Two of the things I'm thinking about recently (in relation to my work) are visual agnosia and blindsight.
Blindsight is a condition where people report being blind but can solve tasks with obviously above-chance accuracy, e.g., going through a corridor and dodging obstacles. For me, this really isn't that philosophically confusing anymore; visual qualia is just one possible way to deal with visual inputs, and the idea that the brain also does it differently doesn't present any paradox. What exactly can you conclude from this? This is what I'm trying to figure out. But it's almost certainly relevant.
Philosophical commentary on blindsight tends to be dreadful. E.g., Dennett covered it in his book, and his rebuttal of Blindsight's philosophical significance was that if a Blindsight person were to be trained to volunteer guesses as to what they see (which is generally not the case; they don't realize that they have information about objects they see), and if you could slowly close the behavioral gap, then they would eventually claim to see like normal people. He presents no evidence for this, but putting that aside -- I actually agree with it, so evidence isn't the sticking point -- it doesn't show what he think it does. All it shows is that visual qualia is tied to a causal function, which I not only agree with, but which isn't even part of the debate; at that point in the sequence we'll have already established that all qualia are tied to causal functions.
The *difficult* question is whether visual qualia is (a) a certain way of processing visual information or (b) a by-product that hops on whenever you do visual processing of any kind. Dennett's argument does nothing to address this, and from what I've seen so far, neither does any other philosophical commentary. But Blindsight itself is certainly evidence for (a).
Conversely, Visual Agnosia is a condition where people feel like they can see normally but have certain clear deficits. It all seems to come down to the ability to integrate visual information. I sort of imagine it like if you looked at an object, say a glove, then you'd see many little parts of it sequentially rather than the entire thing at once. As a result, such people have difficulty identifying objects even though they see that something is there just fine. But they seem to be able to scan them sequentially and maybe deduce the answer; or they might pick something that resembles the object, like mis-identifying an apple for a small ball.
They're extremely bad with faces, which makes a lot of sense because facial recognition is actually extremely difficult and we just don't notice it because we're all so good at it. I think they generally can't even recognize themselves in a photograph, and they certainly can't tell whether someone is male or female. One case report said that the only person that the patient was able to identify among pictures on his wall was Albert Einstein because some feature of his hair or mustache gave it away.
If we differentiate between seeing, which is the part with visual qualia, and recognizing, which is the symbolic part where the brain determines the existence of objects, then Blindsight seems like (very limited) recognition without seeing, and agnosia seems like... seeing with limited recognizing? Seeing with zero recognizing? I'm not sure yet. It's still confusing to me to what extent seeing does things that aren't ultimately about recognition. Maybe they are ultimately about recognition, and it's more like that the brain builds this elaborate data structure to then do more efficient recognition based on that, as supposed to e.g. an image classifier based on a neural network, which does only recognition. In particular, a neural network moves to progressively more abstract representations every step and doesn't keep the pixel-based representation around.