There are at least two ways to correct errors. One is to change the behaviour (for example, reduce backbiting and badmouthing among individuals). This kind of correction requires only single-loop learning. The second way to correct errors is to change the underlying programme, or master programme, that leads individuals to bad-mouth others even when they say they do not intend to do so. This is double-loop learning[1]. If actions are changed without changing the master programmes individuals use to produce the actions, then the correction will either fail or will not persevere.