I'm working my way through these. The first PDF seems to be the most substantive one. But it is also the one that shows his overall thesis isn't as strong as he says it is.
I still don't have a firm opinion, but here are some first impressions.
1) If we live in 'The Matrix', then all perception is only interface and there really is nothing else there. There is no way to tell if this is the case.
2) His model is very simplistic and his terminology is slightly misleading.
For example, he gives a situation (which he later generalizes) where there is only food and water. Fitness is originally determined by the sum of the food and water variables. This changes later.
In this case, his 'truth' perception is one that knows the values of food and water and the 'simple' perceptual system is the one that ranks only one of the two: in essence, it gives a thumbs up if the amount of food or of water is above a threshold.
He also makes an adjustment so that determining the 'truth' takes more energy than the 'simple' system.
What he finds is not too surprising: if the cost of determining the truth is low, the truth perceptual system wins. But if the cost of determining the truth is high, the simple system wins. Essentially, if determining the truth is difficult and there is a rule of thumb that will work, the selection goes to the rule of thumb. But if finding truth is easy, it goes to truth.
This, by the way, is NOT what he claims in other papers. But this one is the one that gives a detailed analysis to prove what happens.
-----
Next, he modifies the fitness function so that both high and low levels of food or water are less 'fit' than medium levels. So, the organism doesn't want to drown or get buried by the food. The most fit situation is where there is some food and water but not too much of either.
Now, he adds another type of perceptual system: the interface system. While the 'truth' knows both the actual amounts of food and water and 'simple' knows gradings of both, these are rigged so that higher levels are seen as better in both systems (what he calls having a homomorphism). The interface system judges the intermediate levels as worthy of perception and less so for the extremes.
Well, again, the result is not too surprising. If the 'cost' of finding the 'truth' is small, the truth wins out. If it is high, the 'interface' version wins out.
-----------------------------------
This is extended to include more than two environmental variables and to allow those variables to not be independent. Similar results are found.
-----------------------------------
OK, a couple of comments.
1. it seems like his 'truth' vs 'simple' vs 'interface' are poorly chosen names. Even the interface version has a correct analysis of the value of the environment to the organism. This is to be expected to promote fitness.
2. In many ways, this seems to be a better description of 'awareness' than of 'perception'. if we pay attention to those things that promote fitness, we are more likely to survive. if finding the 'real story' is costly, a rule of thumb may well be better for survival.
3. In those cases where truth is easy to determine, the system *does* promote truth finding. So, where processing is relatively easy and doesn't take much energy, we can expect the system to promote truth rather than interface.
4. Since both his 'simple' and 'truth' systems are rigged to favor high values in the environment, it isn't surprising they fail when the fitness is to the medium values. But even there, the system correctly perceives the fitness level.
5. We *know* our visual system (for example) takes many shortcuts: it picks out edges, faces, and draws attention to motion. The first and last are easy to analyze and are actually done in th eye and not the brain. So we can expect that these perceptions are correct according to his analysis.
6. His analysis doesn't consider the possibility that correct detection of fitness conditions will, as a side-effect, promote correct detection of some other aspects of the environment. So, for example, we detect redness in fruit in part because it is a sign of ripeness. But that means we are likely to correctly detect redness in other situations simply because it is costly to detect red in one case and not in the others. So, even in cases that are NOT promoting fitness, we can expect that redness is often correctly perceived. The same is true for other easy perceptions (those not requiring excessive energy to determine).
7. We *know* our visual system often 'fills in the gaps' where there is no actual sensory data. So, for example, we all have a blind spot, but we seldom are aware of it because the system 'paints in' the missing information. This is 'non-truth' that is fabricated by our visual system.
8. We know of many other deficiencies of our visual system (I focus on this because it is more extensively studied--many of the same things can be said for other senses). So, for example, we only have three color receptors (some birds have up to seven). This means that we often see two things as having the same color even though they are, in fact, quite different physically (i.e, in reality).
9. This also ignores the simple fact that we cannot see many forms of light at all. Ultra-violet and infra-red are beyond our abilities to directly see. Similar things can be said about sounds.
10. The fact that we can determine such deficiencies in our sensory systems shows that we don't *only* have interface: we can and do often have information about reality.
---
So, my basic conclusion is that there is an important point here: our visual sensory systems are not *guaranteed* to give reliable information. In fact, in many situations we can *expect* them to fill in gaps, use rules of thumb, and minimize energy expenditure. None the less, we can expect those aspects where truth is easy to get and where such is linked heavily to survival to be accurate and for this accuracy to spread out to similar inputs.
I think claiming it is *all* interface is vastly over-stating the evidence he presented and even contradicting some of what his model shows.
