Starclouds

A new book called Blindside will be coming out next week. It’s edited by Francis Fukuyama of “end of history” fame, and is essentially the proceedings of the Blindside conference that was sponsored last year by The American Interest, a quarterly policy journal that Frank co-founded back in 2005. The subtitle of the book (and the conference) pretty well sums up its theme: “How to Anticipate Forcing Events and Wild Cards in Global Politics.” The book ends up with a lot more questions than answers; forecasting is damnably hard, especially the future. But most of the chapters are worth a look.

I mention this because one of the chapters is mine: I use a lot of examples from computer history to explain why it’s hard to do forecasting even for IT, where you’d expect things like Moore’s Law to make it easy. But I’m actually going to save that discussion for a later post. My immediate concern is that this chapter was actually just the middle part of a longer paper (and longer conference presentation) about the perils of technological forecasting in general, which I did in collaboration with Caroline Wagner at George Washington University. Not surprisingly, I think it contains some pretty good stuff. Here’s the original summary passage:

We define the notion of deep uncertainty, which makes prediction effectively impossible in most cases. We explain why that’s OK-because an effort to explore the future can give you a great deal, even without prediction. And then we explore how we can still confront the future quite effectively by embracing deep uncertainty.

Unfortunately, for reasons too mundane to bear repeating, everything besides that central section wound up on the cutting room floor. So, with Caroline’s permission, I’d like to rescue those portions here. Enjoy.

Embracing Deep Uncertainty

By Caroline Wagner and M. Mitchell Waldrop

Never has our species been more obsessed with the future. For most of human history, after all, change was imperceptible: our ancestors lived by the endlessly repeating cycle of the seasons, in much the way their own ancestors had. But today, in an age of rampant globalization and technological ferment, change is constant, unpredictable-and often, it seems, beyond anyone’s control.

Thus our fascination with formal methods of planning and forecasting, with their implicit promise to reduce the uncertainty and give us at least some control over our future.

Given their track record, unfortunately, it is not at all clear that these methods can deliver on that promise. Their roots go back to the late 1940s, when the first wave of enthusiasm for programmed planning and technology forecasting was at least partly inspired by America’s spectacular success in World War II, both on the battlefield and in the laboratory. Indeed, the decades immediately following were a period of striking technocratic optimism in the United States-or striking technocratic hubris, depending on one’s point of view. (See, for example, Robert J. Samuelson’s book, The Good Life and its Discontents.) Rapid advances in economic theory, game theory, social science, and business practices convinced a generation of intellectuals that the world could be rationalized, mathematized, reasoned about, and managed. And in the heady afterglow of victory overseas, followed by years of post-war prosperity at home, that conviction was widely embraced in both government and industry. We now had the tools to maintain the Cold War balance of power, tame the business cycle, end poverty, ensure prosperity, and even suppress insurgencies in far-off lands. In particular, thanks to a variety of forecasting methods, we could now anticipate the direction of technology development in plenty of time to head off any deleterious social impacts.

Things didn’t work out that way, to put it mildly. After chaotic experiences such as Vietnam, globalization, the AIDS epidemic, the unexpected fall of communism and the rise of international terrorism-not to mention the continued existence of poverty-the world has come to seem like a far messier and more unruly place than it did in the post-war years. Messiness has proved to be the norm even in the seemingly clear-cut world of technology forecasting. Mid-century predictions about the year 2000 included assertions that nuclear power would make energy that was too cheap to meter, that the U.S.S.R. would be a dominant superpower, and that the university would replace the firm as the institution most central to economic growth-none of which came to pass. They also included the computer models described in the Club of Rome’s 1972 book, Limits to Growth, which famously painted such a grim picture of future starvation that some pundits were led to consider triage of huge swaths of the human population who were doomed anyway. That hasn’t come to pass, either (yet.) Virtually no one foresaw the coming of interconnected personal computers, the green revolution, climate change, wireless networking or globalization-or the lack of progress in energy. And the few people who did stumble over these possibilities may have been lucky rather than prescient.

