Artificial Intelligence Quotient (AIQ)

Abstract

We introduce Artificial Intelligence Quotient (AIQ)—a person’s ability to use AI to perform a wide variety of tasks—and demonstrate its existence using five studies (archival, lab, and online) across different AIs and samples. Study 1 (18-year global dataset of human+AI chess) and Study 2 (longitudinal study of human+AI renju games) show that individuals have s1 human+AI performance over time (controlling for human’s own capability and AI’s capability), suggesting a stable human+AI capability exists. Study 3 shows that a general AIQ factor can be statistically extracted from individuals’ performance on a variety of tasks completed with ChatGPT, a more general AI tool. Besides replicating Study 3’s findings in larger samples, Study 4 and Study 5 (preregistered) show that the extracted AIQ factor predicts human+AI performance on a new task using the same AI (ChatGPT) on the same day (concurrent validity), and predicts human+AI performance on other tasks using different AIs (renju AI/Gemini) in the future (prospective validity). We ascertain AIQ’s explanatory power by controlling for individual’s IQ, social intelligence, AI knowledge, etc. Together, our findings suggest that AIQ exists and is measurable. By establishing this new type of intelligence (AIQ), we shed light on individual differences in the ability to use AI, which is increasingly important for individuals, organizations, and society.

Ähnliche Arbeiten