To be fair, getting some "false positives" is quite common even with normal, dumb smoke detectors. The real question here is how frequent such behavior occurs.

To be fair, getting some "false positives" is quite common even with normal, dumb smoke detectors. The real question here is how frequent such behavior occurs. Personally, I'd happily tolerate some false positives if in exchange I can be sure that the alarm will always go off in case of real danger. Also, when working with an imperfect measurement device the engineer designing the system will always be forced to choose an operating point that fulfills the desired characteristics of the device best. In case of a smoke detector the most desirable characteristic is probably that it ALWAYS goes off when there is some smoke in the room, so choosing a working point that might incur some false alarms but that will never (or only under very unlikely circumstances) create a false negative is still the best choice. Of course there are other strategies for dealing with failures in a given component (e.g. by using several, distinct and uncorrelated sensors), but those tend to increase complexity and cost as well (after all, no one wants to spend a fortune on a fire detector).