Emergence Notes

These are notes for discussion at the 9/28 Emergence meeting on the role of emergence and complexity in the management of emergencies and disasters.


Duncan Watts:

“Only decentralized but coordinated groups can respond to unexpected catastrophe.”

From Six Degrees:

“How does individual behavior aggregate to collective behavior?” as a pointer to emergent phenomena in human societies (or in interactions of human societies with environments).

Agent-based conception of disaster events: modelling human actors as rule-driven, simultaneous interactions within networks, as both affecting environments and expressing rules differently as environments change

Key concepts: phase transitions, critical point, criticality in complex systems or networks, percolating clusters

“the center emerges only as a consequence of the event itself”; “events are driven not by any preexisting center but by the interactions of equals”

In light of Watts’ remarks, current narratives being offered to explain or understand what happened with Katrina: (see David Alexander, “Symbolic and Practical Interpretations of the Hurricane Katrina Disaster in New Orleans” and Alex de Waal, “An Imperfect Storm”, for more thoughts on this aspect of the problem.

It was Michael Brown
It was FEMA or FEMA being placed into Homeland Security
It was the state of Louisiana
It was the mayor or police force or officials of New Orleans
It was the strength of Katrina
It was global warming causing the strength of Katrina
It was the location and history of New Orleans
It was the long-term underfunding of protections for New Orleans
It was the specific immediate underfunding of protections for New Orleans
It was the specific technical stupidity or design of protections for New Orleans
It was unanticipated technical failures in the levee system
It was unanticipated aspects of the storm surges
It was the lack of available resources due to the war in Iraq
It was the specific flood-control policies of the Army Corps of Engineers combined with the specific erosion of wetlands near New Orleans
It was the structural poverty of New Orleans
It was the racially significant structural poverty of New Orleans
It was that George Bush doesn’t care about black people
It was the lack of automobile ownership
It was the looting
It was the bad decision-making processes of people who stayed
It was bad emergency planning
It was the specific bad planning of the evacuation
It was the failure to request bus transports
It was bad weather forecasting (this seems limited to Rick Santorum)
It was the media
It was the exaggeration of specific facts and the transmission of rumors
It was an excessive focus on New Orleans at the expense of talking about coastal development on the Gulf Coast
It was energy development and effects on a tight oil market plus general effects on economy
It was knowing misrepresentation for political ends
It was the unconscious imprinting of one event with feelings or reactions stemming from some more general social problem or consciousness
It was the big-government urges of the Bush Administration and the failure of small-government conservatism to make its mark

What is “it”?

The complete complex situation which existed about three days after Katrina: the “national narrative” of disaster that was intensely meaningful to so many people.

It isn’t just a question of the numbers of deaths or the insurer’s toll or even the destruction of a city, but the total picture: a story about individual human beings, government, environment, technology.

Yes, Paul Grobstein: it’s story-telling time.

The total picture can’t be understood without something like complexity, emergence, networks: a technical language that allows a great many dimensions of the problem to be in motion at once and interacting in parallel without insisting on subordinating most to the singular real explanation or the isolated variable.

Substitute in Watts’ phrase: the center for “the explanation”. “The explanation only emerges as a consequence of the event itself”.

Note the complications of evacuation in Rita: the “Katrina effect” changed the agent ruleset


Focus in the wake of Katrina on the predictability of the event, BUT:

If we take the analysis of Watts and other network scientists seriously, predictability is exactly the wrong thing to think about, that the particular shape of a disaster will pose its challenges largely through the particular complex architecture it assumes at its critical point. Knowing that the levees may fail in the case of a certain intensity of storm does little to predict the complete complex post-disaster social order in a given locality.

Albert and Barbasi: scale-free networks [most nodes of the network poorly connected, some highly connected] are most vulnerable at their most highly connected nodes; robustly resistant to random failures but not to those which strike at highly connected nodes.

Watts again:

“Percolation models…typically assume that all nodes have the same likelihood of being susceptible. In reality, however, heterogeneity is an important feature of human and many nonhuman populations…cascades of contingent, interdependent failures are more complicated to model…”

Cascades in a context like New Orleans: depend on decision rules that individual agents are using, but also connectivity between individual agents, on the structure of social networks

So for example, it’s possible that the decision to stay in New Orleans was a cascade of neighbors/family assessing multiple factors: assessment of likelihood of looting, capacity to replace lost goods, past experience of hurricanes (Betsy or more recent ‘misses’), knowledge of NO police force’s likely behavior (or failure) in crisis, on the other hand trust in governmental capacity to intervene, physical inability to evacuate, purely individual decision rules (which Watts argues are at the macrolevel effectively random).

Superdome situation: perception of lack of planning + lack of planning + contingent incidents of violence or threatened violence + later decisions by other government agents (suburban cops ‘penning’ NOLA residents in city) + media representations recursively being played back into city = sense of total disaster.


Some policy observations, both general and in specific related domains; “applied emergence”.

Charles Perrow: “normal accidents”—the worst vulnerabilities are not where we think they are, and not a ‘natural’ consequence of particular events, but of systems which fail poorly in cascades. Perrow on Katrina. (In my opinion, actually a somewhat ‘conventional’ story of failure in Perrow’s rendering.)

Doyle and Carlson: “highly optimized tolerance”: real-world networks are commonly both robust and fragile. Possible recommendation from Watts: “a richer conception of robustness”—a consciousness that at highly connected nodes, you have to think differently about failure.

Bruce Schneier on security policy in Beyond Fear:
designing to fail well; “brittleness” is a vulnerability

“automated systems…are brittle…homogenous systems and systems with an overdependence on secrecy tend to be brittle”

consciousness of ‘trade-offs’ in design of security

“security revolves around people”: it’s not systems or technology

Many small little decentralized security systems that rest on the independent judgement of people (but firewall the consequences of bad judgement at any given “node”) are better than big, centralized ones

“detection works where prevention fails”: knowing the state of a system or an asset is better than trying to rigidly protect it

Thomas X. Hammes on counter-insurgent “4GW” in The Sling and the Stone

“superior use of all available networks to directly defeat the will of enemy leadership”

4GW antagonists will identify highly connected nodes and attack them, minimal effort for maximum result; “all an opponent has to move is ideas”; “technology is largely irrelevant”; “the most damaging aspect of our manpower system is its purely top-down evaluation system…we have combined the worst of bureaucratic manpower systems with top-down reporting and rapid personnel turnover”.


What would “failing well” in disaster planning: let’s say, planning for hurricanes, look like?

What are the “highly connected nodes”?
For example, are rural coastal communities nodes which can fail without consequence?

Or could we simply work to make all “nodes” in vulnerable communities weakly connected (e.g., by preventing coastal development of certain kinds in hurricane-vulnerable areas?)

What would need to “fail well”?

For example: is looting an “emergent” method for relief distribution?

Would neighborhood-based systems be better for rescue and verification of resource needs than centralized ones?

Is there such a thing as a decentralized, “improved robust”, system for disaster mitigation and assistance? Is that a governmental capacity, or does it lie somewhere else? Does it require a new structure or unit to government or a reform of existing structures and units?