Why are human animacy judgments continuous rather than categorical? A computational modeling approach