"even were a design school to decide to teach more formal methods, we don't really have a curriculum that is appropriate for designers. Take my concern about the lack of experimental rigor. Suppose you were to agree with me - what courses would we teach? We don't really know. The experimental methods of the social and behavioral sciences are not well suited for the issues faced by designers.
Designers are practitioners, which means they are not trying to extend the knowledge base of science but instead, to apply the knowledge. The designer's goal is to have large, important impact. Scientists are interested in truth, often in the distinction between the predictions of two differing theories. The differences they look for are quite small: often statistically significant but in terms of applied impact, quite unimportant. Experiments that carefully control for numerous possible biases and that use large numbers of experimental observers are inappropriate for designers.
The designer needs results immediately, in hours or at possibly a few days. Quite often tests of 5 to 10 people are quite sufficient. Yes, attention must be paid to the possible biases (such as experimenter biases and the impact of order of presentation of tests), but if one is looking for large effect, it should be possible to do tests that are simpler and faster than are used by the scientific community will suffice. Designs don't have to be optimal or perfect: results that are not quite optimum or les than perfect are often completely satisfactory for everyday usage. No everyday product is perfect, nor need they be. We need experimental techniques that recognize these pragmatic, applied goals.
Design needs to develop its own experimental methods. They should be simple and quick, looking for large phenomena and conditions that are 'good enough.' But they must still be sensitive to statistical variability and experimental biases. These methods do not exist: we need some sympathetic statisticians to work with designers to develop these new, appropriate methods."
(Don Norman, 26 Nov 2010, Core77)