I can't handle the Truth

People involved in litigation tend to think the Truth of their position is glaringly obvious, and can't understand why there is any uncertainty about the outcome. (This is particularly bad in divorce cases.) I finally came up with "The law deals in evidence, not truth," which actually seems to get through to some clients.

"It is actually science's greatest virtue - that it doesn't seek truth, but good theories"

That is nonsense John. Good theories are of course theories that seek (approximate) the truth. You say that science is about "confirmation, reliability, testing, elimination", but the reason that an idea or hypothesis can be tested, confirmed or eliminated is - obviously - that there actually is an underlying truth (or Truth) that determines the evidence.

Morten: that is a view that has some traction. It is known as scientific realism. But it's not only not the sole alternative - some others include instrumentalism, internal realism, structural realism, scientific idealism, and so on - it's not necessary for science to be done. Nobody need think there has to be an underlying reality isomorphic with their theories to do science, although it may be the motivation for many scientists individually.

Science requires only that the theories be empirically adequate, and that they are the best explanation on the board right now. They needn't even be the best possible explanation: there are an infinity of interpretations for any theory, and an infinity of theories. So long as they are parsimonious, and empirically adequate now and in the future, they are sufficient.

Of course a theory must also have explanatory power in various ways - any good introduction to philosophy of science will outline what those virtues are. It doesn't automatically mean that a theory must have a mechanism, or that the mechanisms it does have must be fully explicated (think of Newton's theory of gravity). It merely has to account for what is observed.

Ahem,Morten,

but the reason that an idea or hypothesis can be tested, confirmed or eliminated is - obviously - that there actually is an underlying truth (or Truth) that determines the evidence.

Not so.

I can have a theory that eating Mcdonalds makes kids fat,or that Scandinavians are bad in maths...Now this of course could be tested in all sorts of different ways,but the outcome of these tests would be unrelated or uninfluenced by any "underlying truth",since they would form their own evidence or truth.
I believe this particular philosophical problem was solved in the mid-19th century,give or take....

I'm with Morten here.

You keep saying that theories don't need to be the best possible explanation, only the best current explanation. Of what? Of the way things really are, no? That's why, I believe, Morten used the term "approximate". No scientist worth their salt will say they have privileged access to the absolute truth. Good theories are better approximations of the way things really are, or else they wouldn't mesh as well with the current available evidence or have as much predictive power as worse theories.

I have to go to bed - it's nearly 2am here, but let me say this - the notion that scientific theories approximate truth is a venerable one, but it doesn't easily hold up (I won't say it failed - philosophical theories are rarely disproven - but it hasn't got many advocates since Popper). For a start, by what metric are you measuring truthiness here? Percentage of falsity? How could you know that? What does it mean to say some theory is more true than another it replaces? More bits of it are true? But until we have a final theory, how can you know bits of it are true?

All attempts to talk about verisimilitude resolve down to empirical adequacy (in both retrodiction and prediction). And that means we can eliminate the need for Truth in respect of scientific theories. There's just the evidence and the epistemic aspects, such as Bayesian likelihood.

John: Why just "Bayesian" and not "frequentist," or Neyman-Pearson"? It seems that you are edging toward some sort of inference structure based upon probability conditioned on evidence. If so, huge fractions of thought about conditional probability lie outside of a Bayesian framework.

Surely, there has to be an assumption of some underlying reality which science is attempting to understand, otherwise what is the point of doing it at all? But, given that the nature of this underlying reality - or Truth - is somewhat elusive, it makes sense to go with theories that actually work, at least for the moment. Relativity theory may be a better description of reality than classical mechanics because it explains more of what is observed, but if you want to land a spacecraft on Mars and its quicker and easier to go with Newton rather than Einstein then that's what you do.

M-theory or string theory sound like interesting ideas - at least as far as an innumerate layman can be said to understand them - but they don't seem to actually enable us to do much. That's not to say we shouldn't be working on them because we should. It seems to me that we don't have so many ideas that we can afford to junk any except those that have been shown clearly not to work.

If one theory replaces another, would we not say it is simply a better approximation of reality (if one presumes such things and I do); it explains more, or more accurately, obervations of the world? How else are you defining 'best explaination on the board right now' or 'best possible explaination' if not comparing it to something else?

Perhaps it is the value laden nature of 'truth' that is the problem. If we said scientists were trying to uncover an accurate description of how the observable world operates, as testable by many methods such as predition, retrodiction, baysean liklihoods etc, which is arguably what many scientists mean by 'truth', would this be easier to handle? Otherwise what is it our theories are sufficiently doing?

thanks

John: you seem to equivocate here, on metaphysical and semantic vs. epistemic points. The virtue of science lies in its search for truth, and truth is the ultimate virtue of a scientific theory that trumphs all other virtues (in fact, the other virtues are virtues only insofar as they lead to truth). This is not "realism" - instrumentalists and anti-realists and realists agree on this point, and only radical and incoherent post-modernists would possibly disagree. (Spedding is on the right track here)

A completely different question is how we can know that our theories are true or what it takes for our models of the world to be true modellings of the world - and whether there is some underlying reality isomorphic to our theories (and what on earth 'isomorphic' could mean in this context).

