fear of revealing data and positions at a time of looming legal conflict involving 

 the Truckee-Carson water problem. Finally, middle management in govern- 

 ment and industry often represent a bulge of incompetence that frustrates 

 change within an organization, however much desired above and below. In the 

 words of the Vice-President of a major international mining corporation, 

 "there is a good reason why many middle managers never become senior ones." 

 Above all. implementation requires patience. It requires time for ideas to 

 gestate, for inter-personal and inter-institutional adjustments to occur. It 

 requires time for key unlocking events to occur — a crisis, an election, a public 

 hearing. Some can be planned, most occur as surprises. At one point the 

 budworm study seemed, at best, to have only changed data collection 

 programs, albeit significantly. All efforts to institute policy change seemed to 

 be frustrated at the eleventh hour. In despair. Baskerville wrote an explicit 

 critique of Federal. Provincial and our own activities for the II.ASA policy 

 seminar. '" There were three responses: first, it is not true; second, it is true but 

 we cannot do things differently; third, Baskerville, how would you like a job as 

 Assistant Deputy Minister. 



WHERE ARE WE STUMBLING 



Life is ever delightfully uncertain and ambiguous: the act of bridging gaps has led 

 to new gaps and to new problems. At the moment we can hardly define whether they 

 are important or transient, so they are presented here as potential problems only. 

 Perhaps they will disappear. 



Some Models Have Predicted Too Well 



All work on GIRLS stopped in 1970 and, moreover, the model was initialized with 

 data from 1900. Yet the model has tracked, surprisingly well, changes in selling 

 prices, rates of development and rates of environmental decline since 1970. Similarly, 

 the budworm model, in a more qualitative way, accurately predicted radically 

 different behaviors indifferent regions of North America. That is surprising because 

 we always argued that simulation models were lies, whose quantitative predictions 

 could not be trusted and whose usefulness was in giving insight and mediating 

 constructive dialogue. We could argue that the reason for this high predictive power 

 came because we insisted on a process structure that relied on well-tested and 

 carefully generalized presentations of those processes. But we are simply not sure. 

 The reason why this is a problem is precisely that. One cannot, a priori, identify the 

 limits of predictive power or robustness, no matter how much effort goes into 

 invalidation. It was much easier when we could automatically disbelieve the results! 



Being Adaptive is Essential, But — 



There is certainly no doubt that one cannot predict everything, anticipate all 

 surprises. That is why we argue for an adaptive emphasis that allows probing, 

 experimentation, learning and change. But we encounter two problems. The first is 

 that we are living in an unforgiving world that penalizes error, gambling, and hence 

 learning. The very word adaptive has been attacked by elements of the USFWS and 

 the Bureau of Land Management. Some who feel beleaguered in their defense of the 

 environment believe in an all-or-none world, and that an adaptive sequence will lead 

 lo afait accompli for the developer. Give the developer a pilot study and he will take a 

 project! 



The second problem with an excessive emphasis on an adaptive approach is that 

 for certain developments the actual costs of experiment error can truly be too large 

 for society to bear; chemicals that trigger cancer decades later, or nuclear power 

 plants. We can keep trying to develop new designs that are more forgiving of error, 

 but we are stuck with many that demand some kind of predictive screening devices. 



89 



