To summarize Helene’s introductory post, this project aims to develop a time-sensitive approach to strategic anticipation and warning, using the evolving relationship between Iran and Saudi Arabia as its substantive framework. But how does one undertake strategic anticipation within the context of the complex, chaotic, fast changing politics of the contemporary Middle East? Just over the past few months, we have witnessed an extraordinary series of surprises in or strongly affecting the region, such as the plummeting price of oil, the Saudi intervention in Yemen, the deployment of Russian forces in Syria, and, most recently, the emergence of fears of widespread sectarian warfare. Referring to the last development, but offering a judgment of more general applicability, the Middle East politics analyst Rami Khouri has said that “this is unprecedented, and we don’t have a road map.” (“As Conflicts Flare Up, Leaders Fan Sectarian Flames in Middle East,” New York Times, October 17, 2015).
The proposal here is, in fact, to build a map—a conceptual rather than a cartographic one—of political/security dynamics of the Middle East with a focus on the Iran-Saudi relationship. The aim will be to highlight and to make sense of the key influencing variables and their inter-relationships, and then to flag changes in these factors through continued horizon scanning and evaluative reassessment. The evolving map should be useful even in the near term by promoting analytic “mindfulness”—continuing focus on the broad range of driving forces and possible outcomes—that can help to avert analytic misjudgments resulting from inattention and rigid mindsets. (W. Fishbein and G. Treverton, “Making Sense of Transnational Threats“, Sherman Kent Center for Intelligence Analysis, 2004). But this effort can also lay a firmer foundation for strategic anticipation by progressively building understanding of both the influencing relationships and the observed trends that will shape future developments and their timing.
The map will, in effect, be a model of the Iran-Saudi relationship, but one intended to promote reflection and discussion rather than to calculate firm judgments based on hard data. To reiterate Helene’s comments, our goal is to produce broad foresight on plausible directions and outcomes rather than to engage in prediction. To this end, we will keep the end-state variables fairly general and wide-ranging—Iran-Saudi conflict or cooperation—leaving room for refining them and for considering intermediate or alternative states as we proceed. In a situation as fluid as the one we are addressing, it is better, as Keynes is said to have remarked, to aim to be “roughly right than precisely wrong” (see footnote * below for the proper attribution).
Let me briefly describe three major of components of the project: the map itself, follow-on scanning, and a “Bayesian-influenced” (to be explained below) rolling assessment of implications for foresight. As this is a very much a work-in-progress for a rapidly evolving situation, modifications of “the plan,” such as it is, are likely. We welcome reader comments and suggestions on any aspect of the project.
The map will be constructed in stages building upon the skeletal framework shown below. At its center will be a node containing the project’s focal question(s): Is the Iran-Saudi relationship heading for conflict or cooperation? (and, sooner or later, adding the temporal dimension: “within what time frame”?). The other nodes depicted here broadly capture the key political, economic, and security-related variables in several domains—domestic, regional, and global—that will influence Iranian and Saudi government decisions. These nodes will be progressively converted into clusters that will more finely capture the key actors, driving forces, and constraints in each of these domains. The nodes will be connected by arcs (included here only for basic Iran and Saudi influencing factors) indicating direction of influence and, where the evidence warrants, assessed strength. In the increasingly interconnected global security environment, many arcs will depict boundary-crossing linkages that often carry great potential for generating surprise.
As an illustration of how the map will develop, I have provisionally added several nodes based on thoughtful observations offered by a RTAS reader in response to Helene’s initial post. For instance, this reader notes that falling oil prices are depleting Saudi Arabia’s foreign reserves, probably forcing Riyadh to borrow from international bodies within five year’s time. And he also notes that the difficult and costly intervention in Yemen will increase the strain on Riyadh’s resources. And so we need to include variables for Saudi foreign reserves, Yemen war costs, and, following on logically, Saudi public spending. The connections among these nodes give rise to obvious questions to be addressed in the commentary to accompany the map: Could these combined financial pressures lead to an even earlier financial crisis for Saudi Arabia? Or will they prompt changes in the very Saudi oil production policies that are contributing to low oil prices or, perhaps cutbacks in Saudi internal or external largesse? (And what are the potential knock-on effects for Iran’s economy and regional strategies)?
In building the map, it will be important to seek a balance between two somewhat contradictory requirements for usefulness. One is that the map be robust enough to encompass the range of influencing factors so as not to miss less prominent ones from which surprise might emanate. The other is not to clutter it with so many nodes and arcs that it becomes difficult for even the informed observer to grasp and understand it. We shall experiment with different ways of presenting the map so as to allow us to see both the big picture and fine-grained detail.
