TB: Reference Class Forecasting

Make predictions by starting from base rates of a reference class of similar projects or events, then adjust for the specific features of the current case. Developed by Daniel Kahneman and Amos Tversky; advocated by Philip Tetlock. Counters the "planning fallacy" by anchoring on the outside view (what happens to projects like this?) before allowing the inside view (what's special about our project?).

The planning fallacy is the tendency to underestimate costs, time, and risk of future actions while overestimating benefits. It stems from focusing on the specific plan (inside view) rather than on the distribution of outcomes across similar past endeavors (outside view). Reference class forecasting is the systematic correction: find the relevant reference class, look at the historical distribution, anchor your forecast there, then adjust by specific features of your case. The result consistently outperforms intuitive planning. The inside view should only be a small adjustment to the outside view base rate, not the primary estimate.

• The Black Swan (Taleb): the planning fallacy as a canonical instance of ignoring base rates; Taleb's critique of Gaussian models of uncertainty applies here — using the reference class distribution rather than building a model from first principles
• Seeking Wisdom (Bevelin): Bevelin's treatment of base rates as a corrective to inside-view thinking; "what is the base rate of success for projects like this?" as a checklist item

• Survivorship Bias — reference classes must include failures, not just successes
• Second-Order Thinking — the outside view asks "what usually happens next?" which is exactly second-order thinking applied to forecasting
• Pre-mortem — pre-mortem and reference class forecasting are complementary: the former finds specific failure modes, the latter anchors the overall probability of failure