Epistemic Representation: The Basics

Now that we have reached a rough taxonomy of modalities, let us talk about representation. I think it will be enough to have this rough picture in mind, which, I reckon, should be uncontroversial aspects of representation:

Inferentialism:

An epistemic representation (such as a scientific model, a map, a realistic drawing) can be used to make inferences about target systems of a relevant kind in concrete situations.

Misrepresentation:

It is possible, for a representation, to represent its target inaccurately.

I take inferentialism to be a minimalist theory of representation. This does not tell against theories that attempt to give more substance to representation, but only defines a common ground. As for the possibility of misrepresentation, it has been challenged notably on the ground that representation would be a “success term” or a thick term (Poznic 2018; Chakravartty 2010, pp. 209–10), but I’m utterly unconvinced that this is the case. “Cruel” is a thick term, and one can be more or less cruel, or truly cruel, but would we say that the Ptolemaic model is less a representation of the solar system than the Copernician model, that the latter is truly a representation of it? Or that “bad representation” is as much of an oxymoron as “good cruelty”?

In any case, accepting this minimalist picture is enough to introduce a distinction between conditions of relevance, specifying what the representation is about, and conditions of accuracy, specifying what it says, or what its target must be like for the representation to be accurate. The two must be distinct if misrepresentation is possible. We can flesh out the distinction as follows:

Conditions of Relevance:

A vehicle of representation conveys implicit norms saying in which kind of situations it is appropriate to use it, or what system or kind of system it is apt to represent. For example, a city map of London is appropriate for navigating London, not Lisbon. Note: this need not be a reference to a single object. Most (if not all) epistemic representations are applicable in many potential situations, sometimes involving different objects. For example, the model of the ideal pendulum can be used represent any pendulum in any situation where there is one.

Conditions of accuracy:

A vehicle of representation is associated with implicit or explicit norms of use that define its scope (what inferences are about) and how to coordinate the representation and a physical situation. For example, a city map comes with a legend for interpreting its symbols in terms of streets, intersections, metro stations, etc., from which we know that the map is not about trees or the colour of buildings and it has a wind rose indicating how to orient the map. Ideally at least, yogether with its norms of use, the content of the vehicle of representation determines the inferences that are allowed in any given context of use (including conditional inferences when there are parameters, as in most scientific models). The representation is accurate if these inferences lead to true conclusions about the target situations. In this sense, the content and norms of use determine what the situation must be like for the model to be accurate, i.e. its conditions of accuracy.

If the vehicle is used to represent an object that does not fullfil conditions of relevance, we have a case of misuse, but that does not mean that the representation is inaccurate. On the other hand, if the vehicle is properly used, but the conclusions of the inferences are false, then we have a case of misrepresentation.

I think that this rough picture is capable of handling most cases of representation. I assume that the target of representation can be actual, hypothetical or fictional. However, we do not need to believe that all models are fictions in order to account for idealisations, as the fictional trend would have it (see for example Frigg 2010). A metro map that distorts distances is idealised, but its norms of use tell us that we should not consider that distances are representational, and I think the same goes for idealisations in science in general. The metro map represents the actual metro, not a fictional network.

It will be helpful in the following to compare epistemic representation and linguistic representation: applying a model to a situation is analogue to saying “This K is P” where the kind K corresponds to conditions of relevance and P to the conditions of accuracy of the model (including conditional conditions of the type “if I then O”: this will be important later).

Several modalities were involved in this presentation, and we might wonder what kind of modality they were. This is what will be examined in the following posts.