Agent-Based Modeling: Computation and Cost?

Agent-based modeling (ABM) can do ANYTHING — the basic claim being made by Eric Bonabeau in his article, Agent-based modeling: Methods and techniques for simulating human systems.  And indeed, it does appear that ABM is quite useful, particularly when examining heterogeneous populations, as we can see in “virtually every example in this article”, to quote the author himself.  While I still wonder about the validity of ABM in certain situations, and can’t help but feel unsure about the authors’ exuberant claims in his writing, there was one thing particularly that I found missing from this article: computation and cost.

While Bonabeau does devote one or two sentences at the very end of the article to the high level of computational power required for these types of models, he does not, in my opinion, adequately express not only how important this one factor may be, but also all the additional factors inherent with data-heavy models such as this.  For example, he makes no reference to the amount of data collection that must go into creating these models.  Even a basic GIS user understands that a superficial layer of data is not interesting, but anything more than that requires a lot of commitment to collecting data.  In this case, working with human systems, to me that implies surveying people about their behaviours, how they make decisions, and so on.  This means time and monetary commitment.  And this leads to my larger criticism: the most telling aspect was how the companies he referred to were primarily established, and I would assume, wealthy, companies or organizations who could afford to use ABMs to make better management decisions.  Despite this, nowhere does he discuss cost.  Surely this technology does not come cheap?  And if it does, wouldn’t that make it even more desirable, and worthwhile to include?

With this knowledge, the reader (and potential user) could make a more informed decision about if ABM is not only useful, but at all possible, for them.  In the end, an interesting overview of applications of ABM, but lacking in answers to a few important questions.

Bonabeau, Eric. “Agent-based Modeling: Methods and Techniques for Simulating Human Systems.” Proceedings of the National Academy of Sciences of the United States of America. 99.10 (2002): 7280-7287. Print.


Tags: , ,

One Response to “Agent-Based Modeling: Computation and Cost?”

  1. Milo_Aukerman says:

    Nice reflection – I really enjoyed your thoughts on the cost of ABM (whether the tools or the data). Glad to see you caught the over-excitedness in the article.