Human control redressed: Comparing AI and human predictability in a real-effort task
Predictability is a prerequisite for effective human control of artificial intelligence (AI). For example, the inability to predict the malfunctioning of AI impedes timely human intervention. In this paper, we employ a computerized navigation task, namely, a game called lunar lander, to investigate...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2023-05-01
|
Series: | Computers in Human Behavior Reports |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2451958823000234 |
_version_ | 1797803882988961792 |
---|---|
author | Serhiy Kandul Vincent Micheli Juliane Beck Thomas Burri François Fleuret Markus Kneer Markus Christen |
author_facet | Serhiy Kandul Vincent Micheli Juliane Beck Thomas Burri François Fleuret Markus Kneer Markus Christen |
author_sort | Serhiy Kandul |
collection | DOAJ |
description | Predictability is a prerequisite for effective human control of artificial intelligence (AI). For example, the inability to predict the malfunctioning of AI impedes timely human intervention. In this paper, we employ a computerized navigation task, namely, a game called lunar lander, to investigate empirically how AI's predictability compares to humans' predictability. We ask participants to guess whether the landings of a spaceship performed by AI and humans will succeed. We show that humans are worse at predicting AI performance than at predicting human performance in this environment. Significantly, participants underestimate the differences in the relative predictability of AI and, at times, overestimate their prediction skills. These results raise doubts about the human ability to exercise control of AI effectively — at least in certain contexts. |
first_indexed | 2024-03-13T05:28:44Z |
format | Article |
id | doaj.art-bdf5f0c09ae1402ea2562adba3cb2c00 |
institution | Directory Open Access Journal |
issn | 2451-9588 |
language | English |
last_indexed | 2024-03-13T05:28:44Z |
publishDate | 2023-05-01 |
publisher | Elsevier |
record_format | Article |
series | Computers in Human Behavior Reports |
spelling | doaj.art-bdf5f0c09ae1402ea2562adba3cb2c002023-06-15T04:56:45ZengElsevierComputers in Human Behavior Reports2451-95882023-05-0110100290Human control redressed: Comparing AI and human predictability in a real-effort taskSerhiy Kandul0Vincent Micheli1Juliane Beck2Thomas Burri3François Fleuret4Markus Kneer5Markus Christen6University of Zurich, Switzerland; Corresponding author.University of Geneva, SwitzerlandUniversity of St. Gallen, SwitzerlandUniversity of St. Gallen, SwitzerlandUniversity of Geneva, SwitzerlandUniversity of Zurich, SwitzerlandUniversity of Zurich, SwitzerlandPredictability is a prerequisite for effective human control of artificial intelligence (AI). For example, the inability to predict the malfunctioning of AI impedes timely human intervention. In this paper, we employ a computerized navigation task, namely, a game called lunar lander, to investigate empirically how AI's predictability compares to humans' predictability. We ask participants to guess whether the landings of a spaceship performed by AI and humans will succeed. We show that humans are worse at predicting AI performance than at predicting human performance in this environment. Significantly, participants underestimate the differences in the relative predictability of AI and, at times, overestimate their prediction skills. These results raise doubts about the human ability to exercise control of AI effectively — at least in certain contexts.http://www.sciencedirect.com/science/article/pii/S2451958823000234Human controlAI predictabilityLunar lander gameHuman-computer interaction |
spellingShingle | Serhiy Kandul Vincent Micheli Juliane Beck Thomas Burri François Fleuret Markus Kneer Markus Christen Human control redressed: Comparing AI and human predictability in a real-effort task Computers in Human Behavior Reports Human control AI predictability Lunar lander game Human-computer interaction |
title | Human control redressed: Comparing AI and human predictability in a real-effort task |
title_full | Human control redressed: Comparing AI and human predictability in a real-effort task |
title_fullStr | Human control redressed: Comparing AI and human predictability in a real-effort task |
title_full_unstemmed | Human control redressed: Comparing AI and human predictability in a real-effort task |
title_short | Human control redressed: Comparing AI and human predictability in a real-effort task |
title_sort | human control redressed comparing ai and human predictability in a real effort task |
topic | Human control AI predictability Lunar lander game Human-computer interaction |
url | http://www.sciencedirect.com/science/article/pii/S2451958823000234 |
work_keys_str_mv | AT serhiykandul humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT vincentmicheli humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT julianebeck humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT thomasburri humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT francoisfleuret humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT markuskneer humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask AT markuschristen humancontrolredressedcomparingaiandhumanpredictabilityinarealefforttask |