It is the task of science to generate knowledge. There are different kinds of knowledge; but in order to be scientific, it must be objective.

In a first approximation, a piece of knowledge may be called objective iff it corresponds to verifiable facts. This definition, however, raises at least two problems:

1. Whether something is a verifiable fact and whether some proposition corresponds to it, is not something established independently of the judgement of human beings. On the contrary, for something to be a verifiable fact means that persons can verify it; and for some piece of knowledge to correspond to such facts means that persons can ascertain such a correspondence.
2. The most important propositions in scientific theories are generic propositions or laws. Such propositions do not directly match any particular facts. They may, in fact, be related to particular facts by the intermediation of several degrees of abstraction or of other operations of the mind. In short, it must be spelt out what ‘correspond’ means.

From consideration #1 it follows that objectivity essentially means intersubjectivity: a piece of knowledge is objective iff – given certain conditions – it is conceded by everyone. Since scientists grant the truth of something only on the basis of some method from which it results, intersubjectivity becomes a property of a method. More specifically, it is a particular aspect of the reliability of a method which is sometimes called inter-rater reliability. This means that if a certain method is applied to some set of data by different scientists (“raters”), the outcome is the same.

Although it may at first sound counterintuitive, the notion of intersubjectivity must be kept apart from the notion of understanding. Given a proposition p, it is possible for two persons a and b to find p true independently of each other on the only basis that each of them accepts and applies the same set of criteria. This is what is meant by operationalizing objectivity in terms of intersubjectivity. It is, however, equally possible for a and b to reach agreement on p on the basis of mutual understanding. Understanding is a direct relation between a and b which typically involves empathy. Empathy is something highly subjective. Therefore understanding is to be kept out of the notion of intersubjectivity. Quite on the contrary, it is the hallmark of the objectivity of a certain statement that anybody can control its truth without applying empathy and good will to the task.

From consideration #2, it follows that there must be criteria that relate pieces of knowledge – scientific propositions – to verifiable facts. The essential thing in intersubjectivity is that if anybody applies the set of established criteria, he can verify the piece of knowledge in question. Intersubjectivity is, therefore, not simply the introduction of democracy into science, in the sense that ‘something is true iff an overwhelming majority finds it to be true’. What is crucial, instead, is that anybody who accepts the criteria will necessarily agree to the proposition in question.

Science therefore needs two kinds of criteria in the evaluation of pieces of knowledge:

1. Criteria by which one may verify whether something is a fact.
2. Criteria by which one can relate a proposition to a (verified) fact.

A closer characterization of criteria of the first kind presupposes a definition of the terms ‘fact’ and ‘verify’.

Criteria of the second kind serve to operationalize scientific notions.