Is it true that properties such as cuteness do not correspond to anything? In a sense it is false to deny that any such correspondence exists: such properties do correspond to the cuddle-provocativeness of a baby, the eating-provocativeness of an apple, etc., as a cause of the behavior of agents. They are “lovely” properties (Dennett 1991, p. 379), and there is a way to measure them: we can use ourselves as detectors. But the reason we, intuitively, do not accept a robot as a subject like ourselves is because we know how the robot does it: we know that it calculates, maybe even in a PP-manner—we know that it does not react directly to the properties that seem to exist and that seem to count. Neither do we, or the beings described above. But their own prediction of themselves treats such complex properties as simple, because there is nothing to be gained by being more precise than is necessary for sufficiently accurate prediction.[16]
This is my reconstruction of Dennett’s claim that the mind projects its dispositions onto the world via Bayesian prediction. I want to draw attention to some of the features ascribed to those properties that this story predicts:
These properties are “given directly” to a person
The overall generative model depicts the whole organism as a unified object that reacts directly to the posited properties in the world. Any system that represents itself in such a way is bound to believe that there are properties of the world given directly to the object, which it takes to be itself. In subpersonal terms this object and these properties, as well as their relation to each other, are postulated entities that explain the sensory input. For instance, the fact that others talk about the system as someone with beliefs and desires (which is rooted in the same principle) can be explained by predicting itself in the same way.
These properties are irreducible to physical‚ mechanical phenomena.
Since the generative model does not depict these properties as built up from simpler ones, but simply posits them to predict lower-level patterns, these properties don’t seem (to the system) to be reducible to other properties.
These properties are atomic‚ i.e., unstructured.
There are as many posited properties as there are distinct dispositions to be tracked. This also explains why one can learn to find structure in formerly unstructured qualia (cf. Dennett 1991, p. 49) once new discriminative behavior is learned.
These properties are important to our lives/beings as humans/persons
This felt importance is obvious, given the putative role they play in the explanation provided by the generative model. These properties seem to be the causes of all our behavior: if one did not feel the painfulness of a pain, one would not scream; if one did not sense the funniness of a joke, one would not laugh, etc. Since the model is still needed for interacting with others, despite theoretical advances in the sciences this felt importance of qualia to our lives is very difficult to overcome.
These properties are known to every living human being; it is not possible to sincerely deny their existence
This is due to the fact that our brains predict the behavior of others via a model that posits direct interaction between “agents” and first-order, non-relational object properties—the entities that are then named “qualia”.
This list has considerable overlap with lists of features ascribed to qualia (e.g., Metzinger 2003, p. 68; Tye 2013), lending support to the thesis that we don’t need a revolution in science to accommodate qualia, but rather a change in perspective: we might look at the creatures described above and see that “[t]hey are us” (Dennett 2000, p. 353).