Complexity is certainly a problem. In Refineries and Chemical Plants automatic control is simplified by separating functions:

1. Process Control System DCS (Analogous to the Model in the article)
2. Emergency Shutdown System ESD (Analogous to a Risk of Ruin feature in the article)
3. Alarm Systems (These are early warning systems not tied into the above DCS or ESD. This is looking for things that are not necessarily included in the above controls. Analogous to an unexpected bad report not usually considered in the model)

The DCS or Distributed Control System essentially contains a Model of the Process being controlled and is normally automatic once it is tuned properly. This would also be analogous to a trading system generating buy and sell signals based on a model.

The ESD or shutdown system automatically shuts down the plant (operating unit). This would be like a stop loss provision that automatically closes a position in a trading system.

If an Alarm System is giving an operator a warning he can take actions before the emergency shutdown system. In a trading system this might be a condition that causes a trader to reduce a position. Or in the present situation it was widely known that NINJA (bad) mortgages were being issued. Where were the alarm sirens?

The following is where the analogy starts to breakdown. Refineries and Chemical Plants are dangerous places where there is the potential for immediate loss of life. And while many processes are often not understood to the detail we would like, how to measure and control them is fairly well understood.

The situation with in finance is that primary data is often revised and the operating rules can change with little warning as we see in the current crisis. One of the points of the article is to be on the lookout for incorrect assumptions in the mathematical model or the verbal model of what we think is going on. Sometimes our math models and mental models can blind us to what is actually going on. Or as Will Rogers put it: “It’s what we think we know that just ain’t so that really hurts us.”

the essay did engender some good comments, tho.

1) reality don’t care about all yer fancy degrees.
2 models don’t always work.

Then he takes a remark he remembers from an economist who was probably just knowingly making an overgeneral statement because he was talking to a class of undergraduate non-economics majors, and said ‘Us engineers knew better than that guy’.

Frankly I don’t know why Mr. Retholtz published this, it’s so uninformative.

It does remind me of something though. Go to the wacko Intelligent Design websites like Uncommon Descent and you’ll find that an unusually high number of engineers there, sneering at how dumb and naive professional biologists are about biology, claiming they’re better than those evolutionist eggheads because engineers are out ‘in the real world’ and ‘lives are on the line’ etc. Or read Steven Den Beste’s old essays about why the Iraq War was going to be the best thing ever. He made a big deal about how his engineering training made him a better analyst than the so-called-professionals too.

A typical simplifying ruse in model making is to linearize a representation of a phenomenon (e.g. efficiency of a distribution system vs its cost, number of nodes, modes of transfer ) to simplify the math and its cost of development. At best they may prevent a decision maker from a disastrous mistake — but recall the prescient but incredibly vapid pronouncements by the unlamented and departed Donald Rumsfeld about categories of ‘knowns’ and ‘unknowns’ that we may or may not know.

As regards inquiries about possible outcomes about hypothetical events, a very useful application of models, it’s worth reading the book, ‘The Black Swan’, by Nassim Taleb. Had I thought about its lessons last Spring, and be prescient enough to apply them, I’d be a lot richer today.

aurora17

There is a whole technology relating to those issues including things like Safety Integrity Levels (SIL), failure rates (MTBF), Risk Reduction Factors (RRF) etc. In addition there are often smaller alarm systems that alert operators, technicians and engineers of problems but do not shutdown the plant or operating unit. There is also a great effort to make systems that are “idiot proof”.

But catastrophic failures still do occur. And when they do there are usually serious investigations especially if there are fatalities.

Reference: “Control Systems Safety Evaluation and Reliability” 2nd Ed. by William Goble, isa.org, 1998

This concept would also seem to imply that to reduce systemic risk, we should seek to reduce the complexity of the entire system.

A post, and one person agreeing, is hardly the sort of reflexive echo chamber you see ont he extreme right and left.

But I understand the point you were trying to make . . .

You end up with something like the Pontiac Aztec.