Monday, December 8, 2008

Liberal Arts education, and AI

This journal article about Great Books sparked a conversation about the value of the Western canon to higher education which found me alone on the defense, and most of the elite college grads sitting around the lunch table unfamiliar with the concept.

We had some commentary about the value of cultural context, specialization, and vocational studies. In hindsight, I think the question got bogged down with assumptions about both universities and students. 
  1. Universities currently offer all sorts of classes, and students should be taking those classes because at some margin they are all worthwhile (even an otherwise bad class with few students will become attractive for some students who are particularly benefited by direct interaction with professors). So it's hard to make a case for one class versus another.
  2. The students were represented by anecdotes, or by vague assumptions about what young people want to study.
These are my alternatives, which I only came up with after the fact, for facilitating the discussion about the study of a canon as part of a liberal arts education.
  1. Regardless of the varying levels of merit across current curriculums, this question is about designing a new university by allocating resources to a new set of departments(if any), professors, and courses. Covering everything, even if it is desirable for an undergraduate institution, is not a valid or relevant solution. 
  2. There's something peculiarly difficult about discussing what people ought to be studying. It makes things easier, for me anyway, to think about artificial intelligence instead. Let's say somebody else had built an AI, and I was trying to train it to be able to engage in conversation on a wide range of topics, hold gainful employment (while telecommuting, of course), and in general be able to interact successfully with the world. My resources for teaching the AI would be limited by its design to a few hundred hours of lectures, a few hundred books, periodic questions to guide study of the material, and feedback about how well the AI was able to apply knowledge when tested. Information for the AI would be selected according to how efficiently it maximized the objectives. To the extent that the canon is related to itself, studying parts of it will increase the payoff of studying other parts of it. What is Dante without his cultural and historical context? Likewise, to the extent that the canon has developed as a part of Western Civilization, being familiar with it will facilitate interactions with political, religious, and intellectual phenomenons such as democracy, monotheism, and the scientific method, respectively. Merely developing critical reasoning by studying Mayan basket weaving, or feminist literary theory, even at the doctoral level, will fail to have the same payoffs. While there may be some value to studying any random topic, value is not maximized by randomness given time constraints which only allow a tiny fraction of human knowledge to be studied.
There's an element of critical mass involved. Anything's worth studying if everybody woth socializing and doing business with is also interested in it. However, lacking an alternative, even a neglected canon serves this purpose.

The idea for reasoning about learning by imagining artificial intelligence was derived from Overcoming Bias.



No comments: