Constrained invention

by Neil Rickert

This will mostly be a copy of what I recently posted in a Yahoo groups discussion.  And, incidentally, Yahoo badly mangled that post (stripped out most of the formatting).

As background, I’ll note that in an earlier Yahoo groups post, I had indicated that I was opposed to the view that perception is passive.  This seemed to puzzle some participants in the discussion.  So my post — the one I am quoting — was intended to explain what I mean when I say that perception is active.

The quoted post

You guys need to get out more. You are trapped in a world of logic, and unable to think outside that box.

You both seem committed to God’s eye view thinking, though you may be in denial over that. So you see perception as a system to report to you what is seen by the hypothetical God. But how could that ever work?

Imagine yourself in a two-dimensional world. You happen to know that it is a vector space full of directional vectors. So how are you going to go about discovering vectors to use for a coordinate system?

The answer is, that you cannot. So you do the best that is available to you. You invent vectors to use. You draw a straight line any direction that pleases you, and call that a vector. And that’s clearly a method of invention. Next, you draw another straight line at right angles to the first. And you call that another vector. That was also an invention. For luck, you try to do that a third time, but you cannot. You find out that the two vectors are enough to allow you to represent all of your two-dimensional world. Finding that out, I would count as discovery.

In short, the way to deal with the 2-dimensional world is by invention. And it happens that the world constrains what you can invent (you cannot have three mutually orthogonal vectors).

A lot of what is called “discovery” in mathematics and science is really invention that is constrained by reality (or constrained by axioms).

Now imagine that you are a developing biological organism, trying to find out about the world. You need to design a perceptual system. But it cannot be based on discovery, because you would need a pre-existing perceptual system for that. So the best you can do is rely on constrained invention.

Putting it all in perspective:

When I say that there is a human independent world, what I means is that there are strong constraints on what we can invent.

When I say that there are human independent properties, what I mean is that there are strong constraints on what properties we can invent.

I eschew metaphysics, because I see it as incompatible with the use of constrained invention as a form of discovery.


I came up with the expression “constrained invention” while composing that Yahoo groups post.

My thinking along those lines goes back to when I began to investigate (or theorize about) human learning, circa 1990.  I wanted to use the growth of science as an example of learning.  Roughly speaking, I took science to be human learning writ big.

What was clear to me, from the outset, was the dependence of science on the laboratory and on experiment.  And it was clear that this was not just the critical testing of theories as sometimes described as the hypothetical-deductive method.  Rather, the experimentation was an important part of scientific exploration and discovery.

When I compared this with AI research on machine learning, the experimentation seemed to be missing in AI.  Rather, learning in AI was conceived of as a search for patterns.  So AI was a logic program, looking for patterns in data.  But for scientists, there  often was no data.  They had to start by inventing ways to get data.  And AI was not doing that, nor was epistemology.

In retrospect, now that I have come up with the expression “constrained invention”, I see that as applying very well to the way the lab is used by scientists.  They come up with ideas (inventions), then test them in the lab (find how the world constrains those inventions).

%d bloggers like this: