Is constructive empiricism a credible alternative to scientific realism? Commented [KH1]: ..

In this essay, I will argue that constructive empiricism is a credible alternative to

scientific realism. I will begin with defining both scientific realism and constructive
empiricism. Following this, I will outline two arguments which I believe discredit
scientific realism and give us reason to accept constructive empiricism: the truth-
success gap and the vulnerability criterion.

Scientific realism is the view that our best scientific theories are approximately true,
and the entities and processes postulated by those theories really exist, that science
really does give us an accurate picture of reality. In contrast, constructive empiricism is
the view that our best scientific theories are empirically adequate, meaning that what
the theories say about the observable world is true, but that we should be agnostic
about what those theories say about anything that is unobservable.
I believe all arguments for scientific realism
The most common argument given in favour of scientific realism is the No Miracles boils down to some version of the NMA, so
Argument (NMA). Essentially the argument boils down to the fact that science is very in a good sense it's not only common but the
only game in town so to speak.
successful, and that the best explanation for this success is that accepted scientific
theories are true, and therefore that we should believe that those theories are indeed
actually true. What’s obvious is that the main assumption underlying the NMA is that
there is a connection between success and truth. True theories will be successful, false
theories will be unsuccessful. Is this really the case though?

Newtonian mechanics seems to provide an obvious counterexample to the idea that

But realists would say that Newton's theory is a good
truth and success are correlated. Despite its description of the underlying structure of example of why the NMA is true: it gets it right for the
space and time now being rejected, Newtonian mechanics is able to make remarkably most part, only breaking down in the limit. Better
accurate predictions, such that we would be apt to call it successful. Moreover, even counterexamples would be theories that posit some
unobservable ontology that are abandoned in newer
contemporary science makes active use of idealisations in order to simplify their theories. Maybe Newton's theory can be read like
models and make computation easier. For example, in making models of the interiors of that but it isn't explicit at all.

stars, there may be many simplifying assumptions, such as that there is no radiation
For realists the problem of idealization can be met in other
pressure, that there is no density gradient, that the star is composed wholly of ways, like de-idealization. As far as I know, that's not
hydrogen, and that this behaves as an ideal gas. Idealisations such as these are always possible, but I believe usually scientific realists don't
take idealizing assumptions to be a serious problem
falsehoods but are likewise arguably successful. because they conceive of realism in ways that aren't as
vulnerable to idealization problems, e.g. successful
What about examples of true but unsuccessful theories? If a theory provided a reference of central terms.

completely true description of the unobservable, then it would entail that it also
provided a true description of the observable. The problem with this is that a scientific
realist can’t actually assert that any scientific theory is true, for the reason that our
science is not complete. Therefore, our best theories are just approximately true.
However, it’s not clear that entailment holds between approximate truth and success.
To give an example, let’s take current physics and increase the strong nuclear force by Increase the strength of the force in the world?
2%. On the assumption that current physics is approximately true, the modified theory
would be approximately true of fundamental forces and entities, but would give a
completely inaccurate account of the observable world and make totally inaccurate
But why would it still be approximately true if it gives
scientific predictions (estimations about the amount of hydrogen in the universe being wrong predictions?
less than is the case for example). We therefore have clear reasons to reject the NMA.
Nevertheless, rejecting the NMA on the basis of the truth-success gap doesn’t give us a
positive reason for accepting constructive empiricism, it only leaves us with a reason
for rejecting scientific realism. Why then should we accept constructive empiricism? I
believe that the vulnerability criterion provides the answer.

The vulnerability criterion is the idea that if two theories have the same risk or
‘vulnerability’ of being disproved, we should accept the logically weaker theory, the
theory which makes less claims about the world. When a scientific realist and a
constructive empiricist accept a scientific theory, the realist believes everything that the
empiricist believes but more, specifically in the existence of unobservable phenomena.
Being that the only way we can confirm or disconfirm a scientific theory is through
observation, the realist and empiricist will accept or reject a theory in exactly the same
circumstances. Since the realist and empiricist interpretations of a scientific theory are
both equally vulnerable to empirical tests, on the vulnerability criterion we should take
the stance of the empiricist.

One thing that might be highlighted now is that it's not immediately clear why we should
have any reason for accepting the vulnerability criterion. After all, we could further
argue that we should not believe in what scientific theories say even about things that
are observable but as of yet unobserved. At this point I’d like to introduce some
epistemological ideas. Whenever we decide what to believe, we have to perform a kind
of balancing act. If it were the case that we simply wanted to minimise any false beliefs
we might have, the best thing to do would be to never hold any beliefs at all. To not hold
Good point.
any beliefs, however, first of all is psychologically implausible and second of all would
leave us in a position in which we were unable to undertake any actions.

A distinction between two types of error, drawing on statistics, will come in handy here.
Type 1 errors involve believing what is false. Type 2 errors involve failing to believe what
is true. In order for us to be able to take actions in the world and fulfil our interests, we
require having true beliefs, not just the minimisation of false ones. Attempting to avoid
one type of error will pull in a different direction than attempting to avoid the other type
of error, as applying some kind of method to decrease false beliefs will also decrease
true ones and applying some kind of method to increase true beliefs will also increase
false ones.

My proposal is that the vulnerability criterion of belief is able to strike a good balance.
We are able to avoid type 1 errors through scepticism of the unobservable and we avoid
type 2 errors by holding the strongest possible beliefs given this criterion, which would
be that scientific theories do make true statements about observables. Holding of the
strongest possible beliefs is important since if we were to take on weaker beliefs, such
as scientific theories only making true statements about what has already been
observed, we could no longer make any predictions about the future, which would leave
us unable to act. So that we might satisfy our interests, we need to know what will
happen given different actions we might take, and thereby we need belief about
observables.
What we can note here is that in terms of achieving our goals, we need no beliefs about
the unobservable. After all, the only way the world affects us in any way we can be
aware of is via our senses. The observable world impinges upon us. The unobservable
only impinges on us via the observable world. Because scientific theories describe the
observable world with unprecedented accuracy, we can just take our best scientific
theories as guides to the observable whilst maintaining agnosticism about the
unobservable.

In conclusion, I have argued that the truth-success gap argument provides us cause to
reject the NMA and thereby scientific realism. In its place, I have proposed that
constructive empiricism is the best possible alternative. This is on the basis of the
verifiability criterion, which strikes a good balance between reducing type 1 errors and
reducing type 2 errors.

This is a clearly written and nicely argued essay. I like that there's a clear-cut two step
approach.

I did not understand the part about rejecting NMA, and I think the essay can benefit from a
clearer statement of what constructive empiricism is and what's particular to it (compared to
other antirealism about science).