Authors:
Ronald W. Ferguson
Department of Computer Science
Northwestern University
Evanston, IL 60201 USA
Dedre Gentner
Department of Psychology
Northwestern University
Evanston, IL 60201 USA
Please
address correspondence to: Ronald
Ferguson
(ferguson@ils.nwu.edu)
Abstract
may be considered for PAPER or POSTER.
STRUCTURED
ABSTRACT (for publication)
TITLE: Analogical encoding: A framework
This
paper reviews an emerging area of analogical reasoning, called
analogical
encoding. Analogical encoding
encompasses analogical
processes
occurring during representation, as opposed to processes
involving
the comparison of independently represented items or memory
access. Research on analogical encoding has demonstrated
several
interesting
cognitive phenomena, but these phenomena lack a common
framework,
which this review provides. Current
work on analogical
encoding
includes computer models and psychological experiments
exploring
symmetry detection as a form of analogical mapping, studies
of the
relative effects of concurrent versus sequential presentation
of
business cases, and models of the use of analogy in
juxtaposition-based
diagrams. The wide variety of effects
within
analogical
encoding make a categorization of such processes critical
to this
emerging field of study. It is also
critical to understand
how
analogical encoding relates to known effects in high-level
perception
and transformation-based similarity models.
Our review
contrasts
all these approaches, placing them in a common framework.
In the
end, we argue that analogical encoding is a kind of analogical
processing
at the same status as other analogical reasoning stages,
such as
analogical comparison or access, and provide a description of
analogical
encoding on that basis.
LONG
ABSTRACT (for review)
Analogical
encoding: A framework
This
paper reviews an emerging area of analogical reasoning, called
analogical
encoding. Analogical encoding
encompasses analogical
processes
occurring during representation, as opposed to processes
involving
the comparison of independently represented items or memory
access. Research on analogical encoding has
demonstrated several
interesting
cognitive phenomena, but these phenomena lack a common
framework,
which this review provides.
Several
studies show an interaction between analogical reasoning and
representation. Loewenstein, Gentner and Thompson (in press)
investigate
the effects of analogical encoding on subjects'
representation
of concurrent cases in negotiation.
Subjects were
given
two analogous cases demonstrating a particular principle, and
told
either to write out the commonalities (the analogical encoding
condition)
or to critique each case separately.
The analogical
encoding
group developed a far better understanding and transfer of
the
common principle.
Using a
different approach, Ferguson (1994), argues that symmetry
detection
may be a form of analogical encoding, involving a
specialized
form of analogical mapping. This
specialized analogical
mapping,
called a self-similarity mapping, detects symmetry in a
single
description. Ferguson's MAGI model
detects symmetry using a
structure-mapping
algorithm over low-level perceptual structure,
finding
maximal systems of self-similarity within a single
description. MAGI has successfully simulated several key
characteristics
of symmetry detection, including the detection of
qualitative
symmetry (Ferguson, Aminoff & Gentner, 1996) and the
preference
for vertical symmetry, and can detect non-visual forms of
symmetry,
such as symmetry in narratives.
Finally,
Ferguson and Forbus (1998) show how analogical encoding
techniques
support diagrammatic reasoning. In
their JUXTA system,
repetition-detection
is combined with correspondence-based differences
to
label and critique a juxtapostion-based diagram illustrating simple
physical
laws.
The
variety of effects within analogical encoding make a
categorization
of analogical encoding processes critical to this
emerging
field of study. It is also critical to
understand how
analogical
encoding overlaps with known effects in high-level
perception
(Hofstadter and Mitchell, 1992), and classic
transformation-based
models (Evans, 1968).
Our
review contrasts these approaches, placing them in a common
framework. We argue that analogical encoding is a kind
of analogical
processing
at the same status as other analogical reasoning stages,
such as
analogical comparison or access (Gentner, 1989; Hall, 1989).
We also
describe how analogical encoding makes the comparison process
more
effective.
References
Evans, T.G. (1968). A program for the
solution of a class of
geometric-analogy
intelligence test questions. In Minsky
(ed.),
Semantic
Information Processing. Cambridge: MIT Press.
Ferguson, R.W. (1994). MAGI: Analogy-based encoding using symmetry
and
regularity. Proceedings, 16th Cognitive
Society Conference.
Atlanta,
GA.
Ferguson, R.W., Aminoff, A., and Gentner,
D. (1996). Modeling
qualitative
differences in symmetry judgments. Proceedings, 18th
Cognitive
Society Conference.
Ferguson, R.W. and Forbus, K.D.
(1998). Telling juxtapositions:
Using
repetition and alignable difference in diagram understanding.
International
Workshop on Analogy, Sofia, Bulgaria.
Gentner, D. (1989). The mechanisms of
analogical learning. In
Vosniadou
and Ortony (Eds.), Similarity and Analogical Reasoning.
London:
Cambridge University Press.
Hall, R. P. (1989). Computational
Approaches to Analogical
Reasoning.
Artificial Intelligence, 39, 30-120.
Hofstadter, D. L., & Mitchell, M.
(1992). An overview of the
Copycat
project. In Holyoak & Barnden (Eds.), Connectionist Approaches
to
Analogy, Metaphor, and Case-Based Reasoning. Norwood, NJ: Ablex.
Loewenstein, J., Thompson, L., and Gentner,
D. (In press).
Analogical
encoding facilitates knowledge transfer in negotiation.
Psychonomic
Bulletin & Review.