The map will guide a continuing scan of press, academic, and think-tank publications searching for signals of potential change. This will include both reporting on events and broader analyses of trends and influencing factors. Selected results of the scan will be used to support updating of assessments of emerging and potential outcomes as well as refining the map itself. RTAS scans will be central to this effort but many other sources will also be consulted. Links to the information collected will be archived on this site in connection with the relevant nodes—searchable by clicking on them—so that a trail of facts and analysis can be easily summoned up by those seeking to dig deeper into particular issues.
As an illustration of how this process will work, let us return to the case of Saudi financial pressures. Our scan would undoubtedly highlight a recent article that, while not rejecting the possibility that Saudi foreign reserves may be stretched in a few years, points to the country’s under-appreciated economic strengths, including a solid banking system, progress on industrial diversification, and the recent introduction of a streamlined economic decision making system. (“The Reports of Saudi Arabia’s Death Have Been Greatly Exaggerated,” Foreign Policy, October 20, 2015). This analysis might affect judgments on the near term effects of lower oil prices and Yemen, and prompt an increased focus on the potential longer term effects of diversification.
As foresight is the project’s objective, our assessments based on the map and the scans will be largely qualitative in nature: often just raising questions about possible implications or offering rough judgments about the plausibility or likelihood of possible outcomes. Nevertheless, in keeping with our ultimate aim of including the temporal dimension—which will involve increased specificity as we address onset, sequence and duration—and to reduce the risk of well known cognitive errors associated with a purely intuitive approach to judgment (e.g., confirmation bias—looking for what we expect to see), it would be helpful to add some form of structured analytic thinking to the assessment process. In this connection, we should consider employing elements of Bayesian inference, whose value has been recently highlighted by noted specialists on the art and science of prediction, such as Nate Silver (The Signal and the Noise, 2012) and Philip Tetlock and Dan Gardner (Superforecasting: The Art and Science of Prediction, 2015).
Tetlock and Gardner describe Bayesian inference as a process of “gradually getting closer to the truth by constantly updating [probability judgments] in proportion to the weight of the evidence.” I will spare you a feeble attempt on my part to go beyond this description (but am linking here to a relatively straightforward explanation of the Bayesian idea by George Dvorsky, “How Bayes’ Rule Can Make You A Better Thinker”, IO9, April 2013), and will instead provide the type of example that helps me to grasp (I think) the underlying concepts. Taking our Saudi financial pressures case, let’s say that we are pretty confident at the outset that, due to falling oil prices and the costs of the Yemen war, Riyadh will face a foreign reserves crisis within five years. But we then learn that the Saudi government has embarked on a program of economic diversification that aims to strengthen its financial position. The Bayesian approach would direct us to think about how likely it is that such evidence would exist in a scenario in which there is a foreign reserves crisis as opposed to one in which no such crisis takes place. If our estimate of likelihood were significantly lower for the former, then there would be reason to change one’s judgment—although based on this one piece of evidence, we would probably remain fairly confident given our initial positive estimate (reflecting a variety of evidence). If we judged the difference in likelihood between the scenarios to be small (e.g., we believe that the diversification program would face serious obstacles in a traditional society) then we would only change our view modestly at most. However, if there were a stream of reports about the diversification program having a growing impact over time, we would progressively back away from our starting hypothesis.
The Bayesian approach would appear to be an ideal fit with a methodology that combines development of an initial set of stated or implied hypotheses through the mapping process along with regular updating based on new information. I would not advocate using formal Bayesian calculations as this would require developing numerical probabilities for hypotheses and evidence: a complicated task for a large map/model and leading us in the direction of prediction rather than foresight. But it would be quite doable to add a dash of Bayesian thinking to our scanning and assessment process by paying attention, as appropriate, to the “diagnosticity” of new evidence and adjusting rough probability estimates accordingly. Tetlock and Gardner (Ibid.) indeed emphasize that their “superforecasters”—individuals with proven records of short term predictive accuracy—use Bayesian thinking informally rather than crunch numbers to strengthen their prognostic efforts. By proceeding in this way we might be able to reduce errors in assessment, such as those associated with overvaluing vivid developments or undervaluing the impact of the steady accretion of modest changes.
In future postings we will progressively build out the various groupings of variables and connect more of the nodes. Once again, we look forward to you comments and contributions.
*The original quote comes from Carveth Read: “It is better to be vaguely right than exactly wrong.” Carveth Read, Logic, deductive and inductive (1898), p. 351 (ebook).
Dr Warren H. Fishbein is a Washington, D.C.-based independent consultant focused on foresight and warning for global security issues (see bio). He is part of the team developing the RTAS Temporal Observatory project.
Featured image: Middle East in 2002 by National Geophysical Data Center – NOAA.