AUTHOR=Miyazaki Masato , Ishikawa Ken-Ichi , Nakashima Ken'ichiro , Shimizu Hiroshi , Takahashi Taiki , Takahashi Nobuyuki TITLE=Application of the symbolic regression program AI-Feynman to psychology JOURNAL=Frontiers in Artificial Intelligence VOLUME=6 YEAR=2023 URL=https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2023.1039438 DOI=10.3389/frai.2023.1039438 ISSN=2624-8212 ABSTRACT=

The discovery of hidden laws in data is the core challenge in many fields, from the natural sciences to the social sciences. However, this task has historically relied on human intuition and experience in many areas, including psychology. Therefore, discovering laws using artificial intelligence (AI) has two significant advantages. First, it makes it possible to detect laws that humans cannot discover. Second, it will help construct more accurate theories. An AI called AI-Feynman was released in a very different field, and it performed impressively. Although AI-Feynman was initially designed to discover laws in physics, it can also work well in psychology. This research aims to examine whether AI-Feynman can be a new data analysis method for inter-temporal choice experiments by testing whether it can discover the hyperbolic discount model as a discount function. An inter-temporal choice experiment was conducted to accomplish these objectives, and the data were input into AI-Feynman. As a result, seven discount function candidates were proposed by AI-Feynman. One candidate was the hyperbolic discount model, which is currently considered the most accurate. The three functions of the root-mean-squared errors were superior to the hyperbolic discount model. Moreover, one of the three candidates was more “hyperbolic” than the standard hyperbolic discount function. These results indicate two things. One is that AI-Feynman can be a new data analysis method for inter-temporal choice experiments. The other is that AI-Feynman can discover discount functions that humans cannot find.