Home > The SPLectrum Seed > The seed and an attempt at semantic information

The seed and an attempt at semantic information

An attempt has been made, from inside physics, to define meaningful information in physical terms — to say what it is for a correlation to be about something and to matter to whatever carries it. The developed account of that work, and of the field around it, is on the semantic information piece.

It is a small attempt and a young one, and what it reaches for is meaning approached without a boundary drawn in advance around where meaning could be. That is the seed’s own subject, arrived at from a direction the seed did not take.

What the attempt runs into

The construction needs three things specified before it will produce a number. Its authors name all three and call them subjective choices of the scientist: where the cut between system and environment falls, over what interval viability is assessed, and what standard of doing well counts as viability at all.

They name them as a limitation to be acknowledged. They are instead the shape of what was being done, and each has a place.

The cut — being. Where the boundary falls decides which thing is under test. There is no system in the world waiting to be measured; there is a boundary drawn, and what falls inside it is the thing the question is being put to. P0 has being and language together, and the cut is where that shows: to have a thing at all is to have it delimited in language.

The interval — the heartbeat. Viability is assessed over a stretch of time, and a self-maintaining process has a rate — a cycle, a turnover, whatever it does to go on being what it is. The rate is not chosen for the process; it is what the process runs at. This is where the attempt’s own examples turn. A stone tested over an interval suited to a bird carries nothing that bears on its persistence, and scores nothing. Tested at the rate its own process runs — water, temperature, chemistry, the slow work that will eventually leave no stone there — the correlations that looked inert are the load-bearing ones. The zero is a mismatch of rate, not a fact about stones.

The standard — concept membership. What counts as still doing well is what counts as still being one of those, and that criterion belongs to the concept the thing is held under. It does not come with a sharp edge. There is no moment at which a stone stops being a stone; the process runs, and somewhere along it the concept stops applying, without a line being crossed. How concepts hold their membership this way is a question the seed will have to return to.

The three are not three independent dials. The stone’s case runs through all of them in order: the cut gives a thing, the process runs at its own rate, and far enough along that rate the membership lapses.

Two models, and no third place to stand

What the construction actually does is test a system against an environment. Neither is the world. A system simple enough to be specified is set against an environment simple enough to be specified, and the quantity falls out of the relation between the two.

There was never an alternative. There is no complete environment to test anything against. What a subject has is what it interacts with, and what it interacts with is partial — a facet, not a totality. P1 has language relational, and the relation holds between what is given in language on each side. Two partial things, and the meaning living between them.

The subjectivity the authors apologise for is not a defect in their construction. It is the condition of having said anything at all. A cut is partial; it illuminates one facet of something that has many; and the choosing is not a step that could have been skipped by being more careful. This is what language does everywhere, not what this formalism does badly.

Where meaning is, a relatum is

The construction set out to define a physical quantity, from thermodynamics and Shannon correlation, with no philosophical commitment driving it. The slot for a subject opened because the mathematics would not close without one, and its authors said so in print.

The measure itself is early and thin. What the construction found is that it could not proceed without naming the three things a partial thing in language has to have.