What Are We Arguing About When We Argue About Rationality?

Prompted by the Pinker-vs-Gardner spat, Scott asks what 'anti-rationalism' could even mean, and tries four readings. (1) Rationality as full explicit computation vs heuristics -- rejected, since nobody (including self-styled rationalists) actually opposes heuristics; everyone runs a combination strategy. (2) Explicit computation vs intuition -- also rejected; intuition is trainable pattern-matching (now literally math, as cat/dog and style-transfer AIs show), and even the most explicit modelers argue on intuitive grounds about how to build the model. (3) Yudkowsky's 'rationality is systematized winning' -- clarifies that heuristics/intuition are fine, but makes being pro-rationality nearly vacuous. (4) The one Scott favours: rationality as the study of truth-seeking -- the 'study of study', easily confused with its object level (geology studies rocks; rationality studies studying). On this reading legibility is a rule of the game (like science: you haven't 'scientifically explained' something until others can build on it), which is why the field tilts toward the explicit side and why the best superforecaster need not be the best theorist of rationality.

Quality 70 · Strong. Strong floor. Thoughtful meta-rationality with a couple of genuinely nice handles (the diamond-prospector 'knack vs geology' and 'study of study' framings). Held at the floor because it is exploratory and deliberately inconclusive -- it surveys four definitions without firmly resolving which is right.

Claude’s paradigm shift 50 · Moderate. Moderate. The component pieces (systematized winning, intuition-as-pattern-matching, legibility-as-norm) are in circulation; the fresh synthesis is framing rationality as the meta-level study of truth-seeking, which explains the field's explicit-side tilt.

Real-world impact 2 · Minor. Minor. Within the rationalist epistemics discourse; conceptual only. 2.