Scott Page, one of the two authors of
is offering an online course on modeling. It's free and, to me at least, it's fun. Besides being fun it promises to be a terrific introduction to a style of thinking that contrasts sharply and directly with the uses of thick description in ethnography and history. In pursuit of useful clarity modelers simplify, accepting the consequence described in statistician H. T. Box's notable quote, "All models are wrong. But some are useful." How they might be useful is the topic of Page's introductory lectures, which I have enjoyed immensely. Page is a gifted speaker, who combines great clarity with an easy, conversational tone.
I wonder if there is anyone on OAC who would like to join me in working through this course and discussing what Page has to say.
Well I've been working through the course since I posted it to Anthro-L. Its been a good course, and introduces a lot of different models. Let me know what you have in mind.
I had the pleasure to partake in his and John Miller's Santa Fe Institute Graduate Workshop in Computational Social Science, which was fantastic and yes, I'm in!
To date I have watched the lectures in the first section "Introduction: Why Model?" and the first three of the lectures in the next, the "Segregation and Peer Effects" section. I will likely be stuck at this point until 3/21. Day after tomorrow I fly to LA to participate in the 2012 INSNA Sunbelt conference. Then, leaving 3/19 and arriving in Japan on 3/20, I fly on to Japan. But please, don't wait on me. Charge ahead and tell us about what you think and feel about this course. I will do my best to catch up when I can.
I'm trying to follow a lot of lectures and fill the midterm exam in the next 24 hours.
I have known sometime in the past (almost) every theme but in different contexts: algebra for physics, statistics for sociology, agents in computer models, etc. It is a challenge to grasp the way in which Scott seems to manage such a smörgåsbord of models.
This is to say, John, that there is at least one interested in this matter.
Oscar, glad that you are finding the lectures interesting. I have slipped behind, overwhelmed with paying work at our company and preparation for a couple of upcoming conferences, one in Japan and one in China. What you say about the smörgåsbord is interesting. One of the things I like about both modeling as Scott presents it and the social network analysis I've been working on for the past few years is that the math is utterly indifferent to disciplinary boundaries and offers a genuinely new scientific cosmos to play in. In the case of network analysis, I find totally cool to be using techniques that work equally well when applied to social networks, power grids and protein cascades in cell biology. I find it equally intriguing that models of nuclear explosions work pretty well as models of bar fights or mob violence (the critical variable being the distribution of susceptibility to explosion thresholds — which is why nuclear reactors have damper rods inserted into the fissionable core to slow the nuclear reaction. One wonders how many riots could be prevented by infiltrating enough individuals who don't get terribly excited into the explosive mob. Perhaps a new model for policing. Lots of plain clothes cops calmly strolling through the crowd instead of phalanxes of what look like robocops that squeeze the mob and set up explosive situations....Anyway, lots of fund to think with.
Three or four lectures behind, but still proceeding. Unfortunately, like John, other things have kept me occupied.
It is certainly true, as John said, that formally similar models keep popping up in diverse domains. One can also begin to see higher-level relations between classes of models, e.g. between percolation and epidemiological models of various kinds.
@John : Perhaps doing without the robocops would be sufficient in most cases for things to remain calmer (e.g. Davis, CA; Oakland, CA). On the other hand, if a riot is ongoing, then maybe that would work, though I suspect that each lonely police officer would be terribly anxious not to be outed for fear of retribution of some sort from the folks in the crowd.
As with the nuclear reactor paradigm, it's a question of both the timing and scale of the intervention. One cop alone in an already angry crowd won't do the trick. Imagine a bunch of cops in civilian clothes joining the crowd before things heat up and not joining in when agitators try to start a riot...or sitting down as passive resisters do...or suddenly running in the opposite direction....Which would be the most effective tactic, and what proportion of the crowd would have to be on the cops' side? Its a tipping point question like the one in the standing ovation model. How many cops do you need? How many in a crowd not joining in the "spontaneous" action is enough to dampen the response of the crowd as a whole? Lots of interesting things to think about here.