#acm2y — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #acm2y, aggregated by home.social.
-
Watching a talk by Titus Winters at the ACM2Y meeting, and he floats an interesting idea:
https://youtu.be/lzFxmYn9onM?si=FF-XxHt_asDlPBMV&t=2438
If tokens are a meaningful measure of human cognition and effort, then you can use them to empirically verify claims about thinking and mental effort.
For example, take a claim that making a certain harder change now will pay off in the long run. You could set up experiments: take a common starting point; have one system make the hard change now, the other just leave it. Then run both through an identical sequence of changes -- fix a bug; add a new feature. Refactor something. Upgrade to a new library API. Then just compare total number of tokens.
Of course, the initial assumption about a correlation between human cognitive load and tokens is...a pretty big one. And in the experiment, the final output may not work exactly the same, making a token-count comparison not quite apples-to-apples.
But we do this kind of stuff all the time -- look at fMRI studies, or EEG measurements. We know those are flawed -- think of the "dead salmon in the MRI study" -- but this seems like an interesting avenue.