Experimental Study On Mapping Process Of Analogical Reasoning

Analogical thinking becomes research focus on learning, problem solving, creativity and artificial intelligence now, which is one core ability of cognition development. The core subprocess in analogical thinking is analogical mapping. There is main three computational models which simulate the subprocess: the Structure-Mapping Engine(SME),the Analogical Constraint Mapping Engine(ACME),and the Incremental Analogy Machine(IAM).Keane et compared the three computational models using the Attribute-Mapping problem, but it is lack of validation psychological experiments. The study compared the three computational models from psychological experiments. To compare Keane et al.'s result, the research employed the serial reaction time task and tongue report way which used Attribute-Mapping problem all the same.This research consisted of two experiments. Experiment 1 set up two variable of meaning similarity and singleton order, required subjects to resolve the problem, and noted the reaction time and strategy. Experiment present subjects with two kinds of analogical story and require them the target problem, and noted the reaction time and strategy. The results were as follows:l.The reaction time of subjects have a trend of decrease with increase of meaning similarity item. But it isn't decrease of rule, which indicate: there is isn't significant difference between none similarity group and one similarity group, but there is significant difference between all similarity group and none similarity group, one similarity group. The results of SME obviously differ from our experiment results. To some degree, ACME and IAM forecast our experiment results, but there is still some difference. This indicates: The three computational models can't exactly simulate the analogical mapping thinking of human, but IAM looks better.2.The reaction time is little difference between two groups of order variable, but it isn'tsignificance. It is close to results of SME and ACME while it contrasts to results of IAM. It indicates: to some degree, SME and ACME is close to analogical mapping thinking of human, but IAM is large difference.3.The reaction time of the surface similarity group is short to that of the structure similarity group. This indicates: finding of the structure similarity is more difficult than finding of the surface similarity in the analogical-mapping story; the structure is the basic of the analogical problem.4.Result validated mapping is the key of the analogical problem from analysis of the subjects' strategy; moreover, SME is a pure arithmetic model of "try-wrong"; ACME is an interactive network of parallel process, which is large difference to analogical mapping thinking of human; IAM is closer to the analogical mapping thinking of human, but it still need be modified.