What basis do we have for certainty in current scientific theories?

Question

Given there is much past scientific belief that we now know NOT to be true, what basis do we have for the seemingly increasing certainty in our scientific beliefs held today being true?

On the one hand, we prize Popperian falsifiability and yet still have a sense that we can be quite certain of some things scientifically.

I am thinking less about approximations such as Newtonian mechanics (which, even in light of relativity or quantum mechanics, is still useful) and more like phlogiston theory or Aristotelean physics. See http://en.wikipedia.org/wiki/Obsolete_scientific_theory for a decent list.

Answer 1

Isaac Asimov wrote a great essay related to this, The Relativity of Wrong. This quote summarizes his point:

My answer to him was, "John, when people thought the earth was flat, they were wrong. When people thought the earth was spherical, they were wrong. But if you think that thinking the earth is spherical is just as wrong as thinking the earth is flat, then your view is wronger than both of them put together."

You somewhat touched on this by mentioning Newtonian physics being technically wrong, but still practically useful. This is a key point. Science isn't very concerned with how things actually are. Rather it is concerned with predicting how things will behave under certain specific circumstances. Thinking the Earth is flat is wrong, but operating with a model that the Earth is flat is pretty valid for most of what we do. A more accurate, yet more complex, model is the Earth as a sphere. This becomes the better choice when covering long distances. At the end of the day though, it doesn't matter much that both of these models are "wrong", as long as they continue to provide useful data.

As for the theories you mentioned, the key is that there was never much empirical evidence supporting them. Had there been, then that alone would have made them useful, despite being wrong. Science is not smart people thinking about things and coming up with ideas for how things might be. Rather it is smart people examining data and coming up with models that explain this data and make other testable predictions.

Reading over that wiki list shows a lot of interesting ideas and not a lot of useful models. Relativity is an example of a theory that raises a lot of eyebrows, but has a ton of evidence supporting it. Even if in a generation we discover some serious flaws in relativity it will not fall completely out of favor the way the theories on that list have. The key difference once again, is that relativity agrees with and explains a huge body of evidence. Those "theories" had no real evidence backing them, and often times were simply the first idea people came up with when they had no way to really make detailed observations.

String theory is a good example of an exception that proves the rule. It is a modern theory that may well end up on that list (or its successor). The key is that there isn't much (if any) evidence. It doesn't make any testable predictions yet. String theory may yet turn out to be useful, or it may turn out to be completely bogus. The important difference is that string theory, while widely known, is not widely accepted, and has not provided any useful models.

So, the answer to Are our current theories wrong (in the strictest sense that Newtonian physics is wrong)? is yes. No scientist will argue that we have totally accurate models. But they are useful models, that produce actual results, and are more accurate than any other model previously developed. If that's wrong, I don't want to be right.

Answer 2

It would be hard, I suppose, to improve substantially on Richard Feynman’s discussion of uncertainty in science, from a book of lectures, The Meaning of it All.

The bare bones of his idea are that we cannot be absolutely certain of anything, and thus have to remain open to a revision of our ideas if and when new facts come in, but that we are dealing with predictions to which we assign a certain probability. This probability is a product both of correct predictions in the past and rational theories about how certain facts about the world are related, explaining which factors should influence our assigning of probabilities.

Also, the more cumulative our insights into the workings of the world are, the more certain we can be that we are on to something. In this way, Newtonian mechanics becomes a special case of Relativity, which in turn becomes a special case of quantum mechanics. The trajectory here is from explanatory models with more exceptions to those with less exceptions. The only rational conclusion in light of such evidence is to (provisionally) assume that our models are getting closer to what we might as well call ‘reality’.

Of course, DaleSwanson’s answer above mentioning Asimov’s “The Relativity of Wrong” is just as important a factor in doing the question justice, if not more so.

Answer 3

We have confidence in scientific theories, not unshakeable convictions in our knowledge of their certainty.

A scientific belief simply cannot be justified by any other means than the various evidences produced by scientific analysis and experiment. I'm not sure it's fair to demand that a theory be absolutely true; rather it should predict things, solve problems, teach us to build things -- and ideally indicate further directions to explore.

Accepted scientific theories generally aren't proven wrong outright but rather superseded by more general theories that preserve most existing techniques -- as you suggest Einstein supplanting Newton simply didn't prove that Newtonian physics was "false".

Also, the ultimate "truth" of a scientific theory is probably the wrong criterion (and may even be besides the point.) That is to say, a theory either is or is not effective at predicting things and solving problems. But the fact that a theory lets you predict something still doesn't tell you that the theory is true -- or indeed anything at all about any underlying "reality," for that matter.

Answer 4

I don't think it's true that scientific beliefs are increasingly held to be true. At least in philosophy we have learned the lesson of the "Pessimistic Induction".

