| title | layout |
|---|---|
Notes for Lecture 4 |
post |
The goal of research in programming languages is to come up with reusable insights and solutions. The problem is general and abstract. A problem space instead of a problem. Defining a language to express the problem, enable many different problems and many different solutions to be explored.
As an example, consider this paper by Bastani, Zhang and Solar-Lezama Probabilistic Verification of Fairness Properties via Concentration (forthcoming at OOPSLA 2019). It defines a probabilistic solution to the problem of deciding whether a black-box (such as a neural network) is fair. What is fair? They don't solve the problem for a particular definition of fair, but instead define a language for specifying fairness properties.
What is a language? A language has syntax and semantics. Any powerful general language has primitives, means of composition and means of abstraction. The semantics give meaning to the syntax, ideally denotationally -- by decomposing the meaning of a whole into the meaning of parts.
What is programming? It's manipulating processes that blindly manipulate symbols. A programming language can elevate the level of discourse beyond the mere manipulation of symbols.
For HCI, we need to think of UIs beyond just interaction primitives (such as clicks, gestures). How could they compose better? We can also think of making the programming process less like the processes that are programmed (blindly manipulating symbols!).