The point is that 'truth' is an extremely easy concept. 'S' is true if and only if S. Grasping the concept is also eminently easy; in the terms of inferential role semantics it consists of the grasp of two ridiculously easy rules, the introduction rule 'if S infer that 'S' is true' and the elimination rule 'if 'S' is true, infer that S'. You'll be hard pressed to find a simpler concept than this. There just is no hard question to be asked about what 'is true' means (i.e. the intension of 'true'). This doesn't mean that there aren't hard philosophical questions about truth to be asked, of the kind 'what makes a linguistic unit satisfied by something non-linguistic' (i.e. the 'isomorphism'-question from above; that is where all those standard 'correspondence', 'coherence' or 'pragmatic' theories enter the picture; they are theories about how a sentence is made true, not what its truth consists in (and are, I think, overall confused)), and not the least about which claims are in fact true (the extension of 'truth').

There is no question, therefore, what scientists (or anyone else) is talking about when they are talking about truth. And there is no question that truth is what they are searching for (unless they are faking data for personal gains, or something). There are lots of questions, on the other hand, to be asked about how justified their claims are, and lots of questions of what it takes for scientific theories to be justified. There are some metaphysical questions as well (i.e. realism). But those are separate questions.

There are some deviant uses of 'truth' out there, of course, but these can be easily explained away:

"Truth is such a difficult concept": False, and such a statement usually signals that the speaker doesn't like the facts and wants to disregard them in order to advance some agenda.

"True for me": Meaningless, and usually signals that the speaker lacks good arguments for some stupid but cherished belief and doesn't want the facts or other people to interfere.

Convention T: "S" is true if and only S

Let's call that "T".

"T" is true iff T.

Call that, err, T+...

So the convention itself is subject to a regress. There's another regress:

"S" is true iff S - call that S'

"S'" is true iff S' - call that S''

...

These regresses (the latter one is called "semantic ascent" suggest that Truth is not so simple a concept. Since each sentence is a representation of states of affairs (here, S), the relations here are purely semantic. And since Convention T itself is a representation, it, too, never leaves the semantic prison.

And yet what Truth is supposed to do is show the relations between representations and the world. And it doesn't. In the end, Truth is just a term of approbation for some model or set of propositions. Which is why it is enough to say that an empirically adequate theory will do, whether or not it is true.

I like to make a distinction between truth and reality. Reality is what is tangibly out there, that in which the argumentum ad lapidem is not a logical fallacy. Truth might be the same (which is of course the most parsimonious option) -- but it might also conform to any of a large number of other unfalsifiable ideas, for example solipsism or the Hinduist concept of maya or the idea that God is the solipsist or who knows what. Science is then concerned with reality, but not at all with truth. Reality may or may not be true; I don't know, and I don't care. =8-)

John: "Which is why it is enough to say that an empirically adequate theory will do, whether or not it is true."

An empirically adequate theory may be fine - right here and right now. However, if the theory is false contradictory evidence will accumulate as science progresses making the theory empirically inadequate. Of course then the next step will be to develop a new empirically adequate theory, which will again hold until falsified - unless it is actually true. This process of developing new or refined theories to explain an exponentially growing
body of evidence will inevitably - at least in the long run - steer science in the direction of Truth. And this will happen whether or not we are able to measure truthiness of theories. Denying this is tantamount to saying that current scientific knowledge of the world is no closer to the truth - or at least that we cannot know whether it is closer to the truth - than say 500 years ago (which would in my opinion be pretty silly).

Another way to look at it is to consider two empirically adequate theories explaining the same phenomenon, a situation I cannot imagine any scientist would be satisfied with. And the reason is of course, that s(he) would know that at least one of the theories would have to be false. Empirically adequate theories may be all that is needed in e.g. engineering or medicine. Scientists also want their theories to be true.

will inevitably - at least in the long run - steer science in the direction of Truth.

Of reality, in the terminology I prefer. Truth is only attainable if reality and truth are indeed the same thing, and that's an untestable hypothesis.

I know it's rather pedantic to expect people to use words with some care -- but it's a good habit to cultivate since clear thinking and clear expression of what is thought tend to go hand-in-hand. So consider:
1) The rock I'm holding is real. That makes sense.
2) The rock I'm holding is true. Utter nonsense.

Eliezer Yudkowsky has a wonderful article called The Simple Truth.

Suppose the intelligent species on earth would not come from primates, but from bats. After enough progress, they will discover a curious thing we call "electromagnetic radiation", and will be able to get some truly revolutionary insights on what previously has only been associated with weather (namely, the sky). Furthermore, they will be able to exploit all this theory to, say, better grow plants. But some skeptics would doubt that these findings - which are all built on abstract math and complicated machinery, but no direct sensory data - are in fact true. Rather it's best to just say "the math works".

To paraphrase Eliezer, truth is what a mind has when it is "entangled" with reality. Biological organisms have a lot of truth - they have to, otherwise they wouldn't survive longer than a random atomic mesh. Science is a way to get even more truth, to become better entangled with reality. Theories work not "just because", but rather because they quite literally mirror some aspects of reality, just like your brain absolutely must do if you are to do even the "simplest" things like walking. Or will you claim that life and science are magic?

You, by being alive for quite a few years, are handling a lot of truth (but not all of it). Science, by focusing on "confirmation, reliability, testing, elimination, and other Bayes-like activities", yields even more truth. Why dismiss or downplay this? Because others don't get it? Because they abuse the term?