TechRxiv
IEEE_Intelligent_systemsV2.pdf (1.28 MB)

FUTURE AI SYSTEMS: AN AI OUTPERFORMS A HUMAN AS TEAMMATE

Download (1.28 MB)
preprint
posted on 2023-01-26, 15:56 authored by Joseph MakokhaJoseph Makokha

This paper presents a way in which Artificial Intelligence (AI) might contribute to the concept development stage of engineering design or other contexts involving ideation. We introduce the notion of a “Disruptive Interjector” (DI), an AI system that observes a person’s actions, and makes suggestions based on what the user does; then demonstrate it’s application in a human subject study involving collaborative teamwork. Furthermore, we present the results of team performance in which teams of two Humans (HH) and teams of a Human and Computer (HAI) collaboratively sketch a design based on a prompt, observing that the HAI partnership produced ideas that were consistently rated higher (by trained human judges) than the HH teams. The paper contributes to AI research and practice in at least two ways. First, it informs designers of such systems on design considerations, especially as AI technologies become pervasive in different domains like engineering design and others. Secondly, in addition to presenting a replicable method for conducting studies involving an AI, it contributes to research by providing empirical evidence of an AI outperforming a human as a collaborative partner on a concept development task.

History

Email Address of Submitting Author

josepmak@uio.no

ORCID of Submitting Author

0000-0002-7716-1414

Submitting Author's Institution

University of Oslo, Norway

Submitting Author's Country

  • Norway