power and allowed stabilizing actions that he had previously prohibited .
Most computer modeling in this period represented only force movements and combat — not the longer-run issues that were more quintessentially political , such as the parochial shortcomings of Nouri al-Maliki that led him to squander opportunities . 5
U . S . troops departed over the period 2007-2011 , as promised in a formal agreement reached under President Bush , which President Obama honored , despite calls for leaving forces in Iraq . Subsequently , Iraq fell again into sectarian struggle and the Islamic State emerged to fill the vacuum . We need not elaborate here on how the saga unfolded ( and continues to unfold 16 years later ). This case illustrates the profound importance of questioning the conceptual model behind reasoning , not just the values of parameters in a formalized model .
3 . Myths and Realities about Uncertainty Analysis
Section 2 illustrated why analysis
should consider model uncertainties , as has long been recognized . It has been widely believed , however , that confronting model uncertainty is not feasible and that it is more pragmatic to make assumptions and plunge on , rather than being paralyzed by uncertainty . That stance once had considerable basis in practical experience , but it is now a myth : uncertainty analysis —
Journal on Policy and Complex Systems
5 An exception was work with the political simulation model , Senturion ( Abdollahian , Barnick , Efird , & Kugler , 2006 ).
186
even about model uncertainties — can be very helpful .
Pioneering work on uncertainty analysis in the 1980s was pulled together in a textbook ( Morgan & Henrion , 1992 ). Subsequently , two streams of research evolved at the RAND Corporation : one focused primarily on defense planning ( Davis , 1994 ; Davis , 2014 ), the other primarily on broader social-policy issues ( Bankes , 1993 ). These efforts have led to the methodology now called robust decision-making ( RDM ) ( Lempert , Popper , & Bankes , 2003 ; Lempert , Groves , Popper , & Bankes , 2006 ; Popper , Lempert , & Bankes , 2005 ). Interest in decision-making under deep uncertainty now has strong international interest ( Marchau , Walker , Bloemen , & Popper , 2019 ), as evidenced by important contributions from the Technical University of Delft ( Haasnoot , Kwakkel , Walker , & Maat , 2013 ) and a vibrant professional society ( see
http :// www . deepuncertainty . org ).
Uncertainties can be addressed with the more sophisticated approaches and policymakers can — without paralysis — be assisted in making well hedged decisions . The approaches , however , are more advanced for parametric uncertainties than for model uncertainties . What follows suggests a way ahead , described for uncertainty analysis generally so that our discussion of model uncertainty fits into that larger context .