"what is 17 × 24?"
sample 4 answers
o₁ 408 ✓ score 1.0
o₂ 391 ✗ score 0.0
o₃ 408 ✓ score 1.0
o₄ ~400 ≈ score 0.5
average the scores
(1.0 + 0.0 + 1.0 + 0.5) ÷ 4 = 0.625
advantage = score − average
1.0 − 0.625 +0.38 reinforce ↑
0.0 − 0.625 −0.63 suppress ↓
1.0 − 0.625 +0.38 reinforce ↑
0.5 − 0.625 −0.13 suppress ↓
update weights: more of o₁, o₃  ·  less of o₂, o₄