Skip to content

IRL-CT/calling-for-backup

Repository files navigation

Calling for Backup: How Children Navigate Successive Robot Communication Failures

HRI 2026 | Edinburgh, Scotland, UK

Maria Teresa Parreira* ¹ · Isabel Neto* ² · Filipa Rocha²³ · Wendy Ju¹⁴

¹Cornell University · ²Lasige, Faculdade de Ciências · ³ITI/LarSys, Universidade de Lisboa · ⁴Cornell Tech

*Equal contribution


🌐 Project Website

View Project Website

Calling for Backup - Main Video Vignette

Experimental Conditions

  • Interruption Condition (n=30): Robot interrupts child before they finish answering
  • Control Condition (n=29): Robot does not interrupt
  • All participants experienced 3 successive performance errors

Robot Platform

  • Robot: Simon (Nodbot) - simple 2-axis robot with speaker
  • Control: Wizard-of-Oz methodology
  • Recording: 2 cameras (room view + laptop webcam)

Protocol

  1. Introduction to robot Simon
  2. Video visualization task (6 short videos with robots/humans, successes/failures)
  3. Pre-interaction robot perception survey (5-point Likert scale)
  4. Social error manipulation (interruption or control during Q&A)
  5. Performance error: 3 successive failures to understand "call the researcher" command
  6. Post-interaction robot perception survey
  7. Debriefing and deception explanation

Session duration: ~7 minutes

Research Question

How do children perceive and react to repeated robot error?

To investigate this question, we examine children's responses to two types of robot failures:

  • Social Error (Interruption): Robot inappropriately interrupts the child during conversation, violating social norms
  • Performance Error (Successive Failure): Robot fails to understand the child's request to "call the researcher" three successive times

Key Findings

Similarities with Adult Responses

Like adults (Liu et al., 2025), children demonstrated:

  • Verbal adaptation: Modified prompts (repeating, simplifying, adding specificity) and adjusted vocal tone/cadence (slower speech, demanding/interrogative tones)
  • 📈 Emotional progression: Similar evolution from confusion to frustration across successive errors
  • ⏱️ Response timing: Increased response latencies and variability with successive failures
  • 🚪 Interaction abandonment: Some participants completely discontinued interaction by the third error

Key Differences from Adult Responses

Children exhibited distinct behavioral patterns:

  • 🔴 Greater disengagement: Children showed notably more disengagement behaviors, often temporarily ignoring the robot
  • 🙋 External help-seeking ("Calling for Backup"): Unlike adults who attempted self-directed problem-solving, children looked for or called the researcher, demonstrating different agency expectations
  • 🙏 Politeness strategies: Children more readily shifted to polite language forms ("please") when initial commands failed
  • Earlier disengagement: Children showed disengagement beginning at Error I, while adults primarily abandoned after Error III (n=7)
  • 🔄 Mixed engagement patterns: Children displayed more dynamic patterns, sometimes disengaging then re-engaging, unlike the more linear adult progression

Social Error (Interruption)

  • 84% of children (21/25) in the Interruption condition shifted their answer to the new question, not acknowledging the interruption
  • Only 2 children continued responding to the previous question
  • Children appeared oblivious to or unconcerned with interruption as a social violation

Robot Perception

No significant changes in robot perception across any measured dimension:

  • Willingness to continue interacting
  • Competence
  • Trust
  • Social acceptance
  • Likeability

This suggests children have more flexible conversational expectations with robots compared to adults.

Behavioral Response Patterns

Most common engagement trajectories (Error I → Error II → Error III):

  1. Engage → Engage → Engage (n=15)
  2. Engage → Disengage → Engage (n=5)
  3. Engage → Disengage → Disengage (n=4)

Response categories observed:

  • Verbal reprompting: Repeating, more specific/longer prompts, simpler prompts, term swapping
  • Vocal modifications: Slower speech, demanding tone, interrogative tone, assertive tone, filler words
  • Physical behaviors: Moving closer to robot, standing up, leaving room
  • Emotional displays: Confusion, frustration, amusement/humor
  • Disengagement: Looking for researcher, no prompt, quitting interaction

Design Implications for Child-Robot Interaction

Observed Behavior Design Implication
Children's perception remained stable despite successive errors Design for Error Tolerance: Focus on managing errors gracefully rather than aggressively eliminating them. Allow minor errors without elaborate apologies.
Children frequently call for adult help after repeated failures Facilitate External Help-Seeking: Recognize requests for human assistance as valid and implement graceful handoff mechanisms.
Children use sophisticated verbal strategies (repetition, specificity, politeness) Implement Multi-Layered Conversational Repair: Design systems sensitive to prompt reformulation and increase error confidence based on politeness markers.
Children exhibit dynamic engagement patterns Employ Non-Intrusive Error Recovery: Recognize disengagement as active problem-solving. Avoid intrusive re-engagement; monitor passively.
Response latencies increase with successive errors Utilize Timing as Failure Metric: Integrate response latency analysis as real-time measure of cognitive load and distress.

Repository Structure

.
├── docs/                                    # Project website files
│   ├── index.html                          # Main website page
│   ├── style.css                           # Website styling
│   ├── README.md                           # Website documentation
│   ├── figures/                            # All images and video
│   │   ├── badchild_videofigure.mp4       # Main video vignette
│   │   ├── collage_p*.png                 # Participant vignettes
│   │   ├── doodle_*.png                   # Behavioral illustrations
│   │   ├── experimental_setup.png         # Study setup
│   │   └── vignette_p*.png                # Additional examples
│   └── submission/                         # Paper PDFs
│       ├── manuscript_calling4backup.pdf
│       └── supp_material_calling4backup.pdf
├── submission/                             # Original submission materials
│   ├── manuscript_calling4backup.pdf
│   ├── supp_material_calling4backup.pdf
│   └── figures/                            # Paper figures
└── README.md                               # This file

Materials

Keywords

successive error · robot error · reproducibility · child-robot interaction · error recovery · performance error · social error

Citation

@inproceedings{parreira2026calling,
  title={Calling for Backup: How Children Navigate Successive Robot Communication Failures},
  author={Parreira, Maria Teresa and Neto, Isabel and Rocha, Filipa and Ju, Wendy},
  booktitle={TBD},
  year={2026},

Related Work

This study reproduces and extends the successive robot failure paradigm from:

  • Liu, S., Parreira, M. T., & Ju, W. (2025). "I'm Done": Describing Human Reactions to Successive Robot Failure. HRI 2025.

Contact

For questions about this research, please contact the authors:


© 2026 Authors. Licensed under Cornell University IRB exempt protocol #IRB0010006.

About

A repository for the paper "Calling for Backup: How Children Navigate Repeated Robot Communication Failures"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published