Colour constancy, or in general constancy of any sort, is an untenable proposition. Our guessing meatware sausage is likely the source of this fundamental instability. We can only guess based on memory prior constellations.
We can probably lens this as a min-maxing of sorts. For example, a disc through a continuous veiling lamina filter yielding a chromatic cognition is no different to the “surface” of a diffuse object redirecting a cognitively unbiased energy source.
They are equivalent.
As such, the general idea that our visual system tries to reach into “reality” and deduce these external properties seems problematic.
Instead, the idea that cognition is a looping puzzle solver seems more tenable.
Why?
Because if the “information” is outside of us, it is fixed and rigid. Conversely, the framing of “a puzzle” implies a specific perspectival need-oriented solution, subject to the organism context.
From an art vantage, produced craft is a proposal; a proposed contemplation based on a given perspectival problem.
Cubist work, for example, cannot be merely an assemblage of square-like forms, but rather something forcing the primate to contemplate in a broader context.
The idea that there is always a veridical “information” that exists outside of us betrays us as conscious creatures, with the capacity to manipulate the phase structure of our memory priors.
This became even more apparent to me in my most recent post on hg2dc, where I attempted to demarcate a difference between a state and an operator.
The chosen operators are the generator of meaningfulness, relative to the given contextual need.
A hammer can exist as many different operators, from tool to weapon to a potential energy source to hold down your writing in the wind.
There is no veridical “information” in the hammer, but rather the informational meaning is driven by operations.
troy_s 
ino