Impact of Different Failures on a Robot's Perceived Reliability

Andrew Violette; Zhanxin Wu; Haruki Nishimura; Masha Itkina; Leticia Priebe Rocha; Mark Zolotas; Guy Hoffman; Hadas Kress-Gazit

cs.ROMar 9, 2026

Impact of Different Failures on a Robot's Perceived Reliability

Authors: Andrew Violette, Zhanxin Wu, Haruki Nishimura, Masha Itkina, Leticia Priebe Rocha, Mark Zolotas, Guy Hoffman, Hadas Kress-Gazit

Organizations: Department of Computer Science, Cornell University · Toyota Research Institute (TRI). · Sibley School of Mechanical and Aerospace Engineering, Cornell University

Abstract

Robots fail, potentially leading to a loss in the robot's perceived reliability (PR), a measure correlated with trustworthiness. In this study we examine how various kinds of failures affect the PR of the robot differently, and how this measure recovers without explicit social repair actions by the robot. In a preregistered and controlled online video study, participants were asked to predict a robot's success in a pick-and-place task. We examined manipulation failures (slips), freezing (lapses), and three types of incorrect picked objects or place goals (mistakes). Participants were shown one of 11 videos -- one of five types of failure, one of five types of failure followed by a successful execution in the same video, or a successful execution video. This was followed by two additional successful execution videos. Participants bet money either on the robot or on a coin toss after each video. People's betting patterns along with a qualitative analysis of their survey responses highlight that mistakes are less damaging to PR than slips or lapses, and some mistakes are even perceived as successes. We also see that successes immediately following a failure have the same effect on PR as successes without a preceding failure. Finally, we show that successful executions recover PR after a failure. Our findings highlight which robot failures are in higher need of repair in a human-robot interaction, and how trust could be recovered by robot successes.

Figures & tables

Fig. 1: 6 of the 11 conditions examined in this study. From left to right: a) Success . The robot puts the bottle in the trash. b) Slip . The robot is unable to pick up the bottle. c) Lapse . The robot positions to pick up the bottle, then freezes for 15 seconds. d) Mistake (Thermos) . The robot picks up the reusable thermos and places it in the trash. e) Mistake (Stapler) . The robot picks up the stapler and places it into the trash. f) Mistake (Marker) . The robot picks up the marker and places it into the mug. Not shown: The other five conditions, which are the failure modes (b–f) immediately followed by success.

Fig. 2: The experimental setup. The robot (blue) attempted to place the disposable plastic water bottle (pink) into the trash (green). In some videos, the robot interacted with other objects (dotted red). Participants saw an unlabeled version of this before placing bets.

Fig. 3: The flow of the experiment. Participants were asked to place bets on robot success before and after seeing three videos—one of the experimental conditions, then two success videos.

Fig. 4: Change in participant bet after they see the condition video, compared to the bet before. Each light dot represents one participant, the dark dot is the mean of participants, and the whiskers are the 95% confidence interval of the mean. We see evidence for differences in effect of failure conditions (slip, mistake, and lapse), with no evidence for difference in effect if the demonstration contains success. Mistake and Success are abbreviated as M and S, respectively.

	Failure	Failure+Success
Slip	$-3.40\pm 0.84$	$+0.23\pm 0.69$
Lapse	$-2.46\pm 0.87$	$+1.60\pm 0.70$
Mistake (Marker)	$-1.68\pm 0.82$	$+1.13\pm 0.69$
Mistake (Stapler)	$-0.67\pm 0.84$	$+0.61\pm 0.69$
Mistake (Thermos)	$+0.19\pm 0.90$	$+1.93\pm 0.71$
No Failure		$+1.31\pm 0.71$

TABLE I:

\Delta^{1\rightarrow 2}_{\text{PR}}

, the difference in perceived reliability (PR) after seeing the conditional video across conditions, reported as mean

\pm

standard error.

Fig. 5: Participant betting behavior before any videos (bet 1), after the Slip video (bet 2), and after additional success videos (bets 3 and 4). Perceived reliability (PR) drops after seeing the robot slip. PR rises after seeing success, ending above the baseline from Bet 1.

Fig. 6: Participant evaluation (horizontal) matched our labels (vertical), though participants often labeled lapses and mistakes as slips. Mistake (M) and Success (S) are abbreviated.

Explore similar work

REPAIR-Bench: A Benchmark for Robot Error Perception And Interaction Recovery
Jun 29, 2026Giuliano Pioldi, Yashika Batra, Arman Ibrayeva +4Robot Failure Detection Robot Failure Recovery
Cornell University Cornell Tech New York, NY
Toward Study on the Effects of Robot's Humorous Apology on Human's Trust Recovery
Oct 12, 2026Gayoung Kim, Go-Eum Cha, Wonse JoHuman-Robot Interaction Trust in AI
A+HRI Laboratory, Department of Information and Telecommunication Engineering, Incheon National University, Incheon, South Korea. · Department of Computer Science, Purdue University, IN, USA.
RFM-HRI : A Multimodal Dataset of Medical Robot Failure, User Reaction and Recovery Preferences for Item Retrieval Tasks
Mar 5, 2026Yashika Batra, Giuliano Pioldi, Promise Ekpo +5Human-Robot Interaction Robot Failure Recovery
Cornell University, Cornell Tech, United States of America · Weill Cornell Medicine, United States of America

Jun 29, 2026J/K move · Enter open · S save

REPAIR-Bench: A Benchmark for Robot Error Perception And Interaction Recovery

Giuliano Pioldi, Yashika Batra, Arman Ibrayeva, Yuanchen Bai+3

Cornell University Cornell Tech New York, NY

Robot Failure Detection Robot Failure Recovery Human-Robot Interaction Anomaly Detection

Open paper AskPDF arXiv

Understanding how users perceive and respond to robot failures is essential for building robust and trustworthy robot systems. Prior work, however, (i) often treats failures as independent events, (ii) emphasizes binary failure detection, (iii) with rule-based recovery modeling. We present REPAIR-Bench, built on 214 interaction trials from 41 participants, the benchmark spans four induced failure types and provides synchronized facial action units, head pose, speech transcripts, and post-interaction affect and recovery reports. The benchmark spans three novel evaluation tasks that jointly capture the lifecycle of failure in human-robot interaction (HRI): (i) failure detection over inter-dependent interaction sessions, modeling longitudinal user adaptation across repeated failures; (ii) visual failure-type classification beyond binary success/failure formulations; and (iii) user-centered recovery prediction, inferring users' preferred recovery strategies from interaction context rather than relying on manually designed or rule-based strategies. In baseline experiments, hierarchical recurrent modeling improved failure detection over a single-session model (strict F1: 0.80 vs. 0.68), achieved a failure localization mean signed error of -0.51 s, median absolute error of 2.97 s and, for recovery prediction, a QLoRA-tuned Mistral-7B reached Hit@5=0.76 and F1@5=0.32. REPAIR-Bench provides both the HRI and Medical HRI communities with a standardized framework for (1) evaluating robot failures and (2) building transparent, adaptive, and trustworthy recovery systems.

Abstract

Explore similar work

REPAIR-Bench: A Benchmark for Robot Error Perception And Interaction Recovery

Toward Study on the Effects of Robot's Humorous Apology on Human's Trust Recovery

RFM-HRI : A Multimodal Dataset of Medical Robot Failure, User Reaction and Recovery Preferences for Item Retrieval Tasks