What has changed is the general public's perception of science. They might well be increasingly certain of science. That's due partly to their increasing use of and reliance on the products of science. It's also partly due to some awkward facts about the rehtoric of public scientific discourse. Put bluntly, if you don't sound sure of yourself, you don't secure research funding. And then there's the whole co-opting of science to prop up the militant atheists like Dawkins and co. But let's not go there. No one sensible should take Dawkins seriously as a philosopher. Certainly not as a philosopher of science.

So why might one have an increasing confidence in science despite Phlogiston, Caloric, Ether theories of light, Humour theories of disease and so on... Well, there's always Structural Realism. This is basically the view that while there have been notable false theories, they contained, somehow, a grain of truth. There is an increase in our understanding of the structure of the world, despite revolutions about what we think it is made of. So despite wrong, Fresnel's ether theory of light got some stuff right: namely the equations for diffraction etc...

The motivation for this sort of view is the "no miracles" intuition. The idea is that it would be a cosmic coincidence if we managed to build computers, aeroplanes, microwaves and the like, if we weren't getting something right about the nature of the world.

This all falls far short of certainty. But then, certainty is in short supply in philosophy.

Answer 5

I take a somewhat pragmatic approach. It's unlikely that the ratio of solid scientific knowledge to bogus knowledge is much different today than it was at other points in human history. My basis for the claim comes from reading histories of science and scientists. Invariably, false scientific ideas turn out to be reasonable and potentially correct explanations of the available data. When advancements in instruments, observation technique, and experimentation arise, they tend to make falsified science look wrong. At the time the ideas were proposed, that data didn't exist. There's no reason to think we've reached the limits of observation.

Further, science seems to advance on the backs of creative insights which come from the minds of especially gifted thinkers. (My thinking on this comes from my reading of Thomas Kuhn, especially The Structure of Scientific Revolutions.) Our minds have the remarkable ability to tolerate inconsistent data and we can go a long time accepting mistaken theories before a sudden insight causes us to accept an alternative that better fits the data. The feeling that every bit of new data fits our models is a dangerous bias. We tend to dismiss counter-indicators as "outliers".

Now, we probably do know more in an absolute sense. Given that new knowledge is informed by previous work, we've learned from mistakes and refined theories. Engineering feats that depend on fundamental science validate (from a pragmatic sense) that we know more then we once did. But we also assert many, many more things then we once did. Some things we "know" are based on natural phenomenon that were unknown even a generation ago. (I'm thinking of things like medical claims, computer science theories, and particle physics. There are likely theories that I can't begin to understand in other realms of knowledge.) In order for the ratio of things we know to keep up with the number of things we claim, it's inevitable that we know more in an absolute sense.

A more difficult question is how do we know which claims are true and which claims are bogus. The example of Aristotelean physics suggests that simply picking the most respectable theories isn't a certain way to select true theories.

Answer 6

It seems some great answers have been given, but just to add something else:

Regardless of whether a scientific theory is later proven to be incorrect, the data and testing that previously supported that theory are still just as correct and useful. It is just our interpretation of that data which is brought into question.

I feel this is important to mention, because I have often heard people mention that science's practice of regularly tossing out previously held beliefs means that it is inconsistent and untrustworthy.

If tomorrow, the theory of gravity was proven to not be accurate in describing some newly discovered phenomenon, that does not make it any less accurate at describing the things it used to. It is no longer an all encompassing framework that can be used, but that doesn't change the fact that 99.9% of what it states remains true.

Answer 7

David Deutsch's last book is more or less exclusively on this topic: The Beginning of Infinity

There he argues that although objective knowledge is attainable, justified belief ('certain truth') is not.

That means that we will be able to come ever closer to the truth but never reach it. Explanation after explanation will be better infinitely - but as with all infinities we will always be at the beginning of it (like every number you can conceive of is always small compared to infinity).

Reading this monumental piece is highly recommended!

Answer 8

The basic reason for falsification is not to inspire doubt, but to insist that every addition to our field's complexity is warranted by at least slightly better returns on investment.

The point of Popperian falsifiability, or Kuhnian normal science, is to insist that adding to the theory should improve the odds of any given calculation improving. New theories must always adequately explain what has already been validated, or at least explain why our current reliance on the existing principles appears to work, and clearly predict something else. So, the math says, you must be increasing the number of your predictions that will be correct.

Attempts at falsification use the theory we attempt to extend. So, to a real and important degree, every case of the non-falsification of a falsifiable theory is a re-validation of every theory it was based upon back down to the first principles. If we fail to falsify, we are validating a wide range of computations made according to the underlying science.

So what we have, if we stay in this framework, to continually increase our certainty, is statistics.

Answer 9

Why all this discussion? Like Feynman said we cannot be absolutely certain of anything.... You might be a brain in a vat and you want certainty?