Human Control Redressed: Comparing AI-To-Human Vs. Human-To-Human Predictability in a Real-Effort Task

15 Pages Posted: 19 Jan 2023

See all articles by Juliane Beck

Juliane Beck

University of St. Gallen

Thomas Burri

University of St. Gallen

Markus Christen

University of Zurich

François Fleuret

University of Geneva

Serhiy Kandul

University of Zurich

Markus Kneer

University of Zurich - Institute of Philosophy

Vincent Micheli

University of Geneva

Date Written: January 16, 2023

Abstract

Predictability is a prerequisite for effective human control of artificial intelligence (AI). The inability to predict malfunctioning of AI, for example, impedes timely human intervention. In this paper, we empirically investigate how AI’s predictability compares to the predictability of humans in a real-effort task. We show that humans are worse at predicting AI performance than at predicting human performance. Importantly, participants are not aware of the differences in relative predictability of AI and overestimate their prediction skills. These results raise doubts about the human ability to effectively exercise control of AI — at least in certain contexts.

Keywords: Human Control; AI Predictability; Lunar Lander Game; Human-Computer Interaction

Suggested Citation

Beck, Juliane and Burri, Thomas and Christen, Markus and Fleuret, François and Kandul, Serhiy and Kneer, Markus and Micheli, Vincent, Human Control Redressed: Comparing AI-To-Human Vs. Human-To-Human Predictability in a Real-Effort Task (January 16, 2023). Available at SSRN: https://ssrn.com/abstract=4325339 or http://dx.doi.org/10.2139/ssrn.4325339

Juliane Beck (Contact Author)

University of St. Gallen ( email )

Thomas Burri

University of St. Gallen ( email )

Bodanstrasse 3
Saint Gallen, St. Gallen CH-9000
Switzerland

HOME PAGE: http://www.unisg.ch/en/universitaet/schools/law/ueber-ls/faculty/burri

Markus Christen

University of Zurich ( email )

Rämistrasse 71
Zürich, CH-8006
Switzerland

François Fleuret

University of Geneva ( email )

Serhiy Kandul

University of Zurich ( email )

Markus Kneer

University of Zurich - Institute of Philosophy ( email )

Vincent Micheli

University of Geneva ( email )

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
95
Abstract Views
601
Rank
524,901
PlumX Metrics