Opposed Systems Design

Anyone know anything more?

The DSB study calls for the Pentagon to educate Congress about the problem and to create a new office to advise senior military leaders “of high risk potential red capabilities” and how to handle them. The new office, to be known as the Capability Assessment, Warning and Response Office, would warn senior leaders of high risks, come up with options to counter them, and recommend technological approaches, the study says.
The DSB also recommends that the Pentagon embrace red teaming throughout it structure… In addition to red teaming, the military must place much more emphasis on rapid fielding of capabilities and create a Rapid Capability Fielding Office that would report directly to the undersecretary of defense for acquisition, technology and logistics.

I’m wary of any approach that hopes, however implicitly, to institutionalize strategic thinking. Strategic thinking cannot be routinized, yet that is what one attempts if one tries to fix an undefined challenge like “unexpected threats” through bureaucracy.
No set of instructions, whether capabilities-based or otherwise, substitutes for having active strategic minds tackling a problem. It’s fuzzy and it resists an engineering approach, but that’s the nature of the challenge.
If this Capability Assessment, Warning and Response Office can serve as a center of gravity for strategists - a place where they can hang their hats and collect a paycheck - then it might be a valuable improvement. But you still need to find and recruit strategists to work in the office, which will require a leader with the autonomy to hire the types of people he or she thinks will help.
These people should be voracious consumers of information, excel at rigorous abstract thinking, and be conversant with the uses and limits of quantitative techniques. There should be a wide range of backgrounds, including history, operational experience from the military, business, policy analysis and mathematics. They must have some autonomy in seeking out interesting problems and developing them (along the lines of Google’s one day a week policy of independent projects). The office requires lifelong learners who look deeply at situations and wrestle with undefined problems.
During RAND’s golden age (roughly 1950-1961), new employees generally had at least a year to rattle around the company looking into things and figuring out the lay of the land. It’s a shocking proposition to consider in today’s world of hourly accounting systems and contract auditing. But it reflected a deep faith that RAND leadership had in its hiring managers. If they brought bright, highly motivated and creative thinkers together, they trusted that these people would do something worthwhile often enough to justify the times when they didn’t.
It all flies in the face of scientific management. We can explore new ways to managing this sort of organization - perhaps using tools like prediction markets to assess whether analysts are producing accurate analysis. But one cannot get around the need for finding the right type of thinkers. Otherwise, it won’t matter what their office is called or who it reports to.

This theme is going to keep coming up. The Stimson Center ran a study last summer looking at it. I’m sure lots of other think tanks are getting into the issue. Please share any links to studies you’re familiar with.

Unprecedented collision of two satellites in orbit. A topic to watch.

I can’t wait to get to the book, but my to-read pile has been growing at an alarming rate that absolutely overwhelms my rate of consumption (a fact that prompted a welcome gift that I would recommend to anyone else who suffers from this same challenge - which is to say the entire readership of this blog).

It’s a classy move. Easy to do, but easy to forget too. I’m flattered to be on the list.

In general, the underlying idea is: each system has its own critical point. If I know where it is, I hit this point and destroy the whole system. If I do not know this, I will have to go on hitting different components of the system until I accidentally hit the critical point. The more components I damage, even without hitting the critical point, the closer is the moment when the system disintegrates.
And there is a certain connection between “q” - which is the percentage of component interconnection - and “Q” which describes the probability of the whole system collapse.

When the rate of component failure q is 11%, the probability Q(q) for total system collapse is 50% When q=25%, Q=81% When q=50%, Q=100%

Fig. 4. The Probability of System Collapse

So what does this formula tell us? In case the damage level of the components (q) within the system is 50%, the system will definitely stop functioning. There are simply no systems capable to withstand the malfunctioning of half of its components. In case q is 25%, there is still an 80% probability of its falling apart.
I’ll buy that one could make this sort of assessment of an electrical network or a water distribution system or an oil pipeline network. But a dynamic network of organizational authorities, human relationships and commands? I think we’d have to generalize away all the interesting content to make such a comparison.
Also, there are obviously assumptions about the underlying the network topology behind the above calculations. I’d be fascinated to learn what they are. Maybe when I read the whole thing I’ll find out.

Google and Nasa are throwing their weight behind a new school for futurists in Silicon Valley to prepare scientists for an era when machines become cleverer than people.

Of course, responsible programmers have already been addressing this issue for a while:





Remember: only you can prevent Skynet from killing us all. Program responsibly, kids.

Same goes for the Sim Iraq idea.
Put simply, if we had adequate mathematical models for human and societal behavior, then we could apply them at home to figure out who will win an election before it happens. Actually, that would be an easier challenge because it wouldn’t confront any language or culture barriers.
I’ve got a quant background, so I deeply identify with the desire to rationalize decision making through the use of models (so that we can do substantive what-if analysis). But these problems are so poorly defined and the content so staggeringly broad that I’m skeptical that we have the adequate knowledge to start building a useful tool.