Given the recent concerns on campus about mental health, I wonder what the implications of these theories are for treatment. A place like Stanford is perceived by students as a very high-stakes game. Placing a lot of intelligent people into a competitive environment is bound to produce threat, not just for people who identify as members of groups with negative stereotypes about their performance, but for anyone who might identify as possibly “not good enough” for Stanford. Performing well in school, graduating, getting a good job… all of these things are imbued with such importance by the pervading culture on campus that even an optimistic student will have a crash of confidence if they perceive enough setbacks. And setbacks are almost destined to occur in a situation where there is stereotype threat.

Enough with the rant. That reading also reminded me of an AI thing from 221, which describes a clever means to do planning for a computer called a Markov Decision Process. Essentially, you assign goals a positive weight, depending on how awesome they are, and antigoals a negative weight. Then, for each state touching a goal, you set the value of that state to be the expectation of moving from the state. (That is, if you have a .5 chance to get -1 and a.5 chance to get a 2, you end up with a 1.) You keep recursing until you get the start state’s value. The similarity I think is clear: anti-goals are just goals with negative value, and you don’t need to distinguish between them at all in this model.

Things actually get a lot more complicated: you might not even know how far you are from your goal, and you have to figure it out yourself! Here, the more you act, the more you learn about where you are, and thus about how to get to where you want to be and avoid where you don’t want to be. Thus, you get the same sort of feedback loop that they’re describing, since disturbances and actions affect your internal state, as well as your plans for attaining your goals (and the value of those goals!).

Under this model, though, you don’t need to have their weird assertion that an anti-goal will usually end up giving rise to a goal. They gave the example of a rebellious adolescent who avoids his parents and finds a group of like minded friends. To my estimation, that’s two disjoint goals that don’t need to be related: one can want friends, and one can want to avoid one’s parents. Add some homophily (Erich’s and my topic!) and you get the expected behavior: avoid parents and run to a group of friends who also hate their parents.