-
I don’t think this is a fair account of what ABM really are. An ABM is a collection of chunks of code that we say are "agents". Agents can be humans, animals, particles, everything we want. It’s up to the modeler to tell the story she/he wants. 1/8 @ethanbdm/1107339702233899010Permalink
❤️ 3 Favorites
Mood +3 🙂
-
Every agent is defined by rules of behavior and parameters. Heterogeneity can arise from both. An ABM (at least for a computer scientist) is letting all those agents interact and observing the result - as in a computational Petri dish. 2/Permalink
Mood 0
-
To me, at least with the most recent libraries you can definitely track down agents for every period of the simulation, and know why they did what they did (providing the model is parcimonious enough, aka all ABMS aren’t *good* ABMS). 3/Permalink
Mood +3 🙂
-
As long as you know the behaviors and parameters of agents at every period of the model, maybe we have a different definition of "seeing through" but I would say we do "see through". Even though I agree equation based models are more "transparent" somewhat. 4/Permalink
Mood +1 🙂
-
But I don’t think it’s 0 or 1. Some EBM can be really hard to see through as well. 5/Permalink
Mood -1 🙁
-
My last (and a bit lateral) point would be to say that having the Jupiter’s eye in a computer simulation is better than not having it at all. You have the first best (a clear model), but then you have the second best. 6/Permalink
Mood +9 🙂
-
In my view, non-trivial degrees of heterogeneity among agents can quickly make EBM impossible to solve - at this point, it’s then irrelevant to "see through" because your model just can’t exist. ABMs can answer that, pretty cheaply. 7/
Mood +2 🙂
-
So even if ABMs were actually not "intentional" enough, I don’t think it’s a good reason to dismiss them as in (probably a lot more than we think) occasions, they are the only model that exists. And I think an "imperfect" model is better than no model at all. 8/8Permalink
Mood +2 🙂