How to teach AI to a ten year old with the help of chatGPT - part three - A syllabus

In the first two parts of this post?

How to teach AI to a ten year old with the help of chatGPT - Part One?

And

How to teach AI to a ten year old with the help of chatGPT - part two

I explained the background and approach for teaching AI

In this post, I explain the syllabus?

?

Principles

The syllabus is based on the following principles

• Scenario driven approach,?
• Uses Synthetic data?
• Scaffolding - based on the elements discussed in part two
• Problem solving is central?
• Critical thinking is encouraged
• Teacher driven - teachers play an active part in scaffolding, scenarios, group discussions etc
• Designed to produce either code (Python generated chatGPT code) or data analysis (also chatGPT generated)?
• Group discussion in the cohort is encouraged?
• My teaching is based on the Oxford Tutorial system? Unique to Oxford and Cambridge

?

What do students need to know

• Fundamentals of data driven decision making
• Data pipeline concepts
• Prompt engineering
• What decisions can be made - Cannot be made using data
• How do we know that the decision is acceptable and what are the limits
• What data is needed to make the decision
• Feature engineering
• Model building
• Model evaluation

Outcomes

• Data analysis using chatGPT
• Python code using chatGPT

Pedagogical Enhancements

• Critical Thinking Discussions: Incorporate group discussions where learners debate trade-offs in data-driven decisions.
• Reflections: After every module, learners write a brief reflection on how they applied problem-solving or critical thinking to the tasks.
• Real-World Scenarios: Include case studies and encourage learners to think about ethical, social, and business implications of their decisions.
• Feedback Loops: Encourage iterative problem-solving by revisiting earlier modules with new insights.

Syllabus

Module 1: Understanding Data and Asking the Right Questions

Scenario: A city council wants to reduce traffic congestion. How can data help frame the problem and propose solutions? (note in this case, we only frame the problem - in terms of using the data to formulate the research question)

Learning Objectives:

• Understand the types and sources of data.
• Learn how to ask critical, actionable questions that data can answer.
• Explore data cleaning and validation as part of critical thinking.

Key Topics:

• Types of data: numerical, categorical, textual.
• Data quality: identifying bias, inconsistencies, and missing values.
• Framing problems into data-driven questions.

Problem-Solving Activity:

Given a messy dataset, identify what questions can and cannot be answered. Propose steps to clean the data.

Module 2: Decisions Using Machine Learning, Deep Learning, and Statistical Inference

Scenario: A bank is deciding which models to use for loan approvals. How would you approach the decision? Consider statistical, machine learning and deep learning approaches and their pros and cons

Learning Objectives:

• Distinguish between different methods for data-driven decision-making.
• Develop the ability to match the right analytical approach to a problem.
• Critically evaluate assumptions and limitations of each method.

Key Topics:

• Machine learning and deep learning basics.
• When to use statistical inference vs. machine learning.
• Limitations and biases in algorithms.

Problem-Solving Activity:

Analyze the problem and cross validate against the techniques of solving problems (machine learning, deep learning and statistical analysis). Consider the pros and cons of using the techniques and the tradeoffs of? complexity, interpretability, and accuracy.

Module 3: Feature Engineering and Creative Problem Solving

Scenario: A hiring platform uses features like education level and years of experience to recommend candidates. How could this approach create unfair decisions, and how would you address it?

Learning Objectives:

• Learn to think creatively about transforming raw data into useful features.
• Understand how features impact model performance and decisions.
• Develop critical thinking by evaluating the impact of irrelevant or biased features.

Key Topics:

• Transforming raw data into actionable features.
• Addressing ethical concerns in feature creation (e.g., fairness, bias).
• Automating feature selection.

Problem-Solving Activity:

Given a dataset of customer transactions, brainstorm features to predict customer churn. Discuss which features might introduce bias or ethical concerns.

Module 4: Building and Evaluating Machine Learning Pipelines

Scenario: A real estate company deploys a model to predict house prices but finds inaccurate predictions in rural areas. How would you debug and fix the pipeline?

Learning Objectives:

• Learn to critically analyze each step of the pipeline.
• Develop problem-solving skills to identify bottlenecks and improve efficiency.
• Evaluate trade-offs between accuracy, complexity, and interpretability.

Key Topics:

• Pipeline stages: data preprocessing, model building, and deployment.
• Handling errors and debugging pipelines.
• Automation and reproducibility in pipelines.

Problem-Solving Activity:

Identify potential failure points in a machine learning pipeline . Propose solutions to address these challenges.

Module 5: Model Evaluation Metrics and Decision Analysis

Scenario: A healthcare provider must choose between two models for disease risk prediction. How would you evaluate the models considering both accuracy and patient outcomes?

Learning Objectives:

• Develop the ability to choose the right evaluation metrics based on the problem.
• Critically assess how metrics influence decisions.
• Learn to analyze the trade-offs between model performance and business objectives.

Key Topics:

• Evaluation metrics for regression, classification, and clustering.
• The impact of false positives and false negatives on decisions.
• Balancing accuracy with interpretability and fairness.

Problem-Solving Activity:

Compare two models : one with high accuracy but many false positives, and another with lower accuracy but fewer false positives. Decide which is better and justify your choice.

Module 6: Improving Decisions from a Baseline

Scenario: An online retailer uses a baseline recommendation system. How would you improve it to increase customer engagement while ensuring fairness across demographics?

Learning Objectives:

• Develop critical thinking to compare baseline and advanced models.
• Learn strategies to iteratively improve models and decisions.
• Understand the role of domain knowledge in enhancing decisions.

Key Topics:

• Establishing a baseline (e.g., mean prediction or heuristic).
• Iterative improvement through feature refinement and hyperparameter tuning.
• Incorporating feedback from stakeholders.

Problem-Solving Activity:

Start with a simple baseline model. Propose and implement improvements step-by-step, justifying each decision.

These ideas are based on my teaching and experience at #universityofoxford. But I have never adapted them to a younger audience, I hope to develop these ideas in the Open domain. Many thanks to my colleague Anjali Jain at #universityofoxford??

Image - my hand made diary from fausta?which kind of looks like a syllabus :)