Developing novel tools for applications of argumentation to behavioral economics

Consumer purchasing decisions can be considered as a form of preference-based human reasoning. There are two major schools of thought on preference. While mentalism asserts that preference reflects the true mental state of a person, behaviorism is the view in which preference is considered as a mathematical construct.

According to behaviorists, people's actions and not words are decisive in determining their preferences. Economists foster this behavioral preference of consumers with Revealed Preference Theory (RPT), also known as consumer theory.

As reasoning is involved in preference, it is instructive to generalize RPT to artificial intelligence (AI), currently dictated by mentalism.

With this forethought, Professor Van-Nam Huynh from Japan Advanced Institute of Science and Technology (JAIST) and Assistant Professor Nguyen Duy Hung from Sirindhorn International Institute of Technology, Thammasat University, Thailand, have recently generalized consumer theory to AI reasoning using argumentation—a kind of reasoning that draws inspiration from the process of people exchanging arguments to reach conclusions in daily life.

In an article published in the International Journal of Approximate Reasoning, the researchers present the theoretical foundation and analytical tools for practical applications of argumentation in analyzing consumer mentalism and behavior.

Prof. Huynh highlights the novelty of their work, "This paper bridges two lines of research: AI argument-based reasoning and behavioral economics. In particular, it explores the relationships between economic rationalities and argumentation semantics, between consumer's preference and AI agent's preference, and between consumer's purchasing behavior and AI agent's reasoning behavior."

In this work, the contributions of the researchers are three-fold. They first developed a Revealed Preference Argumentation (RPA) framework. The researchers argued that the existing frameworks are governed by the opposing mentalistic interpretation of preference. Thus, they re-constructed and unified two main approaches to RPT in terms of argumentation, demonstrating that RPT-based consumer analyses, including different rationality checks of a consumer behavior and extrapolations of such behaviors, can be interpreted as computational tasks in RPA.

Following that, the researchers successfully integrated mentalism and behaviorism to present an Integrated Preference Argumentation (IPA) framework. They established that RPA is just a special case of IPA with only 'revealed' preference. This finding is particularly important as existing preference-based argumentation frameworks are presented as IPA frameworks with only 'stated' preference.

Finally, the researchers developed comprehensive IPA algorithms, rigorously establishing their accuracy and termination for a general class of IPA frameworks. The researchers successfully implemented the algorithms in Prolog—a logic programming language associated with AI and computational linguistics—and obtained an IPA reasoning engine. Subsequently, they tested the developed tool to effectively analyze RPT-based consumer behavior.

In summary, this work makes remarkable inroads to a largely unexplored area in consumer behavioral economics. "This paper not only provides a theoretical and algorithmic foundation but also development tools for applications of argumentation to behavioral economics in consumer's behavior analyses such as rationality checks, consumer's preference recoveries, and behavior extrapolations," notes Prof. Huynh.