My central area of specialization is cognitive science. Currently in my research I am focusing on the role of representation and the development of concepts in cognition, especially trying to map these ideas onto recent understanding in neuroscience and cognitive psychology.
Book — On Mental Representation
Advances in neuro-science and improvements in AI technologies are making it increasingly clear that our traditional, vague, understanding of mental representation must be replaced by a more rigorous account. In this book, I develop precisely this. To begin clarification, we need to distinguish between three different senses of mental representation: particular mental representations; individual mental concepts; and, idealised linguistic concepts. The first step along the way is to debunk the ubiquitous assumption that neurons or clusters of neurons function in the brain as representations. I argue, along with a few others (e.g. Andy Clark, Fred Keijzer) that we need to adopt an integrated approach to theorising about representations in which we use both dynamic systems theoretical tools of analysis alongside the more standard reductive ones. Once we do that, we’ll see that conceptualising neurons as representors/detectors is not only inaccurate, but motivates misguided and unfruitful research paths and closes the door to theoretical advance. In developing these ideas, I draw on a large multi-disciplinary body of literature.
Say it Well
We’re adapting Luis von Ahn’s human computation ideas for language translation to the problem of translation between natural language and CycL (the often opaque and complex logic-based language that the knowledge of Cyc is encoded in). On the model of the ESP Game we will develop an on-line game that should prove entertaining to a wide spectrum of gamers. In von Ahn’s version, the result of game-play is a database of appropriately tagged photographs. In ours, the result of game-play will be a database of CycL/Natural language pairs — CycL rules and statements tagged with the correct natural language translations.
Such a database would serve two important purposes:
- It would allow logical sentences in Cyc to be translated into Natural Language for human consumption; and,
- It would allow Cyc to understand the information it gathers from the wild.
One clear use for this would be the creation of a “Semantic Wikipedia,” where the information contained could be used not only by people, but also by machines for automatic inference.
Vague and Unsupported Ideas
I’m interested in getting a better understanding of how case-based reasoning (CBR) is an entirely different kind of reasoning from rule-based reasoning (RBR). My hunch is that CBR is more fundamental then RBR, as a form of reasoning, and is possibly pre-representational. Our capacity for episodic memory, what makes mental representation even possible on my view, is in fact a kind of CBR. Contact me if you are interested in similar ideas.