Do I know that the hand in front of me is a material object in a metaphysically realistic sense?

References:

Annis, David B., A Contextualist Theory of Epistemic Justification, in: American Philosophical Quarterly Volume 15 Number 3 (1978): 213-219.

Austin, J. L., Other Minds, in: Austin, J. L., Philosophical Papers, Oxford 1962: 44-84.

DeRose, Keith, Contextualism: An Explanation and Defense, in: Greco, John/ Sosa, Ernest (eds.), The Blackwell Guide to Epistemology, Oxford 1999: 187-205.

Grundmann, Thomas/Stüber, Karsten, Einleitung, in: Grundmann, Thomas/Stüber, Karsten

(eds.), Philosophie der Skepsis, Paderborn/München/Wien/Zürich 1996: 9-57.

Heil, John, Doubts about Scepticism, in: Philosophical Studies 51 (1987): 1-17.

Moore, G. E., Proof of an External World, in: Moore, G. E.: Philosophical Papers, London

(11959) 31970: 127-150.

Stroud, Barry, The Significance of Philosophical Scepticism, Oxford (11984) 21985.

Willaschek, Markus, Wissen, Zweifel, Kontext - Eine kontextualistische Zurückweisung des Skeptizismus, in: Zeitschrift für Philosophische Forschung, 54 (2000): 151-172.

since december 2000