my research job at CMU

[gusl]

This is the project I am going to work for at CMU. The authors are my new team.

Applying Machine Learning to Cognitive Modeling for Cognitive Tutors

Applying Programming by Demonstration in an Intelligent Authoring Tool for Cognitive Tutors

Building Cognitive Tutors with Programming by Demonstration

What characterizes a better demonstration for cognitive modeling by demonstration?

Programming by demonstration (PD) hints at interesting issues of expressiveness. Programmers often think in terms of examples, and much (most?) of programming work is about trying to infer a logical expression that fits these examples ^{1 2}, so programming by demonstration can lighten this load.

Of course, there is also the "logical" approach to programming, in which one tries to write code that is provably correct: we try to satisfy abstract specifications, rather than test cases. The PD approach is to debug the null program until you have what you want ³, whereas the "logical" approach doesn't require testing: syntactic proofs ensure that the results will go through, even if we can't understand the proofs.

---

^{1 - didn't Polya write about this?
2 - isn't this what ILP is all about?
3 - this reminds me of Sussman's "problem solving by debugging almost-right plans"}