EMERGENT INTELLIGENCE IN OPEN SOURCE WARFARE

Open source warfare, like what we see in Iraq and increasingly in other locations, relies on networks of peers rather than the hierarchies of command and control we see in conventional militaries. This structure provides an open source movement with levels of innovation and resilience that rigid hierarchies can't match. Unfortunately, these attributes are likely not constrained to merely local tactical activity. Open source movements can exhibit emergent intelligence that guides the movement's collective actions towards strategic goals.

Emergent Intelligence

What is emergent intelligence? It is a form of macrointelligence that arises from local interactions. It isn't merely the simple stigmergic interactions (Stigmergy is a term used in biology -- from the work of french biologist Pierre-Paul Grasse -- to describe environmental mechanisms for coordinating the work of independent actors) necessary for the coordination of the swarm activities of local autonomous agents. Rather, it is a form of group intelligence that learns, achieves goals, and engages in self-preservation. There are five simple requirements for emergent intelligence (a good starting point for those that want to dive deeper into this subject is Steven Johnson's book "Emergence"):

• A critical mass of participation is necessary. A certain minimum number of participants, either individuals or component groups, is necessary for micromotives to translate into macromotives. It also means that without a minimum number of interactions between these participants, the statistical nature of macrointelligence won't emerge. More is different.
• A local focus is useful. Open source actors are mainly focused on local activity. The simplicity of this focus is a feature and not a bug since it prevents activity that may upset the entire organism's operation. Local action, global impact.
• Random interactions are necessary. Random interactions between groups and individuals outside is a necessity (this assumes a certain level of mobility and communications capability). These interactions provide a fluidity to the learning process that finds and responds to new information quickly.
• Pattern matching from stigmergic communication. The ability to see patterns of activity from simple signs is a necessity. Gradients of activity provide "maps" to areas of focus for individual actors or groups.
• An openness to interaction. A willingness to interact with others is required. This assumes some commonality between actors that bonds them to each other.

What does this mean?

Analysis of the Iraqi insurgency indicates that emergent intelligence is evident. A complex series of local interactions has led to shifts in its behavior that reflect complex learning, goal attainment, and self-preservation despite a lack of a leadership hierarchy. What this means to the outcome of the war is as follows:

• The insurgency will continue to improve over time. Despite losses, the macro behavior of the Iraqi insurgency will become more complex (virulent) the longer it operates. As a result, it will be difficult to dislodge with each passing month. We will see less activity where it has little impact and more where it matters. Further, this activity will become more efficient.
• Breakout is possible. While it is unlikely that the insurgency will spread horizontally to other countries in an incremental fashion, it is very likely that those trained in this environment will seed other movements (and inevitable that the knowledge of this will initiate activity). Further, this breakout can occur globally and in unexpected locales -- since this neutral method isn't tied to any single motive, it can be applied to any cause.
• It is impossible to discern the motives of this movement (UPDATED). The motives of individual actors are easy to discern. A global motive is impossible to uncover, particularly since it is the culmination of thousands of local interactions. Even observations of the movement's global pattern of activity might be fruitless, since the time horizon is likely too short for accurate measurement. The movement is in a constant process of maturation in response to evolving environmental conditions.

Comments

[enigma_foundry]

A interesting discussion of the phenomena of an emergence is found in Cosma Shalizi's Doctoral Thesis. The interesting question he addresses is how to measure self organization.

The GG is a classic case of an emergent, self-organizing phenomena.

Is anyone busy on establishing metric of self organization that would be applicable to GG phenomena?

[James Bowery]

Evolutionary theory adequately addressed the emergent intelligence problem. It stems from a simple foundation:

Shared fate.

The most instructive examples of emergent stupidity are cheating genes, such as meotic drive, which cheat the meotic lottery.

The ecological example of such stupidty is the evolution of virulence in horizontally transmitted parasites. In vertical transmission parasites share the fate of their host's reproduction. In horizontal transmission parasites are free to emigrate to other hosts.

There are a _lot_ of geopolitical parallels here.

[John Robb]

EF and others: metrics are welcome...

James, I totally agree. I particularly like your example of virulence in horizontally transmitted parasites. On the other hand, if there is advantage in expansion, but at the cost of non-fatal virulence to the host, is it stupidity? Or what if the host is already at a low level of fitness, but the control of the dominant parasite is weakened or replaced by a new more aggressive one?

BTW, I really liked your Prisoner Dilemma simulation for communication:

http://www.geocities.com/jim_bowery/climatepd.html

However, it is a little too one dimensional since it assumes one greater good and not many goods. What if there is trust across a group but hypocrisy with external groups? How will that complicate the picture in a totally mobile/mixed world? Also, can a harsh environment be a failed state? Can a harsh environment be virtual?

[James Bowery]

In vertical transmission of the parasite by the host's reproduction, the ecological intelligence of the host/parasite system increases (or at least the stupidity decreases). But if the parasite can escape the shared fate of the host's reproduction (via horizontal transmission) the consequences are dire for the host's reproduction, to the point that it frequently results in what is called "parasitic castration" of the host, a stupid state for the host/parasite system (which no longer embodies shared fate) but not-so-stupid for the parasite. A question in my mind however is whether such a parasite can achieve an evolutionarily stable state within which the parasite can maintain horizontal transmission virulence without eventually wiping out, or being successfully defended against by, all but one of its hosts, with which it then must coadapt due to shared fate? I'm sure there must have been discussion in the literature but I've not been able to locate it.

