Solving Geometric Analogy Problems Through Two‐Stage Analogical Mapping

Lovett, Andrew; Tomai, Emmett; Forbus, Kenneth D.; Usher, Jeffrey

doi:10.1111/j.1551-6709.2009.01052.x

Cited by 61 publications

(67 citation statements)

References 34 publications

(40 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It should be noted that, although those models of Lovett et al were first developed before 2010 [103,105,106], we find improved or more advanced versions of them from 2010 onwards [104,77,107], so it is somewhat arbitrary to include them in this subsection or in the following one.…”

Section: Later Systems: 2003-2009mentioning

confidence: 99%

“…Similar to Tomai et al, Lovett et al addressed three visual problem-solving tasks: geometric analogies [103,104], Raven's Progressive Matrices [105,77], and odd-one-out intelligence tests [106,107]. With a major goal of modeling human cognition, their models provide novel insights about which cognitive operations are easier or harder for human visual problem-solving.…”

Section: Later Systems: 2003-2009mentioning

confidence: 99%

“…With a major goal of modeling human cognition, their models provide novel insights about which cognitive operations are easier or harder for human visual problem-solving. Lovett et al demonstrated that qualitative spatial representations extracted by using CogSketch [108] (their sketch understanding system) and visual comparisons made by the structure mapping engine (SME) can be used to solve geometric analogy problems [99,103,104]. The authors combine these two models with two complementary theories of how people perform it: (1) visual inference, where the answer of a problem is inferred applying the differences found between the images "A" and "B" over the image "C" in order to obtain the solution "D"; and (b), secondorder comparison, where the differences found between the images "A" and "B" are then compared to the differences between "C" and each possible answer "i", selecting the most similar one.…”

Section: Later Systems: 2003-2009mentioning

confidence: 99%

“…sketches + SME + SEQL itemwise itemwise Lov+OOO Lovett et al [106,107] 2008-10 odd-one-out human cog. sketches + SME + SEQL itemwise itemwise Lov+GEO Lovett et al [103,104] 2009-12 geometric analogy human cog. Table 4: Taxonomy of computer models solving intelligence tests.…”

Section: -10mentioning

confidence: 99%

See 3 more Smart Citations

Computer models solving intelligence test problems: Progress and implications

Hernández-Orallo

Martínez-Plumed

Schmid

et al. 2016

Artificial Intelligence

View full text Add to dashboard Cite

ElsevierHernández Orallo, J.; Martínez Plumed, F.; Schmid, U.; Siebers, M.; Dowe, DL. (2016 AbstractWhile some computational models of intelligence test problems were proposed throughout the second half of the XXth century, in the first years of the XXIst century we have seen an increasing number of computer systems being able to score well on particular intelligence test tasks. However, despite this increasing trend there has been no general account of all these works in terms of how they relate to each other and what their real achievements are. Also, there is poor understanding about what intelligence tests measure in machines, whether they are useful to evaluate AI systems, whether they are really challenging problems, and whether they are useful to understand (human) intelligence. In this paper, we provide some insight on these issues, in the form of nine specific questions, by giving a comprehensive account of about thirty computer models, from the 1960s to nowadays, and their relationships, focussing on the range of intelligence test tasks they address, the purpose of the models, how general or specialised these models are, the AI techniques they use in each case, their comparison with human performance, and their evaluation of item difficulty. As a conclusion, these tests and the computer models attempting them show that AI is still lacking general techniques to deal with a variety of problems at the same time. Nonetheless, a renewed attention on these problems and a more careful understanding of what intelligence tests offer for AI may help build new bridges between psychometrics, cognitive science, and AI; and may motivate new kinds of problem repositories.

