Training AI: Science & Data
Teaching Machines to See the World
Resources & Downloads
Download the PDF versions of this scenario or the associated attachments.
Target Group: 8 - 12 y.o.
Activity Duration: 4-5 lessons (45 min each)
Key Learning Goals:
- Machine Learning: Describe how AI models learn from training and testing datasets.
- Data Collection: Collect, label, and organize datasets (e.g., traffic signs, faces).
- Critical Analysis: Identify limitations and errors in AI models.
- Ethics: Recognize the importance of diversity and human responsibility in AI.
Learning Outcomes
Students will be able to:
KNOWLEDGE & UNDERSTANDING:
- Describe how machine learning works in simple terms (training vs. testing).
- Understand how AI applies to scientific contexts like image recognition.
- Recognize key ethical considerations, such as bias and trust.
SKILLS & ABILITIES:
- Use tools like Teachable Machine to create and evaluate AI models.
- Analyze AI output and compare it with human reasoning.
- Collaborate in teams to design experiments and test hypotheses.
ATTITUDES & VALUES:
- Appreciate the value of combining human insight with technology.
- Show curiosity in exploring how AI intersects with science.
- Demonstrate openness to different perspectives during discussions.
European Dimension / Erasmus+ Connection
- Democracy & Inclusion: All countries contribute equally to shared datasets.
- Cultural Awareness: Recognizing that AI must adapt to regional differences (e.g., different traffic signs).
- Digital Responsibility: Introducing GDPR concepts and ethical image data usage.
1. Resources and Tools
Digital Tools:
- Teachable Machine: For training AI models (Google).
- eTwinning/TwinSpace: For international collaboration and file sharing.
- Padlet/Jamboard: For reflection and sharing results.
Materials:
- Tablets/Cameras for capturing images.
- AI-generated photos of ethnic groups (Att 1.1).
- Video of "AI Fails" (recognition mistakes).
2. Working Methods
- Scientific Inquiry: Collecting and categorizing real-world data.
- Simulation: Mimicking international research collaboration.
- Hands-on Training: Using no-code tools to build working AI models.
- Critical Debate: Discussing whether AI can replace human judgment.
Activity Overview
| Phase | Duration | Activity | Description |
|---|---|---|---|
| Intro | 20 min | Motivation | AI Bloopers: Watching video of AI recognition mistakes. Discussion: "Why did the robot fail?". |
| Activity 1 | 45 min | Many Faces, One World | Human vs AI: Classifying faces without technology. Discussing how machines "see" people vs. how humans do. |
| Activity 2 | 45 min | Exploring Traffic Signs | Data Collection: Photographing local traffic signs. Categorizing them (Regulatory, Warning, Guide). Uploading to eTwinning. |
| Activity 3 | 45 min | Training the Model | Teachable Machine: Training a model with the image sets. Testing with new images. Analyzing where it fails. |
| Activity 4 | 45 min | Ethics & Debate | Can AI Replace Humans? Case study of AI errors. Debate and digital reflection (Padlet). |
3. Introduction and Motivation
AI "Bloopers"
Goal: Show that AI is not perfect.
- Watch: "Funny Object Recognition Fails" video.
- Discuss: "Which mistake surprised you? Why?"
- Concept: Just like these AIs failed, your model might fail if you don't teach it well. You are the scientist!
4. Research and Learning
Activity 1: Many Faces, One World
Task: Students categorize AI-generated photos of diverse ethnic groups.
- Method: THINK - WRITE - PAIR - SHARE.
- Question: "How do machines distinguish people?" (Eyes, hair color, etc.).
- Conclusion: Humans use context and culture; machines look for pixel patterns.
Activity 2: Traffic Signs of Europe
Task: Create a dataset of traffic signs.
- Collect: Photograph signs in your local area.
- Share: Upload to eTwinning folder to share with partner countries.
- Categorize: Group signs by type (Warning vs. Regulatory) or Country.
5. Creative Application
Activity 3: Humans Teach the Machine
Tool: Teachable Machine (with Google).
- Train: Upload your traffic sign images into classes (e.g., "Stop Sign", "Speed Limit").
- Test: Show the AI a sign from a different country (provided by partners).
- Analyze: Did it recognize the foreign sign? Why or why not?
6. Reflection and Evaluation
Activity 4: Ethics & Limitations
Debate: "Can AI replace human reasoning in science?"
- Case Study: Analyze a specific AI error (e.g., misclassified image).
- Reflection: Post thoughts on a shared Padlet/Jamboard.
- Peer Feedback: Comment on reflections from partner schools.
Co-founded by the Erasmus+ Programme of the European Union • 2024-1-HR01-KA220-SCH-000254267
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License.