We–instructors and students–come to this course with a wide range of
beliefs and convictions about large language models (LLMs) like
Claude, ChatGPT, Gemini, etc.
Putting aside concerns about the future of humanity, here’s where we are
at in Fall 2026 for R-DAVIS: Despite all that these tools remain bad (or
let’s face it, just not yet good) at, the fact is that they can
easily handle the level of basic coding expected by R-DAVIS
with a little bit of context and prompt engineering (and an AI agent
could definitely complete this course and earn an A). So what’s
the point of this class?
The point of this class is only sort of to learn to
implement specific functions and techniques in R. The downside of AI’s
coding upside is that nobody is going to hire you because you know entry
level programming anyways. Rather, the point of this class is to
learn how to think about and solve data-driven problems. R, and
other object oriented-programming languages like Python,
Java, Ruby, or C++, represent a
completely different way of thinking and organizing your work. Whether
you end up in a role where you do not, or cannot, use AI for coding, you
go on to rely heavily on AI for coding in the future, or you never
code again, we firmly believe that gaining facility with
thinking in an object-oriented programming language will make you better
at your work. But that requires practicing the thinking part. To that
end, we believe it is a waste of your time, and our time, to turn in
AI-generated deliverables in R-DAVIS.
To some extent, AI use is unavoidable. A core R-DAVIS mantra is
“Google it!”, which now produces AI-generated results at the top. But
Claude Code or Copilot-type situations, or
uploading the assignment into chatGPT, will not help you
learn (and hinders our ability as instructors since we have no way to
tell how things are going for you). Whether the AI generated products
are right or wrong doesn’t really matter; programming might
feel fact-based, as if what you need to do is memorize a set of
commands or functions. But it is not. Programming is practice-based–you
learn by doing, not observing. So having AI handle the programming works
for learning about as well as when I try to learn to jump my
bike by watching YouTube videos.
Finally, separate from the instrumental value of R and
object-oriented programming in general, we encourage you to think about
this as an opportunity to practice learning, or to “learn how to learn.”
Being a graduate student at UC Davis likely means you have been good at
achievement for a long time; you have no doubt honed your skills. But it
might have been a long time since you tried to learn something from
scratch, and started from a place of true confusion and suckitude.
Learning to program will make you feel stupid (at times) and perhaps
seed doubt (“I’m just not someone who can do this.”). You aren’t
(stupid) and you are (someone who can do this). We will help you get
there. And we believe that this experience will pay immense
dividends for learning other things here in graduate school or in your
next job.
Finally, the warning: if in reviewing a submission we believe we are reviewing AI-generated work, we’ll ask to have a chat with you and see what’s up and how we can better help you. If submitted deliverables exhibit hallmarks of outsourced work and one-click “prompt-download-submit” behavior, we will deal with that as an academic code of conduct issue (misrepresenting others’ work as your own).