cs.ROAug 29, 2026

Teaching Robot Policies to Humans Using Erroneous Examples

Authors: Rithika Narayan, Suresh Kumaar Jayaraman, Henny Admoni

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Human-robot collaboration describes the process of humans and autonomous agents working together to accomplish common goals. This process is facilitated best when robot policies, or behaviors in different situations, are made transparent to humans. Demonstration-based explanations have been a focus of human-robot collaboration research, and the field has frequently drawn upon literature from education to improve how humans are taught robot policies. However, no single teaching method has been proven effective across domains, difficulties, learners, and other variables; the question of how humans can most effectively be taught robot policies remains open. In traditional classrooms, learners are shown erroneous examples, in which they reflect on and correct incorrect responses to understand common pitfalls when learning a concept. We propose using erroneous examples to teach robot policies, extending an existing policy teaching framework. We conduct a user study in which participants view incorrect demonstrations of robot behavior and correct the actions to align with the actual policy. Our findings suggest that viewing these incorrect demonstrations and verbalizing one's reasoning in predicting a robot's actions improves retention of the policy over time, in agreement with the effect of erroneous examples in classrooms. We also categorize participants into distinct learning styles and establish that participants using inverse reinforcement learning-like reasoning perform best on policy prediction tasks. With this work, we aim to advance the methods by which robots educate humans on their policies.

Figures & tables

Figure 1.1: Prior work communicates robot policies via a series of demonstrations of a robot’s optimal trajectories in various environments [ 15 ] . We propose integrating self-explanations, or verbalization of reasoning, and erroneous examples, or suboptimal trajectories, into the paradigm to improve a human learner’s ability to predict robot trajectories in new environments.
Figure 2.1: Interaction facilitating a demonstration-style explanation of the agent’s behavior
Figure 3.1: Update of particle filter based on introduction of red and blue half-space constraints in demonstration. Particles consistent with the constraints receive more weight (yellow ring) while particles inconsistent with the demonstrations are weighted with the von-Mises demonstration. [ 8 ]
Figure 3.2: Partial feedback policy teaching loop used in previous work [ 14 ]
Figure 3.3: Policy teaching loop with control condition (green), explanation condition (blue), and erroneous example condition (orange)
Figure 3.4: Possible series of demonstrations shown to participants to teach tradeoff of action cost and traveling through mud in taxi domain.
Figure 3.5: Unit test format under control (green) condition with prompt environment and response trajectory
Figure 3.6: Unit test under second experimental condition with prompt environment, response trajectory, and open-ended explanation
Figure 3.7: Unit test under third experimental condition, including erroneous example of trajectory, explanation of error, corrected trajectory, and explanation of reasoning
Figure 3.8: Example of feedback provided to participants contrasting a learner’s incorrect test response with the optimal response.
Figure 3.9: Example of two incorrect test responses ((a) and (b)) that reveal the same underlying misconception compared to the optimal response (c).
Figure 4.1: The control group is the only group to answer fewer questions correctly on the delayed post-test than on the immediate post-test. The self-explanation and erroneous example groups show improvement in performance.
Figure 4.2: The self-explanation group’s average regret decreases greatly from immediate to delayed post-test. This means that their trajectories deviated less from the optimal ones as time passed.
Figure 4.3: Prevalence of explanation types among self-explanation and erroneous example groups.
Figure 4.4: Categorization of participants in self-explanation and erroneous example groups by learning style.
Figure 4.5: (a) Average number of optimal responses on immediate post-test by style. (b) Average number of optimal responses on delayed post-test by style. IRL-style learners perform better than most other groups.
Figure 4.6: (a) Average regret on immediate post-test by style. (b) Average regret on delayed post-test by style. IRL-style learners perform better than most other groups.
More in this topic

Human-Robot Collaboration

Human-robot collaboration studies how people and robots coordinate tasks safely, effectively, and with appropriate human control.

