An HRI Experiment on Children's Language Learning Using the Furhat Social Robot
Human-Robot Interaction Project
Universitร degli Studi di Napoli Federico II
- Overview
- Research Objective
- The Furhat Robot
- Experiment Design
- Learning Activities
- Implementation
- Project Structure
- Installation
- Usage
- Results
- Conclusions
This project explores how emotional expressiveness in social robots affects language learning outcomes in children. Using the Furhat robot as an English tutor, we compare two interaction modalities: an expressive robot displaying emotions and gestures versus a neutral robot with minimal emotional cues.
The study is grounded in Human-Robot Interaction (HRI) research, which investigates how humans and robots can effectively communicate and collaborate, particularly in educational contexts.
Primary Goal: Demonstrate how a robot's emotional expression influences language learning in children.
Research Questions:
- Does emotional expressiveness improve learning engagement?
- Do children perform better with an expressive vs. neutral robot tutor?
- How do children perceive and respond to different robot personalities?
Furhat is a humanoid social robot designed for natural and intuitive communication with humans.
| Feature | Description |
|---|---|
| Realistic Facial Expressions | Projected face enables lifelike emotions |
| Head & Gaze Movement | Natural head tracking and eye contact |
| Natural Multilingual Voice | Text-to-speech in multiple languages |
| Social Adaptability | Responds to user behavior and context |
| Participant | Age | Focus |
|---|---|---|
| Sofia | 9 years | English vocabulary & grammar |
| Christian | 6 years | Basic English words |
| ๐ญ Expressive Robot | ๐ Neutral Robot |
|---|---|
|
โข Facial expressions (smiles, winks) โข Voice variations (pitch, rate) โข Emotional feedback (joy, empathy) โข Head movements (nodding) |
โข Neutral facial expression โข Monotone voice โข Minimal feedback โข No head movements |
Furhat acts as an English tutor through three interactive exercises:
Objective: Child repeats a word proposed by the robot
Objective: Child guesses the synonym or antonym of a given word
Objective: Child completes a sentence with an appropriate word
| Tool | Purpose |
|---|---|
| Furhat SDK | Skill creation and interaction management |
| Furhat Emulator | Testing robot behavior in simulation |
| IntelliJ IDEA | Kotlin development IDE |
| Aspect | Expressive | Neutral |
|---|---|---|
| Facial Expressions | Gestures.BigSmile, Gestures.Nod, Gestures.Wink |
None |
| Voice Tone | furhat.voice.copy(pitch=1.6, rate=0.9) |
Standard voice |
| Motivational Feedback | "You're doing great!", "Wonderful!" | "That's correct.", "Wrong." |
| Head Movements | Gestures.Nod for encouragement |
None |
Furhat-Language-Learning/
โ
โโโ ๐ Englishteacher/ # ๐ญ EXPRESSIVE VERSION
โ โโโ assets/
โ โ โโโ webTemplates/BASIC/ # Web interface templates
โ โโโ gradle/wrapper/ # Gradle wrapper files
โ โโโ src/ # Kotlin source code
โ โ โโโ main/kotlin/furhatos/app/englishteacher/
โ โ โโโ main.kt # Entry point
โ โ โโโ flow/ # Interaction states
โ โ โโโ nlu/ # Speech recognition
โ โโโ .gitignore
โ โโโ README.md
โ โโโ build.gradle # Build configuration
โ โโโ gradlew # Gradle wrapper (Unix)
โ โโโ gradlew.bat # Gradle wrapper (Windows)
โ โโโ skill.properties # Skill metadata
โ
โโโ ๐ Neutral/ # ๐ NEUTRAL VERSION
โ โโโ assets/
โ โ โโโ webTemplates/BASIC/ # Web interface templates
โ โโโ gradle/wrapper/ # Gradle wrapper files
โ โโโ src/ # Kotlin source code
โ โ โโโ main/kotlin/furhatos/app/Neutral/
โ โ โโโ main.kt # Entry point
โ โ โโโ flow/ # Interaction states
โ โ โโโ nlu/ # Speech recognition
โ โโโ .gitignore
โ โโโ README.md
โ โโโ build.gradle # Build configuration
โ โโโ gradlew # Gradle wrapper (Unix)
โ โโโ gradlew.bat # Gradle wrapper (Windows)
โ โโโ skill.properties # Skill metadata
โ
โโโ ๐ Presentazione_HRI_EV.pptx # Project presentation
โโโ ๐ README.md # This file
| Folder | Description |
|---|---|
Englishteacher/ |
Expressive Furhat skill with emotions, gestures, and motivational feedback |
Neutral/ |
Neutral Furhat skill with minimal expressions and monotone responses |
-
Furhat SDK (v2.8.0 or later)
- Download from Furhat Developer Portal
-
IntelliJ IDEA (Community or Ultimate)
- Download from JetBrains
-
Java JDK 11+
-
Clone the repository
git clone https://github.com/yourusername/Furhat-Language-Learning.git cd Furhat-Language-Learning -
Install and Start Furhat SDK
# Launch the Furhat SDK desktop launcher ./furhat-sdk-desktop-launcher -
Start Virtual Furhat
- Open Furhat SDK Launcher
- Click "Start Virtual Furhat"
- Wait for the emulator to load
-
Open a skill in IntelliJ IDEA
For Expressive Version:
cd EnglishteacherFor Neutral Version:
cd Neutral- Open IntelliJ IDEA
- File โ Open โ Select the skill folder
- Wait for Gradle sync to complete
-
Build the project
./gradlew build
-
Navigate to the Englishteacher folder
cd Englishteacher -
Run the skill
./gradlew run
Or run directly from IntelliJ IDEA (Run โ Run 'main')
-
Interact with Furhat
- The robot will greet you with expressions and emotions
- Follow the English learning exercises
- Receive motivational feedback
-
Navigate to the Neutral folder
cd Neutral -
Run the skill
./gradlew run
-
Interact with Furhat
- The robot will provide instructions with neutral tone
- Follow the same exercises with minimal feedback
While the skill is running, you can access the control panel:
- Open browser at
http://localhost:54321 - Monitor interactions and control the robot manually
| Command | Expressive | Neutral |
|---|---|---|
| Navigate | cd Englishteacher |
cd Neutral |
| Run | ./gradlew run |
./gradlew run |
| Expected Behavior | Smiles, nods, encouraging phrases | Monotone, minimal response |
| Metric | Expressive Furhat | Neutral Furhat |
|---|---|---|
| Emotional Engagement | High (wonder, enthusiasm) | Medium-Low |
| Answer Accuracy | Higher | Lower |
| Average Response Time | Faster | Slower |
| Children's Feedback | "It's fun!" | "Not very stimulating" |
- ๐ฏ Higher Engagement: Children showed significantly more enthusiasm with the expressive robot
- ๐ Better Performance: Accuracy in language exercises was higher with the expressive tutor
- โก Faster Responses: Children responded more quickly, suggesting higher motivation
- ๐ Positive Perception: Children described the expressive robot as "fun" and "friendly"
This experiment demonstrates that emotional expressiveness in social robots significantly enhances language learning outcomes in children.
- Expand participant pool for statistical significance
- Long-term learning retention studies
- Adaptive expressiveness based on child's emotional state
This project is developed for academic purposes as part of the Human-Robot Interaction course at Universitร degli Studi di Napoli Federico II.