Software features can make things better, or worse. In one case, a feature to turn off a smoke alarm with a wave sounded like a good idea. The problem discovered by testing, after the release, was that the algorithm could misinterpret movement near the detector as intentional when it wasn’t. See the write up in IEEE Spectrum for more details.
The question to ponder for quality professionals is how do we test features like this, especially in safety critical applications. Writing tests to see if a feature performs as designed seems pretty straightforward. This seems to be a case where the “internet of things” is exhibiting behavior similar to security holes in webforms. In security testing, the tester can enter unintended strings to see if the software will let them pass or not. The “internet of things’ takes this to a whole new level. How to detect intentional motions, from unintentional? How to make something “easy” to use without creating a safety issue?
These are good topics for discussion. Submit your ideas in the form of an abstract for this year’s PNSQC conference coming in October.
