On This Problem Rational People Do Worse

Quick Overview

The core of Newcomb's paradox hinges on whether one prioritizes evidential decision theory (one-boxing to secure the predicted $1 million) or causal decision theory (two-boxing because the past decision cannot be influenced, guaranteeing an extra $1,000), revealing that rationality at the individual act level (two-box) can conflict with rationality at the systemic or life-rule level (one-box via pre-commitment).

Key Points: Newcomb's paradox divides people almost evenly, with 53.5% of a 2016 Guardian poll choosing one box and 46.5% choosing two boxes. One-boxers follow evidential decision theory, relying on the computer's proven accuracy ("the thousands of people that it accurately predicted before me is evidence enough") to expect $1 million. Two-boxers follow causal decision theory, arguing that since the boxes are already set, their current choice cannot influence the past, meaning they always secure an extra $1,000 over the one-box outcome. The optimal outcome, gaining $1,001,000, requires the seemingly impossible act of pre-committing to one-box but then choosing two-box at the last second, which the predictor would anticipate. The discussion analogizes the paradox to the Prisoner's Dilemma, suggesting that while the rational act is two-boxing, a rational society would be full of one-boxers (cooperators). The speaker suggests the only way to guarantee the $1 million is through pre-commitment, similar to the strategy of Mutually Assured Destruction (MAD) or the doomsday device in Dr. Strangelove.

Context: The video analyzes Newcomb's paradox, a philosophical problem involving a choice between two boxes: one containing $1,000 openly, and a mystery box predicted by a supercomputer to contain either $1 million (if the subject chooses only the mystery box) or nothing (if the subject chooses both boxes). The paradox centers on the conflict between two schools of decision theory—evidential and causal—which leads rational individuals to arrive at diametrically opposed conclusions about the optimal choice.