show abstract

Section: Later Systems: 2003-2009mentioning

confidence: 99%

Section: Later Systems: 2003-2009mentioning

confidence: 99%

Section: Later Systems: 2003-2009mentioning

confidence: 99%

Section: -10mentioning

confidence: 99%

See 2 more Smart Citations

Computer models solving intelligence test problems: Progress and implications

Hernández-Orallo

Martínez-Plumed

Schmid

et al. 2016

Artificial Intelligence

View full text Add to dashboard Cite

show abstract

“…This iterative strategy is used to extract all the relations contained in a given Raven item or as many relations as needed to narrow the number of possible responses. This process has been formalized in models of matrix reasoning (Lovett, Tomai, Forbus, & Usher, 2009;Carpenter et al, 1990).…”

Section: Introductionmentioning

confidence: 99%

Pupil Diameter Tracks the Exploration–Exploitation Trade-off during Analogical Reasoning and Explains Individual Differences in Fluid Intelligence

Hayes¹,

Petrov²

2016

Journal of Cognitive Neuroscience

View full text Add to dashboard Cite

Abstract■ The ability to adaptively shift between exploration and exploitation control states is critical for optimizing behavioral performance. Converging evidence from primate electrophysiology and computational neural modeling has suggested that this ability may be mediated by the broad norepinephrine projections emanating from the locus coeruleus (LC) [Aston-Jones, G., & Cohen, J. D. An integrative theory of locus coeruleus-norepinephrine function: Adaptive gain and optimal performance. Annual Review of Neuroscience, 28, 403-450, 2005]. There is also evidence that pupil diameter covaries systematically with LC activity. Although imperfect and indirect, this link makes pupillometry a useful tool for studying the locus coeruleus norepinephrine system in humans and in high-level tasks. Here, we present a novel paradigm that examines how the pupillary response during exploration and exploitation covaries with individual differences in fluid intelligence during analogical reasoning on Raven's Advanced Progressive Matrices. Pupillometry was used as a noninvasive proxy for LC activity, and concurrent think-aloud verbal protocols were used to identify exploratory and exploitative solution periods. This novel combination of pupillometry and verbal protocols from 40 participants revealed a decrease in pupil diameter during exploitation and an increase during exploration. The temporal dynamics of the pupillary response was characterized by a steep increase during the transition to exploratory periods, sustained dilation for many seconds afterward, and followed by gradual return to baseline. Moreover, the individual differences in the relative magnitude of pupillary dilation accounted for 16% of the variance in Advanced Progressive Matrices scores. Assuming that pupil diameter is a valid index of LC activity, these results establish promising preliminary connections between the literature on locus coeruleus norepinephrine-mediated cognitive control and the literature on analogical reasoning and fluid intelligence. ■

show abstract

Computational models of analogy

Gentner¹,

Forbus

2010

WIRES Cognitive Science

123

119

View full text Add to dashboard Cite

Analogical mapping is a core process in human cognition. A number of valuable computational models of analogy have been created, capturing aspects of how people compare representations, retrieve potential analogs from memory, and learn from the results. In the past 25 years, this area has progressed rapidly, fueled by strong collaboration between psychologists and Artificial Intelligence (AI) scientists, with contributions from linguists and philosophers as well. There is now considerable consensus regarding the constraints governing the mapping process. However, computational models still differ in their focus, with some aimed at capturing the range of analogical phenomena at the cognitive level and others aimed at modeling how analogical processes might be implemented in neural systems. Some recent work has focused on modeling interactions between analogy and other processes, and on modeling analogy as a part of larger cognitive systems. WIREs Cogni Sci 2011 2 266-276 DOI: 10.1002/wcs.105 For further resources related to this article, please visit the WIREs website.

show abstract

Solving Geometric Analogy Problems Through Two‐Stage Analogical Mapping

Cited by 61 publications

References 34 publications

Computer models solving intelligence test problems: Progress and implications

Computer models solving intelligence test problems: Progress and implications

Pupil Diameter Tracks the Exploration–Exploitation Trade-off during Analogical Reasoning and Explains Individual Differences in Fluid Intelligence

Computational models of analogy

Contact Info

Product

Resources

About