Let us take stock. In the posts so far (linked below in order),
- I presented modal empiricism: the view that our best scientific theories are empirically adequate for all possible situations to which they would apply.
- I detailed the conception of empirical adequacy on which the position rests. It is not cast in terms of a model of the universe, as usually, but in terms of situations to which different models apply, and, I think this conception is more connected to scientific practice than the usual ones.
- This conception of empirical adequacy does not commit us to modal empiricism, but I explained how modal empiricism is able to answer the no-miracle argument, while retaining the advantages of empiricism when it comes to theory change.
- Finally, I explained why, according to me, scientific realism is misguided: it rests on (meta-)abduction for its justification, but abduction, however central it is in scientific practice, is not a principle of justification, but a strategic device to select good hypotheses: hypotheses that we should test first, that is. It doesn't exempt us from further empirical tests if we want to justify these hypotheses. But scientific realism cannot itself be tested empirically.
What makes modal empiricism a version of empiricism, not of scientific realism, is that in contrast with realism, the modalities to which it is committed are arrived at by induction on possible situations, not by abduction. Relations of necessity are no explanations to regularities, but regularities extended to the possible.
Why should an empiricit be a modal empiricist?
But let me first disclose an important point to settle the discussion. When I say that modalities are arrived at by induction, I do not mean that induction alone is enough to justify a commitment to the existence of necessity in nature. First, one should accept that there is something like constraints of necessity on observable phenomena before to induce on possible situations. Then we get all the neat features of modal empiricism: its ability to answer the no-miracle argument, and so on. But why should we make the first move?
I don't think that this move can be justified on purely empirical ground, but any way, no empiricist position can be justified on purely empirical ground so that problem is not specific to the position I defend. Van Fraassen is well aware of that: he defines constructive empiricism as a position about the aim of science, and gives arguments based on the rationality of scientific practice. He views empiricism as a stance.
So in order to defend my position against other versions of empiricism, I will not attempt to prove it on the basis of experience, or whatever, but I will rather apply the following strategy: first, I will explain why the traditional arguments against the possibility of modal knowledge can be resisted. This includes the argument that possible worlds are disconnected from the actual world, and that relations of necessity are not given in experience, but also the idea that modal relations would be underdetermined by actual experience, and would be part of an inflationary metaphysics. All my responses will be based on the assumption that there are such things as possibilities in the world to start with, and this requires an independent justification, so in a second step (in the next post), I will provide positive arguments to that effect, in line with van Fraassen's justification of empiricism on the basis of scientific rationality.
The conclusion will be that if one wishes to make sense of scientific practice as a rational activity, one should better endorse natural modalities, and without any argument to the effect that relations of necessity between observable phenomena couldn't be known from experience, there is no reason not to do so. Let us start with the first step: if there is necessity in the world, how could we know about which relations of necessity hold?
The possibility of modal knowledge
As I explained when examining the notion of empirical adequacy, I think one should better think of experience in terms of situations rather than in terms of the universe as a whole, and as a corollary, one should better think of necessity in terms of possible situations in the actual world rather than in terms of possible worlds. Possible situations are alternative ways actual situations could be, or alternative ways they could evolve (or have evolved). We conceive of possible types of situations on the basis of the objective aspects of actual situations (types of physical systems, the way they can be measured, their initial state) that allow us to apply models of our theories to the world. A theoretical vocabulary somehow constitutes a map of conceivable situations (what could be the measurement results in different types of experiments), and physically possible situations are a subset of them. Necessity in the world is just this: which situations, so conceived, so described in terms of conditions of application and measurement results, are possible or not.
If this is so, then the arguments above lose some of their force: possible situations are not necessarily disconnected from actual ones. An actual situation might have possible effects, for example. And if we take into account not only our observations, but also our interventions (which was an important motivation for understanding empirical adequacy in terms of situations), then we surely possess a "modal microscope" to explore possible situations: all we have to do is implement them by intervention. Just create the right conditions of application. Do you want to know what would happen in a possible situation where an electron and a carbon atom collide? Well, make the test!
However, the most important aspect that allows the modal empiricist to answer all these arguments is the fact that, according to her, modal knowledge is arrived at by induction on possible situations, and not by a mysterious cognitive access to alternative possibilities. Indeed, contrast the following two statements:
- Relations of necessity (what is true of all possible situations) are not given in experience, so we cannot know about them.
- Universal generalisations (what is true of all actual situations) are not given in experience, so we cannot know about them.
- Eventually, all situations we experience are actual. We never experience a merely possible situation. So we know only about actual situations.
- Eventually, all situations we experience are observed. We never experience a merely observable situation. So we know only about observed situations.
All these statements are reasonable, but what they threaten is not modal knowledge per se, but induction in general. The lesson, I think, is that modal empiricism fares no worse than other versions of empiricism, so long as they are willing to extend empirical adequacy beyond what is actually observed. Once we accept that modal statements have truth values, there is no principled reason that would allow one to doubt modal knowledge that wouldn't also threaten any form of induction. Now most people would find reasonable to believe that our theories will continue to be empirically adequate in the future, and similarly, the modal empiricist finds reasonable to believe that our theories are empirically adequate for all possible situations. This is merely the principle of induction: an extension of a statement concerning particular situations to a larger domain of situations, and the fact that all situations in this domain are not actually observed is irrelevant.
If this is so, then the only route for an empiricist to deny modal knowedge would be to doubt that there is necessity in nature from the start, but arguments to the effect that we have no "modal microscope" do not go through.
