in popco.communic.speak, choose-propn-ids-to-say calls worth-saying-ids to create a list of propositions that the person is allowed to say, because they're not masked, they're allowed to be uttered at all, and because a random number is less than the absolute value of their activations. Then choose-propn-ids-to-say samples from whatever propositions pass all of these tests. Is this the most efficient method? Is it really necessary to calculate all possible propositions for uttering before sampling from them? Usually, only a single proposition will be sampled. Further, worth-saying-ids loops through every proposition, even if it's not one that could possibly be uttered. The ones that have a zero value after the filtering process get ignored, but still, the code ends up checking them anyway--and generates a random number for each.
in
popco.communic.speak,choose-propn-ids-to-saycallsworth-saying-idsto create a list of propositions that the person is allowed to say, because they're not masked, they're allowed to be uttered at all, and because a random number is less than the absolute value of their activations. Thenchoose-propn-ids-to-saysamples from whatever propositions pass all of these tests. Is this the most efficient method? Is it really necessary to calculate all possible propositions for uttering before sampling from them? Usually, only a single proposition will be sampled. Further,worth-saying-idsloops through every proposition, even if it's not one that could possibly be uttered. The ones that have a zero value after the filtering process get ignored, but still, the code ends up checking them anyway--and generates a random number for each.