[userpic]

... ↑

ailev в посте Openmeta (оригинал в ЖЖ)

Ссылка была на одну из ранних работ Лената (1984г., кажется). Я приведу тут полностью всю его модель постепенного формирования теорий:
The Accretion Model of Theory Formation
1. Given some new (not fully explored) definitions, objects, operations, rules, etc., immediately gather empirical data about them: find examples of them, try to apply them, etc.
2. As this progresses, try to notice regularities, patterns, and exceptions to patterns, in the data.
3. From these observations, form new hypotheses and modify old ones. In a world over which you have some control, design and carry out experiments to test these hypotheses.
4. As a body of conjectures develops, economize by making new definitions that shorten the statement of the most useful conjectures. The entire cyclic process now typically begins anew at step 1, given these new definitions as grist.
5. As the above loop (steps 1-4) proceeds, it will become necessary from time to time to abstract some new soecific heuristics, by compiling the learner's hindsight.
6. On even more rare occasions, it will become necessary to augment or shift the representation in which the domain knowledge is encoded.
7. For all steps in this model, even steps 5, 6, and 7, it suffices to collect and use a body of heuristics, informal judgmental rules which guide the explorer toward the most plausible alternatives and away from the most implausible ones.
А современные работы Лената нужно смотреть в статьях на www.cyc.com -- особенно интересные работы про контексты.

6 комментариев

сначала старые сначала новые