Instructor: Prof. Reza Arghandeh
Institution: Western Norway University of Applied Sciences (HVL), Norway
Course level: PhD/MS
This course introduces causal intelligence as a practical framework for moving beyond statistical prediction toward intervention, counterfactual reasoning, robust generalization, explainability, adaptivity, and decision-oriented AI.
Modern machine learning is highly effective at learning patterns from data. However, many scientific, engineering, and business problems require more than prediction. We often need to answer questions such as:
- What would happen if we changed a treatment, policy, process, or decision rule?
- Why did an outcome occur?
- What would have happened under a different action?
- Which intervention should be prioritized under limited resources?
- Will an AI model remain reliable when deployed in a new environment?
The central message of the course is:
Prediction tells us what is likely to happen. Causality helps us understand what would happen if we act, what would have happened otherwise, and why a system behaves as it does.
The course combines conceptual foundations, mathematical formulation, causal graphs, simulated examples, Python implementation, and project-based learning.
By the end of the course, students should be able to:
- Distinguish between association, prediction, intervention, and counterfactual reasoning.
- Explain why standard machine learning is not sufficient for causal decision-making.
- Formulate causal research questions from domain-specific problems.
- Construct and interpret Directed Acyclic Graphs.
- Understand causal discovery as a tool for proposing candidate causal structures.
- Estimate average and heterogeneous treatment effects using causal machine learning.
- Apply a disciplined causal workflow from research question to responsible interpretation.
- Explain how causal reasoning supports robust, explainable, and adaptive AI systems.
| Chapter | Title | Main Points |
|---|---|---|
| 1 | Foundations of Causal Intelligence | Prediction vs. causality; association, intervention, and counterfactuals; Pearl’s causal ladder; observational vs. experimental data; confounding and selection bias. |
| 2 | Causal Graphs and Discovery | DAGs as assumption maps; forks, chains, and colliders; d-separation; backdoor adjustment; causal discovery as graph refinement. |
| 3 | Causal Machine Learning | Potential outcomes; ATE, ATT, ATU, and CATE; meta-learners; heterogeneous treatment effects; Double Machine Learning. |
| 4 | Causal Intelligence in Practice | Causal question formulation; variable audit; candidate DAGs; identification strategy; estimation; robustness checks; responsible reporting. |
| 5 | Causal AI for Robust, Explainable, and Adaptive Systems | Distribution shift; shortcut learning; causal explainability; intervention effects; decision intelligence; adaptive AI systems. |
| Final Project | Course Project | Project description, deliverables, timeline, and assessment. |
The course notebooks include chapter-specific references and further reading.
The course material may be updated before or during teaching. Some examples, notebooks, references, and project instructions may change slightly to reflect student backgrounds, available time, and recent developments in causal AI.
If you use or adapt material from this course, please cite it as:
Arghandeh, R. (2026). Causal Intelligence Course: From Prediction to Reasoning to Decision Intelligence. Western Norway University of Applied Sciences (HVL). https://github.com/Ci2Lab/Causal-Intelligence-Course
This repository is intended for educational and research use. Please cite the course and instructor when reusing or adapting the material.