Robot Trust Calibration in Human Handover Tasks

ISEF Category: Robotics and Intelligent Machines

Subcategory: Other  ·  Difficulty: Intermediate  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

People trust robots fast, then lose trust even faster when the robot seems unsure. That makes trust a moving target, not a fixed feeling. Your project asks a smart question: does a robot sound more reliable when it stays silent, or when it admits uncertainty? You can measure that with real users and real choices.

What Is It?

This project studies human-robot trust calibration. Trust calibration means matching a user's trust to the robot's real ability. If a robot is good at a task, people should feel comfortable relying on it. If the robot is shaky, people should hold back a little.

You can think of it like a teammate in gym class. If your teammate says, “I’ve got this,” you may pass them the ball. If they say, “I’m not sure,” you may hold onto it. Your study tests whether a robot's words change that decision, even when the arm itself does the same task every time.

The key idea is simple. The robot gives one of two communication styles, silent execution or a verbal uncertainty disclaimer. Then you measure what people do, such as whether they hand over an object, how competent they think the robot is, and how much they trust it after repeated trials.

Why This Is a Good Topic

This is a strong science fair topic because you can test it with human behavior data, clear survey scores, and simple statistics. The variable you change is easy to define, which makes the project clean and repeatable. It also connects to real problems in caregiving robots, warehouse robots, and home assistants, where trust affects safety and cooperation. You can learn study design, consent, survey construction, and within-subject analysis without needing a full robotics lab.

Research Questions

• How does a robot's verbal uncertainty disclaimer change user handover rates compared with silent execution?
• What is the effect of uncertainty disclaimers on perceived robot competence after repeated tasks?
• Does prior exposure to the robot reduce the effect of uncertainty disclaimers on trust scores?
• To what extent do handover decisions match self-reported trust in a within-subject design?
• Which wording of an uncertainty disclaimer leads to the highest willingness to cooperate?
• How does task difficulty change the impact of robot uncertainty cues on user decisions?

Basic Materials

• Simple robotic arm or tabletop robot with repeatable motion.
• Laptop or tablet for showing instructions and recording survey responses.
• Printed consent form and short participant information sheet.
• Timer or stopwatch for consistent trial pacing.
• Clipboard or digital form for trust and competence ratings.
• Randomization sheet for counterbalancing task order.
• Object to hand over, such as a small foam block or plastic cup.
• Quiet room with a fixed setup area and clear table space.

Advanced Materials

• Programmable robotic arm with speech output or text-to-speech support.
• Microcontroller or robot control computer for scripted communication conditions.
• External camera for behavioral coding of handover timing and hesitation.
• Survey platform for collecting Likert-scale trust and competence ratings.
• Audio recorder for checking whether disclaimer wording stays consistent.
• Statistical software for repeated-measures analysis and effect size calculations.
• Optional eye-tracking or motion-tracking system for attention and hesitation analysis.

Software & Tools

• Google Forms: Collects trust ratings, competence ratings, and post-trial feedback quickly and for free.
• Excel: Organizes trial order, handover outcomes, and summary statistics.
• JASP: Runs paired tests, repeated-measures analysis, and effect size estimates without coding.
• R: Supports stronger statistical analysis, plots, and within-subject modeling.
• ImageJ: Helps you code frame-by-frame hesitation or handover timing from video.

Experiment Steps

1. Define the trust signal you will test, such as silent execution versus a spoken uncertainty disclaimer.
2. Choose one repeatable robot task and one clear human response, such as handing over an object or withholding it.
3. Plan a within-subject order that balances practice effects, fatigue, and expectation bias.
4. Build a scoring system for behavior and perception, then decide how you will record both.
5. Design controls that keep the robot's actual motion identical across conditions.
6. Plan the statistics you will use to compare conditions and check whether the effect is consistent across participants.

Common Pitfalls

• Changing the robot's motion between conditions, which makes users react to motion quality instead of the disclaimer.
• Using unclear disclaimer wording, which confuses participants and blurs the trust effect.
• Forgetting to counterbalance trial order, which lets practice or fatigue distort the results.
• Measuring only survey trust and ignoring handover behavior, which weakens the study's real-world value.
• Letting room noise, experimenter tone, or facial cues differ across trials, which adds hidden bias.

What Makes This Competitive

A stronger project would separate what people say from what they actually do. If you compare behavior, ratings, and response time, you can see whether trust changes in more than one way. A competitive version also tests more than one disclaimer style and uses proper counterbalancing and repeated-measures statistics. That kind of design shows you understand both human factors and experimental control.

Project Variations

• Use a collaborative robot arm versus a humanoid voice assistant to see whether the same disclaimer works across robot types.
• Compare verbal uncertainty disclaimers with confidence statements to test which message better predicts handover behavior.
• Swap the task from object handover to a sorting or placement task to see whether trust shifts with task risk.

Learn More

• NIH PubMed: Search for review articles on human-robot trust, trust calibration, and human factors in robot interaction.
• IEEE Xplore: Search for peer-reviewed papers on robot transparency, uncertainty communication, and collaboration.
• MIT OpenCourseWare: Look for free materials on experimental design, human-computer interaction, and statistics.
• NASA NTRS: Search the technical report server for human-autonomy trust and transparency studies.
• OECD iLibrary and government research reports: Look for accessible reports on automation trust, workplace robots, and human factors.
• University library guides: Search your local university's open research guides for human subjects methods and survey design.