Predictive Decision-Making With Incomplete Information

I was on a NYC Transit subway train this morning–on my way to a regular train station to make a specific regular train leaving at a particular time–and we were stopped at a station a bit longer than usual, and then the conductor got on the PA and said, “We are stopped at this station due to a medical emergency on the train. EMS personnel have been called to handle the emergency.”

So now all of the passengers are presented with a complex decision based on incomplete information. The decision–which is revisited from moment-to-moment–is whether to stay on the train and wait it out, or whether to get off the train and figure out another way to reach one’s destination. The prediction on the basis of incomplete information that must be made is, How long will it take before EMS arrives and takes the sick passenger off the train?

The decision calculus also involves the “sunk cost” of accumulated waiting time. It also involves weighing the time it will take to reach the destination by alternative means versus on the subway, once the sick passenger is off the train and it departs, and whether one has to make it to the destination with a “hard stop” to the time of arrival (such as I did, to make my regular train).

Some people got off the train immediately upon the first announcement that we were stopped for a medical emergency. I suspect those people had alternative means to reach their destination that were straightforward, rapid, and not costly. For example, anyone who was going to get off at the next stop–only eight blocks away–could just get right off and walk.

There were cohorts of people who got off the train right after each one of the several announcements that we were still waiting for EMS. I guess those people figured they would wait to hear an announcement, in case it was, “EMS is here. We should be moving shortly.”

After about ten minutes of waiting, more and more people started to get off the train. I guess these people reached a threshold and figured, “We could be here for a *long* time.”

My calclulus was as follows: (1) I am only four stops away from the regular train station, which takes five-to-seven minutes without delays. (2) My regular train doesn’t leave for forty minutes (I always leave early to make travel connections). (3) If I get off and walk, it’ll take probably 25-30 minutes, and I’ll be sweating like a pig when I get to the train station. (4) If I try to find a taxi, I might have to wait for one, and traffic could be horrendous, and there is nothing more stressful than sitting in a fucken taxi stuck in traffic trying to make a train or plane, and you might not make it. (5) If I miss my train, there is another one an hour later. (6) If I take the next train, I would miss the first meeting I have scheduled at my destination, but make it to my second meeting. (7) The second meeting is with someone I report to, and the first meeting is with someone who reports to me.

So I decided I was just going to wait it out.

After about fifteen minutes total waiting time, the conductor announced that EMS was on the scene, and then we left about three more minutes later. I made my train.


  1. Noah says

    This is what transportation engineers like myself have to deal with. It’s also difficult to predict the behavior of the passengers on the train in this situation, especially if you include the updated information they’re getting. Imagine predicting the travel time on a freeway, and telling people there’s an alternate route that would take less time. If enough people reroute, you’ll create congestion on the new route, throwing off your travel time predictions.

    When you predict the weather, at least the weather doesn’t know what you predicted.

  2. says

    I make exactly the same calculations (albeit in a different city). And, because I am very cheap and hate taking taxis, especially because I already have a monthly transit pass, I will almost always wait it out. I always have something to read with me, and I look at it as an opportunity to read another chapter.

Leave a Reply

Your email address will not be published. Required fields are marked *