I'd like to interject a bit of wild surmize at this point: It seems the "scale-free" networks (although I've seen critiques of that particular name, but that's another topic) that characterize systems with emergent intelligence represent a trade-off between robustness and intelligence via subsystem specialization. A recent breakthrough in general artificial intelligence is best described by Matthew Mahoney: "Hutter's AIXI, http://www.idsia.ch/~marcus/ai/paixi.htm makes another argument for the connection between compression and AI that is more general than the Turing test. He proves that the optimal behavior of an agent (an interactive system that receives a reward signal from an unknown environment) is to guess that the environement is most likely computed by the shortest possible program that is consistent with the behavior observed so far. In other words, the most likely outcome for any experiment is the one with the simplest explanation, where "simplest" means the smallest program that could model what you currently know about the universe." (http://www.geocities.com/jim_bowery/cprize.html)

Clearly, any maximally compressed system will not exhibit power-law distributions, as the entropy will be too high: the distribution will be flat or white noise.

In this regard "A New Route to the Evolution of Cooperation" by F. C. Santos, J. M. Pacheo is most fascinating. In prisoner's dilemma models with local connectivity they've substituted scale-free network connecivity and shown that cooperation can emerge victorious within such networks. This is particularly fascinating given the fact that scale-free networks are also known to be open to epidemic attack of any degree of virulence. The missing ingredient needed to consiliate these two opposing views of scale-free networks may be how the scale-free network comes to exist to begin with. In particular shared-fate may be the pre-requisite for the scale-free network comes to exist hence it is not the scale-free network that is the precondition for cooperation but the shared-fate of the system components.

Regarding my particular model of climatic variation: I think cooperation emerges within highly stressful environments because of the shared fate imposed by such environments. Positive sum gamesmanship is not just an optimization -- it is the difference between existence and nonexistence. This idea is pretty old though. In his "Discourses" Machiavelli recommends that the founder of any new city-state choose a site where nature imposes tolerable austerities, as it strengthens the character of the people.

[Cardenio]

http://news.moneycentral.msn.com/provider/providerarticle.asp?feed=AP&Date=20060210&ID=5497110

cyberswarmming !

[John Robb]

James:

Para #1: Right on. Herpes to an extent. Treated AIDs is a better example since it greatly reduces fitness while maintaining life (particularly given the evolutionary effect of technology).

Para #2,3,4: Right on again. Shared fate is a necessary prerequisite for scale free networks. They are remarkably resilient to "random" failure, economically efficient to build, and quickly grown: hence their popularity. However, they are very vulnerable to external pressure (once the shared fate breaks down).

Also, in regards to terrorist networks that exhibit scale free properties. It is possible that the hubs in a terrorist network could have connections that are much more robust than standard communications channels (given the difference in their nature). In the case of al Qaeda, the connections from the hubs (bin Laden, etc.) are via media broadcast (with feedback in the form of actions). These links are almost impossble to disrupt. The broadcast model makes it easy for these hubs to stay hidden until they are ready to make "statements."

Para #5: Failed states are great examples of stressed environments. It is very likely that failing states to create allies is a dominant strategy (as I have pointed out for a while now). Also, within a huge global environment there can be multiple positive sum relationship networks that compete in a suboptimal way due to core incompatibilities. IF this competition is not existentially threatening, it can continue indefinitely.

[Wordman]

If these five factors really are requirements, then the implication is that the elimination of any of them causes the collapse of the "intelligence" (whatever that means). This suggests a way to fight things like the insurgency in Iraq: destroy or prevent one or more of these factors. Taken in reverse order, this might be done as follows:

Openness to interaction is difficult to attack, because it is basically a personality trait, but it isn't impossible. One way is to encourage "closed" men into authority by, say, eliminating rivals providing intelligence or some other means. Another is to foster distrust between leaders. Any of these strategies, though, is fairly risky.

Pattern matching from stigmergic communication is a bit easier, because some of these environmental patterns can be manipulated or forged. Still, most of them can't. Further, while it may be possible to anticipate the results of a pattern match, doing any of this requires high level of understanding of the insurgency thought process, which I don't think the coalition has.

Random interactions are nearly impossible to stop. This is one area where the coalition is actively fighting (by incarcerating suspects to limit their movement). Ironically, this probably helps random interactions more than hinders them, as those who are inevitably falsely imprisoned are certain to interact with real insurgents in an environment the breeds hostility to the coalition.

Local focus is more attackable than most of the other requirements, because it allows local action in opposition as well. Still, a crackdown in one locality likely just moves the "local action" to a different locality, so the best you could hope for here is somehow manipulating the insurgency into a locality of your own choosing. Those who subscribe to the "better to fight them in Iraq than here" mentality might reasonably claim that this has already been done successfully.

Preventing the critical mass of participants seems to me to be the most sure way to defeat this insurgency. Unfortunately, present tactics for doing so (incarceration into public and secret prisons) seem completely wrong, as they motivate individuals to participate who may not have otherwise. Fighting this requirement would seem to take a much longer timescale, essentially replacing the bias of the Muslim world's educational system with one more slanted to the West. This is a timescale on which America is extremely bad at fighting, as its political system churns much faster.