1/ I've heard this theory a lot in the last few years. I believe it's true. Recently, I've had a lot of fearful moments with RL work thinking, "they're just going to eat our lunch, aren't they?"https://twitter.com/seanjtaylor/status/1143578006721638400 …
-
Show this thread
-
2/ If a researcher with a lot of data can induce empirically-grounded agent behaviors and strategies and stick them into a multi-agent simulation, is CSS obviated?
1 reply 0 retweets 0 likesShow this thread -
3/ I haven't really answered that question, but I think my take is wrong. CSS is more about moving back and forth between levels of analysis and earning your intuition for doing so. RL-powered ABMs may be a next methodological step. But, not an obsolescence inducing one.
1 reply 0 retweets 1 likeShow this thread -
4/ Now, I'm mostly excited about doing a lot more RL/ABM fusion work. Something I'm sure
@seanjtaylor has had the pleasure of doing at a serious level :)1 reply 0 retweets 0 likesShow this thread -
Replying to @generativist @seanjtaylor
Why do you think RL will invalidate ABM? They are different views of the same thing. ABM is bottom up, RL is top-down and often quite brittle. One thing to to also note is that for non-trivial RL, just as much hand-tuning is put into the reward/scorer as is required for ABM.
2 replies 0 retweets 1 like
'Obviated' was the wrong word. A better phrase would have been "crowded out" in the competition for tools and tool-builders. But yes, fundamentally, I agree with you: they are complementary methods.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.