Friday, January 9, 2009

Modeling Terrorist Group Behavior: Hamas & Hezbollah

In my day job at the University of Maryland’s Laboratory for Computational Cultural Dynamics I work with a team of computer scientists and socials scientists to build models of terrorist group behavior. As the in-house TerrorWonk my role is to “interpret” the results and see if they yield any useful insights. I’ve co-authored papers on both Hezbollah and Hamas ( only the abstract is posted online).

The models use a system called SOMA (Stochastic Opponent Modeling Agents) that calculates probabilities of a group acting in a given way in a given situation.

Obviously, we hope that our models can achieve a high level of prediction accuracy. But, regardless they can often reveal facets of an organization’s behavior that were not previously evident. Just as military experts say, “Plans are nothing, planning is everything,” I heard one speaker at a conference say, “All models are wrong, but some are useful.”

Following are short summaries of the findings.

Hezbollah: The Model Holds

On hearing the news that rockets had been fired from Lebanon into Israel yesterday morning, I was surprised as the single clearest rule about Hezbollah behavior was that they do not like to target Israeli civilians during election years – and Lebanon has parliamentary elections coming up in April. Hezbollah even kept their rocket attacks down for 1998’s local elections. It turns out the model held, Hezbollah quickly distanced itself from the rocket launch and did not reply to Israeli counter-fire.

Extrapolating, this trend indicates how highly Hezbollah values its legal and political standing in Lebanon and its recognition that this standing is damaged when it is held responsible for provoking Israeli strikes. This provides a working explanation as to why Hezbollah had not renewed hostilities with Israel in the past few years – Lebanon’s presidential selection crisis, while not exactly an election, had some similar dynamics.

Beyond some utility for predicting (on an annual basis) when Hezbollah might launch rocket strikes, it provides real insight into Hezbollah and even a possible counter-strategy. While Hamas’ rhetoric remains stridently anti-Israel, the group is pressed by its need to satisfy its domestic constituency, the Lebanese Shia, who are a bit tired of being the Muslim world’s spearhead against Israel.

Military efforts against Hezbollah have not been effective, but these findings raise two related questions: how popular is Hezbollah really among its constituents and could political efforts against Hezbollah be successful at marginalizing them. It is worth noting that Hezbollah receives something like $100 million annually from Iran and the Lebanese Shia population is only about 1.5 million people, so these resources would buy a great deal of influence. Other Lebanese Shia groups do not possess comparable resources.

Hamas: Anything but Resistance is Futile

The results of our Hamas model were very different. Strategic decisions to reduce violence were not in evidence. The key driver appeared to be capability. First, the likelihood of suicide bombings (the data set does not include rocket attacks) increased after Hamas came into contact with Hezbollah in 1993, and received training in suicide terror. The other factor, which increased the likelihood of suicide attacks was Hamas’ provision of social services (which would seem counter-intuitive – but as Matt Levitt shows, the social services infrastructure is also a critical part of the terror network.)

Interestingly, when Hamas was participating in the Palestinian democratic process they were also very likely to carry out suicide attacks on Israel. However, the sample size was relatively small and capability seemed like a likelier explanation.

Perhaps the most interesting finding was that certain attacks, such as kidnappings and property attacks on Palestinians, tracked with internal Palestinian conflict. Although it occurred after the data was collected, the kidnapping of Gilad Shalit occurred during the Hamas-Fatah civil war. Another round of Hamas-Fatah fighting is likely in the West Bank, so more of these kinds of actions should be expected.

It could be argued that the 2006 war in Lebanon was a relative success – Hezbollah has kept that border quiet since. The likelihood of a similar modus vivendi with Hamas is Gaza seems less likely based on the model and also based on Hamas rhetoric. In an interview given just days before Hamas began launching rockets the deputy chief of Hamas’ Damascus wing stated:
[Your] question implies that the Tahdiah [truce] is a central issue behind [our] decisions, consultations, and mediation attempts. However, the opposite is true… [for us,] resistance is the main [element] in the relations between the Palestinian people and the Zionist occupation.
Reducing Hamas’ desire to commit violence does not seem possible, it is essential that Israeli strategy reduce their capability. In that regard, cutting the Hamas supply lines of the tunnels between Gaza and Egypt (and attacking the broader smuggling network) is critical.


Drew Conway said...


What is the formulation of the model? That is, have you designed a formal model of variables--like a regression--or are you using agent-based techniques.

I am very interested in learning more about SOMA, and would take your answers offline in e-mail.


Aaron Mannes said...

Let me confess, I am not the technical guy on the project. I think it is all magic. SOMA is an agent-based model, check the link embedded in the post, the papers we co-authored, and the LCCD site in general for more background.

Anonymous said...

Yes,but how do we terminate the vermin? Israel needs help.

AZZenny said...

Gilad Shalit was kidnapped June 2006, and the 'official' civil war between Hamas and Fatah was June 2007.
How much of this model is based on formal behavioral analysis -- schedules of reinforcement, negative reinforcement for Israel, etc? The linked piece makes it sound pretty much correlational and logistical regression-based.

Damian Lataan said...

There is something very faux about academics that posit their work as objective analysis yet use derogatory labels like ‘terrorist’ to identify their subjects.

You should remain politically neutral if you want to wear the academics cap to present your work.

If you take political sides – and it really doesn’t matter which side – and then attempt to present your work as ‘objective’ then you are being fraudulent. Your work, as in this case, becomes purely propaganda disguised as academia.

Philip I said...

Palestinians should convert to Judaism

Aaron Mannes said...

Hi all, as always thanks for your comments.

Gilad Shalit's kidnapping happened after Hamas won the Palestinian elections, so that while it was not height of the inter-Palestinian fighting there was a heightened Fatah-Hamas conflict and inter-organizational conflict did have a correlating with an increased likelihood of kidnapping.

While Hamas is a vicious organization, the connotations of calling them vermin is beneath us.

We can have a long (and fruitless) debate about the semantics of calling the group terrorists or not. But SOMA could be applied to any organization - but trying to figure out when groups might act violently is of particular interest at present. Regardless of my politics, I want to understand how Hamas operates and what triggers its behaviors.

As I mentioned before, I am not the technical guy on the project, the formal explanation is:

SOMA (Stochastic Opponent Modeling Agents) is a formal, logical-statistical reasoning language within which we can express knowledge about the behaviors of a group of interest, and compile a set of rules in such a language into an "agent." Within SOMA, we can express rules about the probability that a given person or group will act a certain way in a given situation.

Feel free to bring specific questions about the system to me offline.

Matt said...

Very cool, and thanks for sharing. I was curious, how much does game theory play into SOMA? Or does it? If this is a prediction model, then I imagine various aspects of game theory would be used. On my blog Feral Jundi, I get into some of this stuff. But I am not an academic. Most of my stuff is about security contracting, but I do like checking out new technologies.

On another note, you might get a kick out of another story I posted about Ushahidi--a SMS/Twitter/news collection and mapping program. Al Jazeera is using it for the latest stuff in Gaza.

Mirela said...

I recently came across your blog and have been reading along. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.