Before elaborating on some of Peter Dlugos's useful points, let's quickly summarise where we were. Recall that the original idea was to examine how electronic philosophy will change the way we do philosophy, not just its methods but its content. My original point was a simple one - that for most of this century, analytic philosophy has restricted itself to propositional modes of representation, but that this is far too confining and electronic ways of doing philosophy will help us break out of those modes. As Dlugos and Fetzer have (implicitly) noted, this issue is part of a much broader concern, which is the way that linguistic representations have led to strange, if not absurd, consequences. I'm happy to pursue this wider issue, as long as we keep the original points in mind. This is, after all, a forum on electronic philosophy.
Point 1.: Dlugos puts many of his remarks in the context of representations. I would warn him against trying to break away from the standard ways of doing things whilst retaining the standard way of construing things. Contrary to what so many philosophers seem to believe, representation is not the key problem in philosophy. It's a problem all right, but it's a mistake to think that every issue in this area involves representations, or that even when they do, that representational aspects are essential to the phenomena. For example, laboratory experiments need not be construed as representations of anything. They are bits of the world, standing for nothing beyond themselves. As such, they can provide us with causal knowledge. Moral: Putting the issue in terms of representations entails that there is something else than the representation - the represented - and that we therefore have to solve the problem of the relation between the representation and the represented. Not so.
Point 2.: The fact that we can (let us concede here, although I think this is false) represent all scientific knowledge propositionally, does not mean that we have to, or that we do when we are acquiring that knowledge. The business of ``rational reconstruction'' in philosophy can be so much attuned to ideally rational epistemic agents that it becomes completely unrealistic as a model of human epistemology.
Point 3.: It's not just that we can avoid propositional representations of problems and solutions in this area, but that we often should. Here's one reason from a related area. It is sometimes said that if F, G are properties, then so is (F&G), their conjunction. Why? Because properties, it is said, are closed under logical operations. Why again? I've heard it claimed either a) it's obvious that they are, or b) (better but still not good) that the predicates that represent properties are so closed and so the properties had better be. Thus you get the existence of properties such as being a round square, and the escape clause that such properties are (necessarily) not instantiated. If you like logical smoothness of your representations, this is o.k. Ontologically, it's a non-starter - representations ought to conform to what there is, not vice versa. One would not insist that contingent features of the world, such as the causally efficacious properties, be closed under conjunction, for example. This is one clear example of how propositional representations can get us into philosophical trouble. Standardly, propositional representations lead us into the whole of logical space, and operations can be performed in logical space that simply have no counterparts in the world, or even in what you might call ontological space. Goodman's ``grue'' problem, Putnam's anti-realism argument, and the views that David Lewis was trying to escape by his use of ``perfectly natural properties'' are just three well-known examples of how elevating logical simplicity over almost all else leads to downgrading our other forms of access to the world.
Point 4.: Icons are a good start. You want to know who was at a meeting. Someone wordlessly hands you an analog photograph. You see, non-propositionally, who was there. Where's the propositional content in the answer? In your language of thought, maybe? Not if you believe the preponderance of current evident that facial recognition is not done in classical computational terms, even less so if you believe it to be a biological phenomenon. Again, you want to know the structure of erosion in brains in victims of kuru? Don't describe it, look at the CAT scan. Interested in the turbulent flow within intergalactic jets? Try a linguistic description and you'll soon go over to what all workers in the field use, which is graphical visualisation techniques. The structure of none of these icons is propositional.