From Prediction to Causation: Causal AI for Real-World Data

As artificial intelligence and machine learning continue to transform research and industry, there is a growing need to distinguish between prediction and causation. While predictive models can forecast outcomes with impressive accuracy, many real-world decisions require a deeper question: How would outcomes change if we intervened?

To address this, the Statistical Society of Canada (SSC), through its Accreditation Services Professional Development Initiative, recently hosted the workshop “Causal AI for Real-World Data” on January 19, 2026. The session was led by Andy Wilson and Meghann Gregg, recognized independent thought leaders in the real-world evidence space. The event was open to a broad audience and free to SSC members. Approximately 25 participants attended, representing a mix of researchers and applied practitioners with an interest in strengthening their causal inference toolkit.

Moving Beyond Prediction

The workshop was designed for those working with real-world data who want to move beyond predictive modelling towards credible cause-and-effect estimation. We began with foundational principles of causal inference, including:

• randomized controlled trials as the gold standard for identification;
• natural experiments and quasi-experimental designs;
• core assumptions underlying causal effect estimation.

By grounding the discussion in identification and study design, the workshop emphasized that causal reasoning starts with careful thinking about counterfactuals, not with algorithms.

Modern Tools for Complex Data

Building on these foundations, the workshop introduced modern methods that combine causal inference principles with machine learning techniques. Topics included:

• double machine learning for high-dimensional confounding adjustment;
• causal forests for heterogeneous treatment effect estimation;
• synthetic control methods for comparative case studies.

Participants engaged in hands-on coding demonstrations, primarily in R, working through applied examples that illustrated both the promise and the limitations of these approaches. The emphasis throughout was practical: how to implement these tools correctly, diagnostics to examine assumptions, and responsible interpretation of results.

Real-World Engagement

Participants asked thoughtful questions about applying causal methods in real-world settings, including observational health data, policy evaluation, and industry analytics. Discussions frequently returned to issues of data quality, unmeasured confounding, and the tension between methodological rigour and operational constraints.

Several attendees later reached out to organizers by email and LinkedIn independently to share that they found the workshop valuable and directly relevant to their work. That kind of follow-up underscores the demand for applied, methodologically grounded training in causal inference.

Accreditation and Professional Development

This workshop reflects SSC’s commitment to supporting professional development in areas that are rapidly evolving and central to modern statistical practice. As machine learning tools become increasingly accessible, it is essential that practitioners also strengthen their understanding of identification, study design, and causal assumptions.

By integrating foundational causal inference with contemporary computational tools, the workshop aimed to equip participants not only with new methods, but with a principled framework for evaluating when and how to use them.

We look forward to continuing to develop the Accreditation Services Professional Development Initiative to help bridging theory and practice, and that supporting statisticians in leading the responsible use of data-driven methods in real-world decision-making.