Outcomes
What students should leave with
- Describe the standard machine-learning workflow from data preparation through evaluation.
- Interpret basic model performance metrics and explain tradeoffs in plain language.
- Recognize when prediction tasks need extra caution, review, or human oversight.
Reference map
Core topics to cover
Features and labels
Train/test split
Classification
Regression
Model evaluation
Bias and drift awareness
Human review
Module sequence
ICCT lesson flow
Module 1
What a model actually does
Lessons: 2
Ground students in the core workflow before they ever tune a model.
Features, labels, and data splits
how prediction tasks are framed and evaluated
Classification and regression basics
matching model type to the question being asked
Checkpoint lab
Frame a prediction problem around follow-up priorities and explain the target variable clearly.
Module 2
Reading model quality
Lessons: 2
Learn to judge whether a result is actually useful.
Accuracy, precision, recall, and error
metric choice and practical interpretation
Overfitting and validation discipline
why promising results can still be misleading
Checkpoint lab
Compare two small models and explain which one you would trust more in a school-operations scenario.
Module 3
Responsible application
Lessons: 2
Keep models in the role of decision support, not unchecked authority.
Bias, data quality, and review boundaries
where harm or weak assumptions can enter the workflow
Communicating predictions to stakeholders
confidence, uncertainty, and human oversight
Checkpoint lab
Write a model-use memo that explains when human review is mandatory and why.
Capstone prompts
Projects students can actually ship
Classification Starter Project
Train and evaluate a simple classification model, then explain the results in non-technical language.
Responsible Prediction Brief
Document the risks, limits, and approval rules for a proposed machine-learning workflow.