See all 181 papers

Explore similar work

CardsList
Aug 21, 2026

Teaching is a Process: The TOSS Framework for Modeling Human Teaching Decisions in Human-Interactive Robot Learning

Successful Human-Robot Teaching assumes alignment between robot processing needs and human teaching intent. To better understand this alignment, this work seeks to uncover the underlying logic that humans intuitively apply when teaching. Through an exploratory, bottom-up study with N=34, participants observing two distinct robot Reinforcement Learning (RL) scenarios, we analyze 204 intuitive teaching responses across early, middle, and late learning phases. Results reveal that teaching decisions consist of a nuanced, interconnected network of Triggers (situational catalysts), Objectives (subjective teaching targets), Signals (communicative acts), and Strategies (high-level governance) in which teachers spontaneously adopt diverse roles, acting as coaches, engineers, or designers and prioritize different objectives. Based on these results, we introduce the TOSS Framework, which conceptualizes Human-Robot teaching as a procedural loop between robot behavior and human teaching actions, in which human teaching decisions are modeled as Trigger-Signal responses modulated by teaching Objectives and Strategies. It provides future research with an openly accessible dataset and a theoretical foundation for a) understanding teaching decisions and b) simulating realistic oracles as well as c) designing human-centered teaching settings and novel robot learning algorithms that go beyond the constraints of current robot learning settings.

Bernhard Hilpert, Kim Baraka, Joost Broekens

LIACS, Leiden University, Leiden, The Netherlands · VU Amsterdam, Amsterdam, The Netherlands

Human-Robot InteractionRobot Learning
AskPDF
May 27, 2026

How Should We Teach Robots? A Comparison of Kinesthetic, Joystick, and Gesture-Based Teaching

Instructing robots from demonstrations can be done through different teaching modalities, each with different usability and performance trade-offs. This paper compares kinesthetic guidance, joystick teleoperation, and hand gestures in a user study with eight participants. We evaluate replay success, modified NASA-TLX workload, and common teaching errors across three manipulation tasks. Kinesthetic guidance produced the shortest demonstrations, lowest workload, and highest success on the more orientation-sensitive and contact-rich tasks. Joystick teleoperation performed best on simple peg picking. Hand-gesture teaching, although less reliable overall, performed better than expected and in some cases achieved results comparable to kinesthetic guidance.

Petr Vanc, Jan Kristof Behrens, Václav Hlaváč +1

Czech Institute of Informatics, Robotics and Cybernetics (CIIRC CTU) Jugoslávských partyzánů 1580/3, 160 00 Dejvice

Robot TeleoperationRobotic ManipulationLearning from DemonstrationRobot Imitation Learning
AskPDF
May 21, 2026

Robots That Know What to Ask: Recovering Misaligned Rewards through Targeted Explanations

Learning reward functions from demonstrations assumes that demonstrations provide adequate supervision over all features -- or task-relevant aspects of behavior. In practice, demonstrations are often imperfect: humans may under-emphasize certain features due to cognitive load or physical difficulty, or the training regime may fail to sufficiently cover all relevant situations. In either case, important features may be underspecified, leading to ambiguity in the learned reward function and misaligned behavior at deployment. We propose a framework that detects such underspecified features and actively solicits targeted corrective demonstrations. Our key insight is that demonstrations implicitly reveal which features are well specified: features that are consistently optimized show little variation across demonstrations, while features that are underspecified vary widely. We leverage this statistical signal to infer which features may have been insufficiently demonstrated. The robot then explains which features it is uncertain about in natural language and queries for demonstrations that explicitly address the identified gaps. We evaluate our approach in a simulated tabletop manipulation domain and in a user study with a real Franka robot. Targeted, explanation-guided queries significantly improve reward recovery compared to random querying and passive data collection, reducing ambiguity that would otherwise persist in learning from imperfect demonstrations.

Helena Merker, Nick Walker, Andreea Bobu

Massachusetts Institute of Technology

Reinforcement LearningLearning from DemonstrationInteractive Imitation LearningRobot LearningLearning from Suboptimal Demonstrations
AskPDF