Information Technology Reference
In-Depth Information
steps) and establish themselves in niches in Value Space far from the centroid. In
contrast, the dispersion of ideal products in the 'Identical Replacement' treatment
appears to be ephemeral movement of individuals without persistent clustering. The
clustering behavior is not as visible or apparent in the space of Utility Weights (top
row), but other analysis or visualization methods might reveal some effects.
11.6.3 Setting 3: Producers and Consumers Interacting
is the only one of the three rules that show significant mutual influence. For Rule
1—Minimize cost, adding Consumers interaction made no difference since it didn't
alter the cost function. For Rule 2—Maximize performance, adding Consumer
interaction also had no influence because Consumer demand did not directly
influence the Producer's performance metric.
The situation is different for Rule 3 because variations in Consumer demand can
change the performance/cost ratio as measured by Producers. Comparing Figs.
11.8
and
11.9
, it can be seen that movement along the design trajectory is slower when
Consumer values are dynamic and situated (Fig.
11.9
), compared to the setting
particular parameter settings for Producer and Consumer learning, the design
trajectory is not altered by Consumer interaction. However, that does not rule out
the possibility that there are parameters values that can result in a dramatic shift in
trajectories. Using full parameter sweeps, we will look for critical parameter values
in future research.
Step = 2335
Fig. 11.8 A snapshot for Setting 1—Producer only, showing the active set of designs (black dots)
using Rule 3—Maximize performance/cost ratio, recorded when the target metric reached 25
Search WWH ::
Custom Search