[ home | resume | contact | science | art | personal | email ]

---

Science of Imagination (General)

• Title: A Vision for the Science of Imagination
  Abstract:
  Imagination is a crucial process for hypothetical thinking, planning, dreaming, counterfactual thinking, and creativity. In this talk I will present how imagination can be studied scientifically, and the various endeavors we are currently pursuing in the Science of Imagination Laboratory in the Institute of Cognitive Science at Carleton University.

Artificial Intelligence

• Title: Visuo: Quantitative Estimation of Spatial Magnitudes Using Analogical Reasoning
  Abstract:
  Visuo is an implemented Python program that models visual reasoning. It takes as input a description of a scene in words (e.g., “small dog on a sunny street”) and produces estimates of the quantitative magnitudes of the qualitative input (e.g., the size of the dog and the brightness of the street). We claim that reasoners transfer quantitative knowledge to new concepts from distributions of familiar concepts in memory. We also claim that visuospatial magnitudes should be stored as distributions over fuzzy sets. We show that Visuo successfully predicts quantitative knowledge to new concepts. Based on work done with Jonathan Gagne.
• Title: Modelling English Spatial Preposition Detectors
  Abstract:
  When shown a picture or asked to describe a scene people will use spatial prepositions to identify relations between objects. In this paper we present five algorithms for the detection of spatial relationships within an image. We discuss how best to represent human intuitions about spatial relationships using fuzzy ‘belief’ values and how such representations can be modeled after English spatial prepositions. The relationships modeled are above/below, adjacent to, occlusion, between, and close to.
• Title: Constructive Adaptive Visual Analogy
  Abstract:
  I present Galatea, an implemented LISP program that implements adaptive problem-solving through transfer from distant analogs. It is the first program to transfer procedural problem-solving solutions using only visual information. I show that visual similarity can provide a useful means to access otherwise semantically-unrelated solutions. I present how the program works, how it models the results of two experiments, and its psychological plausibility.

Philosophy

• Title: The Cognitive Importance of Testimony
  Abstract:
  As a belief source, testimony has long been held by theorists of the mind to play a deeply important role in human cognition. It is unclear, however, just why testimony has been afforded such cognitive importance. We distinguish three suggestions on the matter. According to the number claim, testimony’s cognitive importance is the consequence of human cognitive agents typically acquiring at least as many beliefs from testimony as from any other distinct belief source. According to the reliability claim, testimony is cognitively important because it is at least as truth-conducive as any other source. According to the scope claim, testimony is cognitively important because, as least as much as any other source, it affords us the ability to represent significantly more of the world than we would be able to in its absence. After laying out these three suggestions, we go on to argue that there is little hope of grounding testimony’s cognitive importance in either the number claim or the reliability claim. We conclude with a tentative exploration of the basis and plausibility of the scope claim. Based on work done with David Matheson.
• Title: The Microtheory Model of Belief Contamination
  Abstract:
  When imagining fabricated or hypothetical beliefs, we almost never use them to infer things about the real world. Somehow these beliefs (say, about future imagined events, or things that are true in “The Fellowship of the Ring”) are kept functionally separate from the beliefs we hold about the real world. It has been suggested that such beliefs are kept in quarantine in a “pretense box.” Drawing on solutions to this problem used in AI (particularly the CYC project) we propose that this theory is insufficient, and offer our own, the microtheory model (MTM) of beliefs, which holds that each belief is tagged with the microtheories in which it is true. This is philosophically-oriented talk based on a paper in preparation, co-authored with Jeanette Bicknell.
• Title: Task-Based Distributed Cognition
  Abstract:
  The limits of the extended mind hypothesis preclude it from including distributed cognitive systems without human beings, such as ant hills, computers, and brain areas. Intracranialist views suffer the same problems. This paper describes the task-based theory of distributed cognition, which solves these problems, providing a comprehensive framework for better defining the bounds of cognition in any conceivable system, at several levels of abstraction. I evaluate the theory by comparing anthropological data of a biomedical engineering laboratory and human mental imagery.
• Title: When You Need To Be Nice To Your Robot: A Fuctional Proposal for Pain Description
  Abstract:
  I examine the limits of current ethical frameworks as they relate to artificial intelligences (AIs). As AIs become more sophisticated, they may be capable of experiencing trauma, which means that there will be a need to treat them with ethical consideration. This talk will demonstrate the need for an ethical framework, summarize existing paradigms and their limitations with regard to applications to Artificial Intelligence research, and conclude by recommending functional criteria on which to base ethical considerations, which will, in turn, allow the boundaries moral community to extend to include some AIs. Work done with Xander Miller.

Education

• Title: Don't Waste Student Work: Using Classroom Assignments to Further Wider Educational Goals and Research Abstract:
  Every year college and graduate students across North America work on millions of assignments, and instructors and teaching assistants spend millions of hours grading them. These assignments help the students learn, but have no benefit for anyone else. In his talk, Dr. Davies will describe several kinds of assignments he has developed that he believes:
  • facilitate learning
  • are particularly motivating
  • contribute to the greater educational and research communities.
  Any discipline can make use of the methods Dr. Davies will describe. You can see an older version of this talk and download an associated techinical report at http://jimdavies.org/research/publications/technical-reports/Davies2009.html