A Model for Understanding Recognition Validity

Abstract

In most decision-making domains, people recognize some stimuli but not others. The validity of recognition as a basis for making decisions about the stimuli is empirically unclear, but has strong implications for the development of theories and models of decision-making. Following the framework used to motivate the recognition heuristic (Gigerenzer and Goldstein, Psychological Review, 103(4), 650–669, 1996), we identify two possible sources of failures in the connection between the signal provided by recognition and the criteria by which stimuli are judged. One is a failure of sending, in which the mediating information environment does not provide recognition signals that follow the criterion. The other is a failure of receiving, in which there are individual differences in the way people use environmental signals to learn to recognize stimuli. We test these two hypotheses, with a new model of individual-level recognition based on the concept of Guttman scales and the Rasch model from psychometrics, using seven new data sets. The data sets cover various real-world domains, including domains involving prediction, and span a range of levels of overall recognition and correspondence between recognition rates and criterion values. We show that the model is able to describe the patterns of recognition across people and stimuli very well and thus provides a capability to measure the noise inherent in sending and receiving recognition signals. These modeling results lead to a clear conclusion that, across all of the data sets, there are at best small individual differences in receiving environmental signals, but often large differences between recognition rates and criterion values in the signals the environment sends.