Anna Koop

Despite a few technical snags, my proposal is in the hands of my committee members and thus beyond my control for the next two months. Time to shift gears and focus on paper-units for a while.

Since I put up a draft of my abstract, I thought I should go for broke and post my proposal as well. It’s shifted a bit in focus . . . lemme see if I can summarize it (this version has no abstract):

Knowledge. Kinda important issue for artificial intelligence and cognitive science. But a slippery beast—we don’t really know that much about what it could or should or might be. And one particularly intriguing question is how the detailed, transient, particular signals of an intelligent agent’s sensorimotor experience are related to the abstract, stable, and general summary information that we call knowledge. This is the aforementioned problem of knowledge and data.

Now, in artificial intelligence research, there’s been a lot of work on the representation of knowledge. In particular what kind of structure should be used for representation and what should go in it. Then, given different kinds of representations, how it is grounded (or given meaning) and how the content might be verified (and what it means to be true in the first place). That research has been helpful and interesting. But there’s an even-more-basic question about knowledge representation, which is what the knowledge is about—what does the representation represent? This choice about what knowledge refers to has consequences.

So although we’ve spent lots of time on figuring out the problems and advantages of different representational schemes, there hasn’t been as much talk about different referents. But I think it matters, for grounding and verification and usability. In the proposal I briefly talk about the differences between taking an objective stance, which says knowledge is about the objects and laws of the physical world, and taking an empirical stance, which says knowledge is about patterns in sensorimotor data.

That’s the setup. The actual thesis work I’m proposing has two parts. First, I want to analyze what we’ve done in AI for knowledge, particularly with respect to how knowledge and data interact. What are the strengths and weaknesses of different choices of referent and approaches to the problem of knowledge and data? After getting a clear handle on what we’ve got, I want to see what I can do. I want to implement a predictive representation specifically for general knowledge representation and see if our various tools for abstraction can actually turn around some of the current weaknesses in empirical representations.

That’s the gist. It’s more carefully laid out in the proposal. Let me know what you think! Love it? Hate it? Reserving judgement over whether or not this is actually a comp sci thesis? After incubating the ideas for ages I’m looking forward to hearing what people think!

I’ll be giving a practice candidacy talk in a Tea-Time-Talk soonish. Patrick will keep us all posted…

The Problem of Knowledge and Data (pdf)

A random thought, from reading this article on cheating (Mike’s fault)
and bumping into this quote: “Propositional calculus is a system for deducing conclusions from true premises. It uses variables for statements because the logic works regardless of what the statements are.”

Which is standard stuff but it struck me that the whole problem of this view is the problem of definition (Plato’s problem in the Margolis and Laurence survey).

Logic works regardless of what the statements are as long as the variables mean what you wanted them to mean. “If P, then Q. P, therefore Q.” This is true so long as the entities you want to sub in for P and Q can properly take a true or false value. So we get told modus ponens as if it’s “this is a universal truth” and well, it’s more like 1+1=2, isn’t it? That *can* be one of the universals. Doesn’t have to be (dangit, I have to read up on Gödel one of these days). And even so, it rather hinges on the definition of 1 and 2. One cup of water and one cup of sugar doesn’t make two cups of anything.

Reading further (the article’s quite interesting) I see I’m not alone in this wait-a-minute reaction and now I have to look up the Wason selection task and David Buller’s critique of it.

Anyway, apparently, “meaning matters” is going to be my new hobby-horse.

Random picture test:

Just watched a brilliant Ted talk by Daniel Kahneman: The riddle of experience vs. memory.

The upshot: there’s a difference between your experiences and your memory of your experience. So much most people know already. But it has ramifications far broader than we realized. Your experience determines your transient happiness or well-being. The story you tell yourself determines your long-term satisfaction. Probably this relates to Seligman’s distinctions in kinds of happiness: the pleasant, engaged, and meaningful life.

He has a simple example in the talk: colonoscopies. Used to be quite painful. Patient A had a quick one that ended on a high-pain note. Patient B had the same high-pain but the treatment went on longer, ending in middling pain. Guess who had a better memory? Patient B. Because the ending is the part that sticks with you. This matches Dan Ariely’s findings that pulling off a bandage slower is better. We remember intensity more than duration.

I’ve been toying lately with the idea that the conscious self (glances around quickly to see if Rich is watching) is more like the story, the construction or projection we make from our experience. This gets tricky to talk about because of course you assume I mean the conscious “I” when I use first or second person. Oh well.

In these terms: we have our sensorimotor experience and our mind makes of that what it will. Part of what our mind makes of our sensorimotor experience is the elaborate explanations for it, including ideas about chairs and tables and “I”. “I” am the remembering self, not the experiencing self. The experiencing self has a transient and dynamic existence. But it does inform the remembering self, of course. They’re just not mapped together exactly.

Something like “I think, therefore I am; my agency experiences, therefore it is.”

So who is the boss? Kahneman makes the point that the experiencing self makes a lot of sacrifices on behalf of the remembering self—three weeks of vacation for a few hours of memories spread over a lifetime? On the other hand, when we pursue pleasure over purpose we’ve flipped those priorities around. So probably the classic: It. Depends.

People probably don’t want to think of themselves as emergent. But being emergent doesn’t mean less real than being constructed directly. Nor less important.

Iteration eleventybillion of my proposal draft. Comments of all kinds welcome. I think I need to support some of the statements therein and talk up some of the “why”s but I’m not sure how much is needed in the abstract (or if proposal really should have abstracts)

-=-
Update: proposal toned down in scope and claims, oddball definition of data disposed of, algorithm component added back (extending what work we have on empirical knowledge representation to identified areas of interest).

No abstract in the new version but I’ll post a summary soon.
-=-

The problem of how to represent general knowledge in artificial systems remains unsolved. There have been many different approaches to knowledge representation, but these approaches are difficult to compare. No universally satisfactory solution has been found.

For the first part of my thesis research, I aim to analyze the strengths and weaknesses of a broad range of approaches to knowledge representation. I will use an inclusive definition of knowledge as the general, abstract, stable stuff of the mind. This allows me to consider a wide range of knowledge representation: from the logical knowledge bases of good old-fashioned AI to the models of control theorists and discriminative functions of supervised learning. An analysis of knowledge representation that takes such an inclusive stance is rare, as research generally focuses on the fine points of representational detail within a subfield.

I will be investigating the problem of knowledge and data: how the content of knowledge should be related to the data of sensorimotor experience. Existing analyses of knowledge representation frameworks focus on differences in structure and ignore differences in semantics: what the knowledge is meant to represent. I will be comparing knowledge that is concerned with representing objective reality, which is by far the dominant approach in artificial intelligence, to knowledge that is concerned with representing patterns in sensorimotor experience, a relative newcomer in AI.

I hypothesize that both of these approaches to the meaning of knowledge have distinct practical benefits. Knowledge about objective reality lends itself to general, abstract, and stable content, which I have given as definitive characteristics of knowledge. At the same time, knowledge about objective reality seems to require an external source of data for grounding and verification. Knowledge about empirical experience should have an easier task in grounding and verification, being about internally accessible data. However, constructing general, abstract and stable content from the ephemera of sensorimotor signals seems problematic.

Having completed the analysis and clearly identified the strengths and weaknesses of these two semantic approaches, I will propose developing an empirical representation that allows for the construction of general, abstract, and stable concepts. This will build on previous work in my Master’s thesis and provide a strong foundation for the emerging field of empirical knowledge representation.