Chapter 2: Conversation and Computers (pages 13-25...
# reading-together
p
Chapter 2: Conversation and Computers (pages 13-25) A Small Matter of Programming by Bonnie A. Nardi
It's pretty neat how current this book reads being published in 1993. I like Nardi's counter examples of driving cars and playing music as alternatives to natural language ("mundane conversation") as the default medium of human machine communication. Reminds me of something I read recently in Make It Stick—that there is no empirical support for the claim that instruction in one’s preferred learning style (e.g. "I'm a visual learner") improves learning outcomes. It's the subject matter that dictates the best medium, e.g. diagrams for geometry, words for poetry
🍰 1
j
Definitely an interesting chapter, though I struggled a bit with some of the linguistics analysis. We know now that computers are capable of mundane conversation, so I feel like there are some follow up questions worth asking
Nardi seems to suggest that programs require too much precision to be described without formal semantics. But we often talk to each other about how code works (e.g. specifications aren’t written in code, but are an informal means of meta-programming). I there agree that code is a good direct interface with the computer, but for many of the computer users who want to program but don’t know how to code, I feel that conversational model in the AI era has a lot more promise than this chapter foresaw.
p
Yeah, the chapter felt like the AI chapter with Nardi as the AI skeptic. I am craving a blog post or something with her thoughts 30 years later on how the chapter aged (looks like she hasn't blogged in over a decade though, so I'm not getting my hopes up haha).
d
Make it stick was pretty focused on learning/retention rather than communicating/understanding. In teaching/communicating, understanding the shared context is needed to maximize efficiency (which is why tutoring/sales generally has better outcomes than class/marketing) Related, I think we can talk about how code works with lower precision because we have a lot of shared context, which she stressed in other parts of that chapter. We'd have very different conversations with someone non technical (different shared context and probably a different model). I believe the lack of blog posts can be explained by her retirement in 2018
👍 1