mood_congruence_factor multiplies a same-sign memory by 1.06 in a negative mood and 1.03 in a positive one. Mood-congruent recall itself is Bower (1981), but no citation in docs/foundations.md supports the asymmetry — it is one product's tuning.
bench/ experiment E puts it in a regime built to favour it: correct memories share the mood's sign 75% of the time. It is worth +0.012 MRR. It does not hurt, and it does not clearly earn an unexplained constant.
This is falsification item (4). Two acceptable outcomes:
- Find a regime where it matters, add it to
bench/, and the constant is justified.
- Show it never matters much, and collapse
congruence_negative_mood / congruence_positive_mood into one symmetric value — or remove the term.
Either is a real contribution. The current state — a constant nobody can defend and nobody has removed — is the worst of the three.
mood_congruence_factormultiplies a same-sign memory by1.06in a negative mood and1.03in a positive one. Mood-congruent recall itself is Bower (1981), but no citation in docs/foundations.md supports the asymmetry — it is one product's tuning.bench/experiment E puts it in a regime built to favour it: correct memories share the mood's sign 75% of the time. It is worth +0.012 MRR. It does not hurt, and it does not clearly earn an unexplained constant.This is falsification item (4). Two acceptable outcomes:
bench/, and the constant is justified.congruence_negative_mood/congruence_positive_moodinto one symmetric value — or remove the term.Either is a real contribution. The current state — a constant nobody can defend and nobody has removed — is the worst of the three.