ATM Alarms and Signal Detection Theory

My favorite folks over at The Consumerist posted a nice writeup on why ATMs do not have alarm buttons.

According to a FTC report, at about $1,500 a pop, they’re expensive to retrofit, the police usually don’t have the resources to respond in real-time, and they result in more false alarms than real calls. In one pilot program, an alarm-equipped ATM resulted in 500 false calls to the cops, and zero real robbery notifications.

For some human factors fun, try to calculate d-prime. 🙂

4 thoughts on “ATM Alarms and Signal Detection Theory”

  1. Well, according to the grid, there were 500 uses of the alarm button in total, so 100% of uses were False alarms and 0% were hits. The non-uses of the button would be misses and correct rejections, no?

    Fudging the numbers a bit so we can get z-scores:
    z(.00000001) – z(.99999999) gives me about -11.2 for d’.
    c is 9.1458E-10 by this estimate, so a slight bias to say that there is no alarm 😉

    1. Of course that’s pretending there are no people who used the thing without pressing the button….

Comments are closed.