The underdetermination of modalities
I will consider that two theories are underdetermined if both are empirically adequate, but cannot be both correct at the same time. What is understood by empirically adequate and correct can vary. For example, we can say that empirical adequacy in the future is underdetermined by empirical adequacy so far, because two theories can be empirically adequate so far while it could be impossible that both will be in the future (if they make contradictory claims about future events). The argument, then, is that insofar as underdetermination is ubiquitous, empirical adequacy (of a certain kind) is not enough to infer correctness (of a certain kind), in pain of contradiction, and assuming that empirical adequacy is our only means of justification, then correctness cannot be justified. We don't know how to choose between competing theories.
Underdetermination can in principle be used against any form of induction, depending on what kind of empirical adequacy we consider warranted. Underdetermination in general is too strict a criteria to be used as a mark of "inflationary metaphysics". But still, there's something about inflationary metaphysics that has to do with underdetermination: some metaphysical claims, according to the empiricist, are inflationary because they cannot be verified or disconfirmed even in principle, in contrast with a different metaphysical claim. One would be better off remaining agnostic about the issue at stake. But in the case of future empirical adequacy, well, we will eventually know: if we are confronted to two underdetermined theories, we can just remain agnostic so long as we haven't performed the right kind of test that would allow us to discriminate between our two theories. It's only a matter of time when two alternative, competing theories are considered. Perhaps there are unconceived alternative too, but perhaps we shouldn't worry too much about this "hypothetical" underdetermination until we are confronted to a real case--unless we want to be sceptical about induction in general.
Inflationary metaphysics is different: it happens when we postulate things even if alternatives exist, and even if no possible experience could help us decide between alternatives. Then we have to resort to non-empirical criteria to decide (exactly those that make a "good" explanation in abductive reasoning), but the empiricist will be sceptical about the idea that these criteria would be markers of truth, rather than mere pragmatic devices. Rightfully I think.
Underdetermination is usually cast in the litterature not in terms of what was observed so far, but in terms of all possible observations: two theories are underdetermined if no possible experience could help us distinguish the two. They have exactly the same empirical consequences. I think this focus on all possible observations is legitimate: what is important is in principle discriminations between different theories. But obviously, if we focus on possible observations, then modal empiricism is not threatened by underdetermination: if we have two modally incompatible theories, i.e. if we know that both cannot be modally adequate at the same time, then this means that there exists at least one possible situation where the two theories would make contradictory predictions, and then, this means that possible observations could help us decide between the two. Actually, all we have to do is implement this type of situation for which they make contradictory predictions, and we will know which theory is modally adequate!
These comments are not unrelated to the fact that we do possess a "modal microscope" to explore possible situations: we only have to implement them. In both cases, the way out of underdetermination is the same: make an experiment to decide between the two theories, and empirical adequacy is not underdetermined in any problematic way.
A formal argument
Here is an illustration for this claim. Imagine two populations that would believe in two modally incompatible theories. For example, T1, which predicts that golden spheres of 1km of diameter would collapse, and T2, which is identical to T1, except that it predicts that golden spheres of 1km would be stable. Imagine that their theories are both very well tested, but in order to keep good neighbourly relations, these populations refuse to make the crucial experiment, the one that would allow them to choose between the two theories: they never attempt to create golden spheres of 1km of diameter. Would we say that these populations both know, by induction, that their theories are empirically adequate? Or would we say that they didn't perform enough tests to know that?
The latter I contend: what if the crucial experiment naturally happens somewhere in the universe? What if an alien civilisation is creating big golden spheres somewhere? Then, surely, one of the two theories is not empirically adequate, and since the two populations do not know if this crucial experiment actually happens or not, they cannot claim to know that both theories are empirically adequate. The result is the same if they cannot perform the test for technical reasons. Any suspicion they could have against modal adequacy would be a suspicion against traditional adequacy, since they do not know which types of situations are actualised everywhere in the universe.
This simple illustration can be turned into a formal argument, as I did in my PhD dissertation. The most contentious premiss of the argument is the following: that we cannot know, by induction, that a certain type of situation is never exemplified in the whole universe. I think this premiss is true for most practical purpose: only a cosmic fact (something that has to do with the initial conditions of the universe for example) would allow us to know such thing. Local facts are not enough, because scientific theories never refer to particular times, places or objects, and no local fact could tell us that a particular model of a theory does not apply somewhere in the universe. If two theories are underdetermined by a cosmic fact, then we are not far from inflationary metaphysics indeed (we are talking about the hypothetical predictions of our theory in a universe completely different from our universe), and I'm ready to accept that we should suspend our beliefs in such cases. However, this is far from being the general case in science.
The conclusion of this argument is the following: in general (i.e. apart from cosmic underdetermination), if we cannot know that a theory is modally adequate, then we cannot know that it is empirically adequate either. The converse is even more telling: if we know, by induction, that a theory is empirically adequate, then we know that it is modally adequate as well. The traditional empiricist has no reasons not to be a modal empiricist.
As I said above, all this works if we start with the assumption that there is necessity in nature. It also rests on the idea that empirical confrontation is a matter of situations rather than the universe as a whole: importantly, if physical theories were making assumptions about particular objects, times and places, the argument above wouldn't work, but this is ensured by spatio-temporal symmetries: model predictions are insensitive to the specific time, place or object to which the model applies. Finally, it rests on an emphasis on interventions, not only observations, in experimentation.
An empiricist can still believe that there is no necessity in nature, but at least, her belief cannot be supported by principled arguments against the possibility of modal knowledge, and it looks like a dogmatic position. In the next post, I will give more reasons to think so: I think scientific practice hardly make sense if we do not think that there is necessity in the world, and that scientists are attempting to reveal relations of necessity between possible interventions and observations.