1. How much people like or hate you depends on their estimates of your features.His conclusion is that an agent acting in this model should open up when first meeting somebody, but if they become friends, there is no longer any incentive to discuss passions and life-goals.
2. You don't know who likes or hates what bundles of features, so you don't know who will like or hate the things they learn about you.
3. Whether you are above or below certain friend and enemy thresholds matters more than how far above or below those thresholds you are.
4. You can do certain things to reveal more or less about yourself to people.
5. Strangers start out giving you a middle estimate, of neither a friend nor an enemy.
My intuition is that assumption #3 will not hold unless the situation is strictly social. There are some people towards whom our desire to impress will almost never drop off. For example, even if Michael Arrington were already my friend, I would still want him to become an even better friend so that he will link to me from Tech Crunch and boost my Technorati ranking. Even if the CEO of the company I work for is already my friend, I will still want to impress him so that I am promoted.
Of these assumptions, which are the ones that you think will hold up worst with respect to reality?