Abstract
In a largely forgotten 2008 conference paper, IS researchers demonstrated that a language model could predict the correlations between behavioral constructs. Since then, these findings have been extended into Organizational Behavior (OB) research and Psychology, and the results have solidified on pace with the size of the underlying language models. We will start from the premise that most IS behavioral variables are constructs and the axiom that such behavioral constructs are linguistic due to the psychometric method. My provocation is that correlational relationships between constructs are usually linguistic rather than empirical. We will discuss the implications on behavioral research and how the IS discipline may overcome a setting where most of our existing and future findings are directly calculable within LLMs. We will discuss how to disentangle false relationships, linguistic relationships, linguistic causality, and empirical causality between behavioral constructs to understand what human researchers can do that LLMs can’t.
Recommended Citation
Larsen, Kai R. and Yan, Sen, "Is Behavioral Cross-Sectional Information Systems Research Just Linguistic Manipulation?" (2024). ECIS 2024 Proceedings. 4.
https://aisel.aisnet.org/ecis2024/debates/debates/4
Is Behavioral Cross-Sectional Information Systems Research Just Linguistic Manipulation?
In a largely forgotten 2008 conference paper, IS researchers demonstrated that a language model could predict the correlations between behavioral constructs. Since then, these findings have been extended into Organizational Behavior (OB) research and Psychology, and the results have solidified on pace with the size of the underlying language models. We will start from the premise that most IS behavioral variables are constructs and the axiom that such behavioral constructs are linguistic due to the psychometric method. My provocation is that correlational relationships between constructs are usually linguistic rather than empirical. We will discuss the implications on behavioral research and how the IS discipline may overcome a setting where most of our existing and future findings are directly calculable within LLMs. We will discuss how to disentangle false relationships, linguistic relationships, linguistic causality, and empirical causality between behavioral constructs to understand what human researchers can do that LLMs can’t.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.