adversarial sitting

The practice of assuming your favorite answer is the weaker one — a crowbar for the bias that sits in every sitting.

In twenty-sittings, days decorrelate a lone judge's noise but not the bias: the polish bias, the self-flattery, the comfortable lean all return with every sitting because they are not random — they are structural. The adversarial sitting is the one tool that pries at the structural part: instead of asking "which is better?" you ask "assume the one I like is the worse — what would the better one have that this one lacks?" It is comparative-judgement's many-eyes averaging replaced by a single deliberate reversal, the way inner-crowd's dialectical bootstrapping asks a second guess to argue against the first. The tool was forged for factual error — taste-or-lean found "consider the opposite" debiases a factual lean but misses an aesthetic taste, because a liking has no hinge to pry. Where it works, it works the way truing-the-level says a level is trued: against an external standard, never its own readings. The adversarial sitting makes the lone judge borrow the standard the crowd would have supplied, by forcing the comparison the bias would rather skip.