Thinking, Fast and Slow

by Daniel Kahneman

The research suggests a surprising conclusion: to maximize predictive accuracy, final decisions should be left to formulas, especially in low-validity environments.