The Fire

Friday, May 28, 2010

Editor's Note: I'll be enjoying the company of my family in a remote location next week. I may have limited or no Internet access, so posting, comment moderation, and responses to email may be slow to nonexistent.

It may seem strange to derive bloggy inspiration from an apartment fire, but that's what's on my mind this morning and I'm due to catch a flight later in the day, so that's what I'll write about now.

Last Sunday, my wife and I started our morning with the following emotional mini-roller coaster: the fire alarm for our building goes off, capping a string of annoying false alarms over the previous week; we quickly discover that said alarm is not a false one; and, within an hour, we experienced the great relief that our unit was unaffected by the blaze, to the extent that we more or less had a pretty normal day afterwards.

By coincidence, I had been doing lots of thinking about the subject of induction for a course that weekend, and the problem posed by the false alarms happens to be related to several aspects of a rather complex induction (completely unrelated to the course material) that I have been pondering off and on for quite some time.

False positives -- for which early car theft alarms were notorious -- and false negatives are two different types of statistical error that have to be minimized for things like automated detection systems. In the case of a false negative, if your building catches fire while you're asleep and the alarm fails to go off, you could easily die.

On the other hand, false positives are not categorically better than having no alarm at all. If the alarm goes off all the time for no good reason, you end up wasting lots of time checking for fires or becoming complaisant about the alarm. If the alarm happens to coincide with a fire, the latter could be deadly. If not, the former can seriously undermine your quality of life, especially if, for example, you are on a tight schedule and need your time for better things than tracking down nonexistent fires. Ditto for the (head-splitting) alarm itself if it frequently goes off during the wee hours.

No matter how well-designed an alarm system is, the fact remains that it will produce both kinds of errors. The fact is, its output (or lack thereof) really represents only a single piece of evidence from reality that must be rationally evaluated by those it is meant to aid. Our apartment, fortunately, has a window facing the only part of our building readily accessible to fire engines. That fact made evaluating last Sunday's alarm as a likely true positive very easy for me: I heard much more machinery noise than usual under our window after the alarm went off, and upon looking outside, saw something I had never seen before: the ladder passing directly over our window. I didn't even have to leave my apartment to discover that we should ... leave our apartment. Had we been in the other side of the building, noise, the smell of smoke, heat, or any number of other things might have provided the same information or obviated the need for an alarm altogether.

I see parallels between such automated cognitive aids as alarms and animal instincts: Both function automatically, but since there is no such thing as an automatic means of knowledge past the perceptual level, neither has the advantages of a volitional, conceptual consciousness, which allows for error-checking in the form of the consideration of all evidence, which includes questioning whether one's premises are in error.

And in the sense that premises aid cognition, although the analogy is far from perfect (It fails, for example, with certain very basic types of premises, and it also does not adequately cover refinements of essentially correct principles made possible by new knowledge.), I see a parallel between one's premises and a sort of built-in, sophisticated alarm system that one can fine-tune or even completely rebuild from scratch. The key elements in deciding whether to do this are (1) to be on the alert for times that one's premises might be wrong, and (2) to carefully prioritize and weigh all the evidence one has at his disposal.

This is an interesting line of thought, but, alas, more mundane things need my attention, so that preliminary bit of mental chewing will have to be it for now.

Have a great Memorial Day weekend!

-- CAV


Darren said...

Unrelated, but still something to say: I'm glad you added the Twitter support. I've slowly moved away from RSS and more towards watching Twitter and what people I "follow" are interested in. I tend to catch your tweets easier than picking them out of my huge Objectivist RSS reader feed.

Gus Van Horn said...


Thanks for the feedback.

I'm glad this is proving useful.