Sunday I lead a philosophy of science round table discussion with the Seattle Analytic Philosophy Club. I really enjoyed the discussion, which could have gone on far longer than the three hour session. It was attended by enthusiastic individuals from a variety of backgrounds, including computer science, artificial intelligence research, psychology, medicine, and physical science, and I was very happy that everyone was willing to contribute.
Some themes from the discussion:
1. Why do we theorize?
- To obtain the power to act intelligently in the world by predicting regularities.
- Human (finite) memory requires that we generalize from experience instead of remembering individual instances.
- We have evolved to theorize for it usefulness and survival advantage.
I would also like to add:
- Extrapolation of knowledge from past situations to new, unfamiliar situations. This is important because it highlights that we do not simply react in a knee-jerk way to observed regularities, but that we use our theoretical vocabulary in a sophisticated way that cannot be replaced purely by stimulus-response rules.
2. Should we believe in our theories?
I was happy to find that the discussion here gravitated more around dimensions of what qualities reinforce our belief in theories, including:
- Internal consistency
- Predictive ability
- Integration (between adjacent areas of thought, or domains)
There was also a really cool Jigsaw Puzzle analogy proposed for theorizing, in which belief in each element of a theory is somewhat independent of other elements, but also somewhat interdependent on how it fits with other elements.
There was some discussion of:
- Ontological question: Is there really a world there to be known?
- Epistemological question: Can we really know the world with a high degree of certainty?
- Semantic question: Can our language really reflect reality?
I summarized standard positions on belief:
- Skepticism: we can't really know anything
- Empiricism: we can know some things from experience
- Hybrid view: We can be skeptical of things we cannot directly perceive while being empirical about things we can perceive. (Bas van Fraassen holds this view and calls it Constructive Empiricism.)
Variants of empiricism:
- Scientific Realism: good scientific theories may be largely true.
- Constructive Empiricism: good scientific theories may only be "empirically adequate."
We discussed some arguments surrounding belief, including:
Arguments against strong belief:
- Underdetermination: For any given set of experimental evidence, there are multiple theories that could explain the evidence.
- Historical argument: Over time, our theories may be rendered inadequate by future discoveries.
Arguments for strong belief
- No Miracles Argument: for our best theories to have the predictive ability they do, it would be a miracle if they were not true.
I would like to add two that we didn't cover.
- Counter-Historical Argument: Are beliefs refundable? Can we believe in something now but reserve the right to change our opinion when new evidence is offered?
Counter-"No Miracles" Argument: Bas van Fraassen has suggested that good scientific theories evolve through a kind of scientific Darwinism to reach a high level of predictive ability irrespective of its relationship to Truth.
3. Relationship to Quantum Theory
I briefly summarized some of the themes in the book, including:
- Quantum theory was motivated by positivism. Authors of the theory did not want to produce a mechanistic model to explain the behavior of electrons in the atom, instead opting for a mathematical scheme to calculate observables.
- Quantum theory became an algorithm for mapping mathematics to observables, instead of a physical theory of nature.
- Quantum theory is an inadequate theory for calculating observables such as the ionization energies of atoms and ions beyond hydrogen, as well as the spectrum of atoms.
- Quantum theory has become self-reinforcing, because the weird interpretations that come from the theory are assumed to be due to Nature, instead of due to the inadequacy and conceptual murkiness of the theory itself.
One of the major theses from the book is:
Theories designed to be empirically adequate are more likely to allow scientists to ignore new evidence that is inconsistent with the theory because any new evidence can, with trouble, be absorbed into an algorithm of sufficient complexity, and made sense if the conceptual framework is sufficiently vague.
4. Is it Rational to Resist New Ideas?
A major theme from the book is the Semmelweis Effect, in which some paradigm-shifting proposals encounter seemingly irrational resistance by the scientific community, followed by a delay of decades until acceptance.
I do not point to one single cause of this, but rather a constellation of causes that may be unique in each case. In the book you will find the following themes:
- Worst possible timing
- Lasting association with cold fusion
- Wrong credentials
- Scientific culture of biomedicine vs physics
- Long time since major discovery in physics had immediate technological potential
- The voice of scientific authority
- Terms for collaboration on a trillion dollar invention?
- Too much – new theory + new application of new theory
- Theory inertia
- Ability of quantum theory to feign success
- Philosophical assumptions underlying theories
In our discussion there was the proposal that some amount of resistance is "rational" in a Bayesian sense, because the failure to adopt a new theory quickly is made up for the energy savings of investigating every new theory that comes along.
5. The Nature of Theories
Unfortunately, we did not have time to discuss the nature of theories. In the book I discuss the Semantic (or Model-based) View and make some small modifications to the position held by Suppes and Giere.
The basic idea is that a theory is made up of a cognitive model composed of:
- Concepts and their definitions
- Principles: A number of meaningful propositions that artfully express the important relationships in the model.
- Narratives composed of principles, primarily for the purpose of showing deductions or guiding thought.
My unique position, advocated in the book, is that each model is composed of an open-ended set of principles. Principles not important to the key relationships of the model are tacitly assumed but always interwoven with wider spheres of knowledge.
Every model is a model-class that applies to a range of systems specified in the principles, but may be narrowed down in scope until we are speaking of a physical concrete.
Is the idea of a "model" also a good explanation for everyday concepts? Perhaps all knowledge is structured the same way, and scientific theories are only everyday knowledge taken to a higher degree of refinement.
- Dropbox folder containing a few papers on the Semantic View
- My undergraduate thesis: Scientific Realism, Empiricism, and Quantum Theory
- The book: Randell Mills and the Search for Hydrino Energy