Fig. 3. Double-loop learning (Brockbank and McGill 1998).