Modeling Multiple Strategies for Solving Geometric Analogy Problems
Andrew Lovett, Kenneth D. Forbus · eScholarship (California Digital Library) · 2012
We present an improved computational model for performing geometric analogy. The model combines two previously modeled strategies and makes explicit claims about when people will use one strategy or the other. We compare the model to human performance on a classic problem set. The model’s strategy shifts, along with working memory load, account for most of the variance in human reaction times.