Watson for Oncology: Memorial Sloan Kettering’s “bias” [Operational Drift]

Watson for Oncology was sold as a system that could “digest” massive data and guide cancer care, but its recommendations were trained by physicians at Memorial Sloan Kettering Cancer Center—meaning the system could scale a specific set of human preferences while being perceived as global, automated intelligence. This episode documents how that gap formed, how it was normalized in marketing and hospital adoption, and how the absence of independent, third-party evaluation left accountability unclear once the tool was already in use.

Topics Covered

• 🔍 What Watson for Oncology actually did versus what it was said to do
• 📋 How training by one hospital shaped “recommendations” worldwide
• ⚖️ Optional oversight: no requirement for clinical trials before sale
• 🔬 “Concordance” studies and what they do not prove
• 🏥 Workflow reality: costs, integration, and how hospitals used it

Neural Newscast is AI-assisted, human reviewed. View our AI Transparency Policy at NeuralNewscast.com.

• (00:12) - Introduction
• (00:12) - Act I — The Signal: “Worldwide consultation” that was not worldwide
• (00:12) - Act II — The Drift: Training, bias, and optional evaluation
• (00:12) - Act III — Authority without outcomes: concordance and deployment
• (00:12) - Conclusion