This challenge to overconfident amateurs is relevant and well argued. In short, get out the pencil and paper, it's time for a renaissance of sketching. In Ch 3 he critiques [amateur design technique] which he views as too cerebral. In Ch 4 he evangelizes a more pragmatic, physical approach which he calls cognitive interaction. The paper is involved - he backs up his thesis by citing its philosophical, evolutionary basis among other tediousness - so you might want to skim Ch 0 and skip to Ch 3.
From Ch 5:
Why are experiments (and simulations) in the physical world superior to models and simulations in the head? The reason is that you want to find out both what you can figure out and what you can't figure out, i.e. what you cannot simulate mentally. That is, you want to know also about the effects of your actions that you cannot predict or forsee.
[Then he talks about iterating over "fine-grained pieces of activity, a continuous attention to feedback that replaces complex pre-planned actions, and simpler and smaller actions..."]
Why is this better? Remember dead reckoning: it starts from a known position, but the shortcomings of prediction yield an accumulating error that makes the computed position deviate more and more from the actual position. Position fixing instead reestablishes accuracy each time a fix is made. So the more often you make fixes, the more often you can make the proper adjustments to your course, and the less will your measured position drift away from the actual one.
Beyond the critique of the amateur, he's also tackling larger trends of thinking about design methodology. The trend is from formal to informal, from abstract to concrete. From ideas without context to ideas that receive immediate feedback from their environment...
Gedenryd quotes: "The writings of design theorists imply that the traditional method of design-by-drawing is too simple for the growing complexity of the man-made world. This belief is widely held and may not require any further justification. (Jones 1970, p. 27)"
But recently we've seen a resurgence of such "simple" methods like sketching, storyboarding, model-building, etc. Here's where we see the Renaissance of Sketching.
Basically it seems to me the trend is that a bunch of theoreticians in the mid-20th century thought that abstract thought and logic could (by themselves) produce perfect designs. But in fact recent research and writing about design are suggesting that the world's best designs come about because they were designed using processes like sketching that have a direct, real-world, tangible feedback. My favorite passage is maybe "Quist's demonstration of sketching" in ch 4, which directly addresses this dichotomy of formal/abstract vs informal/specific design processes.
I think many people have realized that overly formal methods do not produce designs that are all that beautiful or engaging. Most people hate cube-farms. Most programmers can't stand the waterfall model. Most consumers hate corporate bloat-ware. Maybe it's always better to use paper prototypes, sketches, extreme programming, and so on.
In some ways this whole debate was foreshadowed by Spock's musings in Star Trek I: it is an embrace of what is most human.
From Ch 5:
Why are experiments (and simulations) in the physical world superior to models and simulations in the head? The reason is that you want to find out both what you can figure out and what you can't figure out, i.e. what you cannot simulate mentally. That is, you want to know also about the effects of your actions that you cannot predict or forsee.
[Then he talks about iterating over "fine-grained pieces of activity, a continuous attention to feedback that replaces complex pre-planned actions, and simpler and smaller actions..."]
Why is this better? Remember dead reckoning: it starts from a known position, but the shortcomings of prediction yield an accumulating error that makes the computed position deviate more and more from the actual position. Position fixing instead reestablishes accuracy each time a fix is made. So the more often you make fixes, the more often you can make the proper adjustments to your course, and the less will your measured position drift away from the actual one.