Source: pp 79-80, Think: A Compelling Introduction to Philosophy (1 ed, 1999) by Simon Blackburn
The most popular current approach to this is to concentrate upon the way in which we can attribute thoughts to the well-functioning person. It should be something about a person's behaviour that enables us to interpret him or her as thinking about yesterday, or concentrating upon the weather predicted for the weekend. Thoughts are expressed in both linguistic and non-linguistic behaviour, and perhaps we can hope for some kind of reduction:
[1.] "X thinks that p" if and only if
[2.] X's plans or desires or behaviour are somehow in line with the world being such that p.
The trick would be to fill out the "somehow in line". It is fair to say that nobody has successfully done that. But there are suggestions about how to go. We say that an intelligent system, such as a guided missile, thinks that there is a plane a mile away and two hundred feet up if its systems point it in a direction that is appropriate to there being a plane in that place -- given its aim (or function) of bringing down planes.
[3.] Similarly we might say of a person that she thinks the weather will be fine at the weekend if her behaviour is appropriate, given her aims (or functions), to that being the weather at the weekend.[4.] The difficulty would be to fill out this thought without relying in other ways on other mental states of the subject, and this is what nobody knows how to do.
Though seeing 'somehow in line' in [2], I do not understand [4]. What is difficult?
I exemplify with 3. The adjective 'fine' is unclear; so suppose that we ask that person and then know her definition of fine (eg: 24°C = 75.2°F, 55% relative humidity, clear skies, etc...)
and her aims (or functions) (eg: canoeing in a wildlife area). Then how is 4 true?
I think the problem (generally, considering your other questions) is that you are trying to parse these texts with formal logic and definitions only. But that is not how philosophical writing works. There always is a content in some kind of meta-textual level, the context of the propositions
? Or I can post a separate question.