Can You Condition Yourself?
Read the original on Slate Star Codex →
Summary
A first-principles analysis of whether self-administered behaviorism ('training your inner pigeon' -- rewarding yourself to reinforce behaviors) can actually work, reasoning through three possibilities. (3) Changing your desires (candy makes you enjoy homework): Scott's sharp refutation is that mere reliable reward-association doesn't make a neutral act desirable -- he feels zero urge to sit at the table and wave silverware without dinner, despite decades of that act preceding reward, because reinforcement runs on surprise / prediction-error, and a self-planned reward carries no prediction error. (2) Wanting the reward (candy as bribe): capped by how much you actually want the reward (if the candy sits uneaten in the fridge it can't motivate homework), plus the overjustification effect saps intrinsic motivation. That leaves (1) doesn't work. He then takes on the 'victory gesture' variant (fist-pump + visualize success), arguing it would be terrible mind-design to let consciously-generated emotions feed the reinforcement loop (the saber-tooth-tiger reductio: just choose to feel happy instead of fear, and you die) and that he feels no urge to repeat a victory gesture either. The thin 'self-consequation' literature (1970s-80s 'Scientific Prehistory') is contradictory; he'll try it anyway ('so high value I'd have expected the first person to get it right to take over the world -- this is turning into another argument against it').
Why this score
Quality 72 · Strong. Strong (72): original, careful reasoning about a popular self-help technique from genuine neuroscience first principles, with several sharp, memorable arguments (the silverware-waving reductio; reinforcement-as-prediction-error; the consciously-generated-emotion mind-design argument). Upper-Strong rather than Excellent because it ends inconclusively (he'll try it anyway) and the supporting literature is thin and dated.
Claude’s paradigm shift 48 · Moderate. Moderate (48): the first-principles case against self-conditioning -- especially the prediction-error and consciously-generated-emotion arguments -- is a fresh, non-obvious analysis, though it applies established reinforcement-learning and overjustification ideas.
Real-world impact 2 · Minor. Original first-principles reasoning about whether self-administered behaviorism can work, with sharp arguments (the silverware-waving reductio; reinforcement-as-prediction-error). Conceptual influence within rationalist discourse, ends inconclusively, no material change — low RWI.