Science and scientific theories

by Neil Rickert

This is partly a comment on “The Knight’s Song, or What is a [scientific] theory?” and partly a post on my own view of science and how it differs from what philosophers of science say.

If we follow the Shannon-Weaver theory of communication, then

  • we start with semantic information (the natural world, as studied by science);
  • we encode that in a symbolic form (syntactic information, Shannon information, linguistic representation);
  • that syntactic information can then be transmitted or recorded;
  • a final receiver of the syntactic information can decode it to recover the semantic information.

With science, the method we use for symbolically encoding nature is what we call “measurement”.  This process of encoding produces the data on which science very much depends.  I also discussed this way of looking at measurement in an earlier post.

When we look at science in this way, the encoding step, or measurement step, requires some kind of a protocol, or a set of agreements on how to encode.  We usually describe this protocol as a measuring convention.  And there is something necessarily conventional about the encoding.  The world does not come to us in linguistic form, nor in symbolic form.  In order to have a symbolic representation, we need a convention to define how that encoding is done.  Nature does not itself dictate what that convention will be.  The convention necessarily depends on human decisions.

When we get to the decoding or interpretation step, the same conventions are used.  If we know how to encode the world as a symbolic representation, then that symbolic representation tells us something about the way the world was when the encoding was made.

As I see it, the most important part of a scientific theory resides in the establishment of those encoding protocols.  Without those, we could have no data at all about the world.  Where the protocols are used to generate multi-dimensional data, and when the design of those protocols is sufficiently systematic, there may be implied relationships in the symbolic representations.  These relationships are often described as scientific laws.

Here are some consequences of looking at science and scientific theories in this manner:

There are no laws of nature.  There are scientific laws.  But those scientific laws express syntactic relations that occur in the data.  When we talk of laws, we expect those laws to be law-like, which is to say that we expect them to express something like mathematical relationships.  But those relationships occur only in the syntactic data that results from our encoding of the nature of semantics.

There might well be natural relations in the semantics.  But these relations would be in the form of connections and relevancies between various semantic aspects of the world.  They could not be law-like, because law-like relations can only be relations between syntactic elements.

Data is theory laden.  More precisely, data is necessarily protocol laden.  Without the protocol for encoding semantics as symbolic data, there could be no data.  To the extent that a scientific theory is part of the data encoding protocol, that data will appear to be theory laden.

Many scientific laws are analytic.  That is to say, they are true by virtue of the meanings of their terms.  This will be true whenver scientific laws are deduced from the encoding protocols, for those encoding protocols define the meanings of the terms.

Philosophy of science tends to describe science as operating entirely on propositions.  That is, it sees science as operating on syntactic representations.  It tends to describe scientific laws as arising by induction from data.  I cannot find any evidence to support that view of science.


One Comment to “Science and scientific theories”

  1. Hence my argument that we are talking about something different when we discuss theories in philosophy from when we discuss scientific theories.

    An aside: for measurement, Fisher (Cramer-Rao) information is better than Shannon.


%d bloggers like this: