Scientific Progress

Let’s take ‘scientific progress’ to mean something like ‘moving closer toward a ”correct” conception of reality.’  If this is so, we might ask whether progress is better achieved through methods of falsification or through Kuhnian-style revolution.  I suspect the former, and will briefly explain some of my thoughts.

Popper claims that the goal of science is to move to some sort of objective truth.  This seems right.  Intuitively, there is some one way that things are, that we might better describe or represent as we move  from some scientific model M to a better scientific model N.  Revolutions, on the other hand, are predicated on some sort of ‘gestalt-switch.’ A revolution isn’t just the emergence of a better fitting theory, it is a complete replacement of the particular scientific enterprise’s conceptual apparatus and framework, incompatible and incommensurable with the theories that came before.  A reconception does not entail an advance. 1

If we think that there is some true, or ‘correct,’ conception of physical reality, then falsificationism will reliably lead us toward some correct picture.  The feature that gives falsificationism its tight relationship to reality is its reliance on a sort of infallible, deductive method, modus tollens. 2  The affirmation of the consequent of a hypothesis does not entail the truth of that hypothesis, but the negation of the consequent of the hypothesis does entail the falsity of the hypothesis.  We cannot definitively ascertain the truth of any hypothesis (for confirming the implication of a hypothesis does not entail the truth of hypothesis — perhaps there are other ways that the test implication could have turned out true), but we can definitively know its falsity.  We move toward a more correct conception of reality by deducing ways that the world could not be — eliminating false hypotheses — in order to narrow down the scope of possibly true hypotheses or ways that the world could be.  Narrowing the range of possible truth marks an advance toward a correct conception of reality.

It is important to observe that we only consider salient or otherwise relevant hypotheses.  Or else you might suppose that there are an infinite number of ways that the world could be, and the elimination of any number of hypotheses doesn’t narrow down the field (for $\infty - n$ is just $\infty$).

Simply falsifying theories is not enough, however.  For some falsifying results are due to instrumental error (for instance, when it was ‘discovered’ that neutrinos could travel faster than the speed of light), or perhaps the falsity of some background hypothesis.  In the spirit of Lakatos, we should not throw out a theory until we have a better one to replace it (one with more explanatory power, or is consistent), which has excess content over the unrefuted content of its predecessor. 3  This acknowledges the history of scientific progress — it is not necessary that we have falsified the hypothesis under question, for we may have falsified a background assumption instead.  Consequently, our tests don’t tell us that our hypotheses are false, but merely inconsistent.  (That is, we have a set of hypotheses we take to be true.  In testing a hypothesis, its falsification only tells us that the hypothesis does not fit with how we take the world to be [and how we take the world to be is inevitably steeped in other theories.)  The correct conception of reality is going to be one that maximizes explanatory power and consistency, (and arguably simplicity), for there are myriad ways to describe laws and relations in the natural world and intuitively there seems no reason to suppose that, if two theories have equal explanatory power, consistency, and simplicity, then one of them must be false.  Rather, it seems like the same story is being told in two different languages.  Both are correct conceptions.  But we are spurred toward closer representation of reality by advancing through more and more complicated scientific models as we develop a consistent set of theories with explanatory power.  This is a move toward best-representation, if you will.

Because revolutions are more gestalt-switch 4 than they are progress re vera, they don’t move toward a correct conception of reality.  Indeed, Kuhn argues that no paradigm is closer to ‘the ideal,’ but are rather selected for their ‘puzzle-solving’ capabilities.  This sort of thing should not move toward a closer conception of reality.  Puzzle-solving is something arbitrary.  What is or is not considered a puzzle is contingent upon the current paradigm. 5  Instrumental applicability in a domain relativizes the scientific enterprise to a specific, arbitrary set of problems.  A revolution is not required to respect the puzzles of the paradigm it is revolting from — the reconception reconceives the problems of science.  So we only ever explore some new way of conceiving of reality rather than progress toward some better model, picture, or representation of it.

A revolution provides a new paradigm, a new way of observing and analyzing, but progress must still be made through falsification.  It is by observing effects that were not predicted by our theories (a sort of falsification) that forces us to come up with some better explanation — an explanation which may or may not already be given by the model/framework we are in.  Revolution is not necessitated — the falsification provides all the impetus we need.

Consider the noble duck-rabbit (a visual gestalt).

Seeing the duck-rabbit as a duck and then afterward as a rabbit does not bring you any closer to a correct conception of the duck-rabbit, just a different one.  But by discovering what won’t do as an explanation forces you to work toward some solution that is not subject to the same counterevidence, and so the model continually becomes a better fit or picture of reality.

The result of a revolution is that we see the data with a different conceptual framework.  But the result of (Lakatos’s) falsificationism is that we develop a model that more accurately represents the data.  The latter brings you closer to some goal.  The former does not.

1. An example of a revolution would be our replacement of the Newtonian mechanics paradigm with relativity or QM.  A complete evolution of our scientific, conceptual apparatus.
2. Modus tollens is the following.  $P \rightarrow Q$. | $\neg Q$. | $\therefore \neg P$.  Contrast with modus ponens: $P \rightarrow Q$. | $P$. | $\therefore Q$.  What is not a logical rule of inference is: $P \rightarrow Q$. | $Q$.  | $\therefore P$
3. Theories build on each other.  That a theory is falsified does not mean that all the content of that theory is false.
4. A gestalt is when only one theory or another can be applied toward a set of data at a particular time.  The switch is when you move from applying one concept and instead apply another.
5. The conceptual framework of a scientific enterprise determines the nature and direction of the work — that is, what puzzles are important to solve.  ‘Normal science’ consists in the application of the paradigm to more practical scientific problems and questions.  When a puzzle is encountered that the paradigm cannot possibly answer, then a crisis ensues, followed by some kind of revolution.
Advertisements