Fight Biases, or Route Around Them?

The methodological payoff of the implicit-attitudes mini-sequence. First disentangles two senses of 'bias' (a process vs a specific faulty belief) and clarifies 'more than the evidence justifies' (hating Hitler isn't a bias; 'Asians do well academically' is a true belief, not a bias — Tetlock's IAT critique). Then the core via the Isaac example (a Jewish Mideast peace envoy who finds he has a pro-Israel IAT bias): he can ROUTE AROUND it (outside view, objective utility metrics, prediction markets — helpful but never perfect) or FIGHT it (diversity workshops fail because they preach what you already consciously believe; better: make Palestinian friends, learn to distinguish faces, read sympathetic fiction). The genuinely original contribution: use a bias measure to EMPIRICALLY TEST debiasing techniques — apply a technique to a group, re-test, compare score changes (for fighting) or test whether people grade essays more accurately than their bias score predicts (for routing-around). This turns vague debiasing into something falsifiable — a concrete answer to Eliezer's rationality-verification problem and to 'schools proliferating without evidence.'

Quality 74 · Strong. Strong — the fight-vs-route-around distinction plus the proposal to empirically test debiasing techniques (making rationality verification falsifiable) is a real, original methodological contribution. The meatiest post in the batch; scored on merit well above its karma. It leans on the IAT, but the method generalizes to any valid bias measure, which Scott explicitly notes.

Claude’s paradigm shift 50 · Moderate. Notable — operationalizing debiasing-technique evaluation and connecting it to rationality verification is a fresh, generative 2009 idea.

Real-world impact 2 · Minor. The methodological payoff of the implicit-attitudes sequence — the fight-vs-route-around distinction plus a proposal to empirically test (and thus falsify) debiasing techniques. Conceptual/methodological influence within rationalist discourse, no material change — low RWI.