Salvato in:
Dettagli Bibliografici
Autori principali: Zhou, Jieyu, Roy, Aryan, Gupta, Sneh, Weitekamp, Daniel, MacLellan, Christopher J.
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.05307
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918487281106944
author Zhou, Jieyu
Roy, Aryan
Gupta, Sneh
Weitekamp, Daniel
MacLellan, Christopher J.
author_facet Zhou, Jieyu
Roy, Aryan
Gupta, Sneh
Weitekamp, Daniel
MacLellan, Christopher J.
contents Existing AI agents typically execute multi-step tasks autonomously and only allow user confirmation at the end. During execution, users have little control, making the confirm-at-end approach brittle: a single error can cascade and force a complete restart. Confirming every step avoids such failures, but imposes tedious overhead. Balancing excessive interruptions against costly rollbacks remains an open challenge. We address this problem by modeling confirmation as a minimum time scheduling problem. We conducted a formative study with eight participants, which revealed a recurring Confirmation-Diagnosis-Correction-Redo (CDCR) pattern in how users monitor errors. Based on this pattern, we developed a decision-theoretic model to determine time-efficient confirmation point placement. We then evaluated our approach using a within-subjects study where 48 participants monitored AI agents and repaired their mistakes while executing tasks. Results show that 81 percent of participants preferred our intermediate confirmation approach over the confirm-at-end approach used by existing systems, and task completion time was reduced by 13.54 percent.
format Preprint
id arxiv_https___arxiv_org_abs_2510_05307
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle When Should Users Check? Modeling Confirmation Frequency inMulti-Step Agentic AI Tasks
Zhou, Jieyu
Roy, Aryan
Gupta, Sneh
Weitekamp, Daniel
MacLellan, Christopher J.
Human-Computer Interaction
Existing AI agents typically execute multi-step tasks autonomously and only allow user confirmation at the end. During execution, users have little control, making the confirm-at-end approach brittle: a single error can cascade and force a complete restart. Confirming every step avoids such failures, but imposes tedious overhead. Balancing excessive interruptions against costly rollbacks remains an open challenge. We address this problem by modeling confirmation as a minimum time scheduling problem. We conducted a formative study with eight participants, which revealed a recurring Confirmation-Diagnosis-Correction-Redo (CDCR) pattern in how users monitor errors. Based on this pattern, we developed a decision-theoretic model to determine time-efficient confirmation point placement. We then evaluated our approach using a within-subjects study where 48 participants monitored AI agents and repaired their mistakes while executing tasks. Results show that 81 percent of participants preferred our intermediate confirmation approach over the confirm-at-end approach used by existing systems, and task completion time was reduced by 13.54 percent.
title When Should Users Check? Modeling Confirmation Frequency inMulti-Step Agentic AI Tasks
topic Human-Computer Interaction
url https://arxiv.org/abs/2510.05307