What Lies Beyond Machine Learning and AI: Decision Systems and the Future of Data Teams

In this episode of High Signal, Chris Wiggins—Chief Data Scientist at The New York Times and Professor at Columbia University—discusses moving beyond machine learning to build robust decision systems that drive real-world outcomes. Drawing on his experience scaling data science teams, Chris explores the shift from prediction to prescription, the role of interventions in understanding causality, and what it takes to integrate data science into large organizations.

Key Takeaways**

• From Prediction to Prescription: Driving Action with Data Science**

• Prediction is not the end goal—true value comes from prescribing the right interventions** to influence outcomes.

• Chris states: “The question is not just how you can predict what's going to happen in the absence of treatment. The question is how can you prescribe the optimal treatment in order to drive some event that you want.”* 

• Building Data Science Teams That Align with Organizational Goals**

• Success in data science requires empathy, communication, and alignment with partner teams across the organization.

• Chris highlights: “Other teams are the keys to getting things done at scale.”* He stresses that understanding others’ goals drives adoption and impact.

• The AI Hierarchy of Needs: Focus on Fundamentals Before AI**

• Start with strong data engineering, logging, and observability** before building advanced models.

• Chris underscores: “If the input falls down and there’s no observability, what’s the point of fancy AI?”* Citing Monica Rogati’s framework, he emphasizes the importance of reliable foundations.

• Empathy as a Critical Skill for Data Scientists**

• Data scientists must communicate the value of their work** clearly to ensure organizational alignment.

• Chris advises: “Make sure that everything you're working on, you could explain to anybody in the organization why it's an important thing to work on.”* This clarity drives trust and buy-in.

• From Proofs-of-Concept to Production Systems**

• One-off prototypes may be provocative, but the real challenge lies in systems integration** and scaling solutions across the organization.

• Chris explains: “Doing one-off fancy artificial intelligence can be very useful as a provocation... But it’s not the same thing as a prototype, build one complete solution, then work on systems integration and process integration.”* 

You can read the full transcript here.

00:00 Introduction to Chris Wiggins' Journey

00:07 Building Data Functions at The New York Times

00:48 Early Challenges and Evolution

01:07 The Importance of Prescriptive Analytics

02:07 Optimization and Personalization

03:25 AI Hierarchy of Needs

04:01 Effective Data Science Teams

08:02 Chris Wiggins' Career Journey

14:01 Building a Data Function at The New York Times

16:44 Predictive to Prescriptive Analytics

22:03 History of Data Science

23:46 Statistics and Power

27:00 Building a Data Function: Practical Insights

39:42 Choosing the Right Tool for the Job in Machine Learning

40:35 Causal Inference and Reinforcement Learning

42:02 The Importance of Randomized Control Trials

43:59 Principal Component Analysis and Data Function Priorities

46:47 Empathy in Data Leadership

52:42 Generative AI in Education

58:05 Interdisciplinary Collaboration in Academia and Industry

01:10:34 Future of Data Science at The New York Times

01:15:11 Closing Thoughts and Advice for Data Science Leaders