The accuracy of the predictions hasn’t improved much since then the 1970s, despite any number of studies, books and methodological developments. To put it most simply, technological forecasting has failed to anticipate the future in any reliable way because it uses linear logic to understand a highly nonlinear process. Or to put it a slightly different way, the forecasters have failed to embrace deep uncertainty.

Ordinary uncertainty is the kind we get in, say, a weather report: “There’s a 30% chance of rain tomorrow.” Many things about the future are unknown, but at least they are known unknowns. We have a good idea of where the uncertainties lie, and a framework for thinking about them.

Deep uncertainty, a term coined in 2001 by Steven Popper of the Rand Corporation, is when we don’t even have the concept of “rain,” or “tomorrow.” The future is a fog of unknown unknowns. We can’t begin to agree on what our mental framework ought to be, or what the probabilities are, or whether a given outcome is good or bad. And even if we could, the real world is rife with complexity-meaning not just “complicated,” but full of non-linear and emergent phenomena that make long-range prediction all but impossible. (”For want of a nail, the shoe was lost…”)

This might sound like a counsel of despair for anyone trying to anticipate the future-which, at some level, is all of us. But in fact, the lesson is not that we should give up. Quite the opposite: foresight is essential, and deserving of constant effort. The lesson is that we should give up any notion that foresight will help us predict and control the future. Instead, we should embrace deep uncertainty. Accept the fact that we’re always going to be exploring our way into the future-and that we should value foresight for what can give us even without prediction.

For example, our efforts to anticipate the future can help us-

• Assess the evolving situation and identify critical factors;
• Think through the significance of those factors;
• Reveal and challenge hidden assumptions;
• Identify the warning signs (or hopeful signs) to look out for;
• Think through our alternatives for response, both tactical and strategic;
• Identify and begin to build a network of experts, partners and allies.

Indeed, by giving up on prediction as the goal, our efforts to look at the future can open the way to a different, and arguably more fruitful, approach to dealing with deep uncertainty. Some principles:

Maintain Eternal Vigilance. Foresight shouldn’t be seen as just a one-time effort, nor should the process be treated as a black box that delivers The Answer. Foresight should be an on-going process, with predictions that are constantly updated, reevaluated, and reassessed. A good model is what psychologists call “situation awareness,” an individual’s evolving picture of his or her surroundings. According to one commonly accepted definition, situation awareness involves three levels: perceiving the critical factors in the environment; understanding what those factors mean, particularly in relation to the decision maker’s goals; and projecting what is likely to happen in the near future. All three levels have to be maintained constantly if we’re to function in a timely and effective manner.

Minimize Potential Regret. Foresight shouldn’t be seen as a way to anticipate the most likely future, or set of futures, as the basis of planning, which is how decision-makers typically use it today. It should be seen as a way to avoid entrapment, by choosing strategies that are adaptive and robust over a broad range of scenarios. What policy courses are available to us, and how do they fare under many alternative futures? What are their failure modes? What are the warning signs? And how can we maximize our ability to respond to the unfolding of events?

The good news here is that information technology, computing, and modeling have greatly advanced since the early days of forecasting, as has our understanding of human decision-making, complexity, and social organization. The result is a new generation of tools that can help for more insightful and robust analysis. One example is the Computer-Assisted Reasoning system. Developed by Lempert, Bankes, and Popper, and described in the RAND report Shaping the Next One Hundred Years, it allows users to map hundreds of possible futures against a landscape of desirable and undesirable outcomes.

Seek Collaborative Advantage. In a world governed by complexity and deep uncertainty, no organization (or nation) can ever be big enough and strong enough to go it alone all the time. Sooner or later, the organization will need more resources, information and expertise than it has in-house. And to get them, it will need to build alliances, coalitions, and partnerships-that is, to seek collaborative advantage rather than competitive advantage.

This is certainly the case when it comes to assessing the future, where there is always another perspective that could prove vital. Indeed, this “million minds” approach is deeply related to how the scientific community makes progress, and how other open systems move towards consensus. Entertaining multiple points of view and multiple forms of expertise not only helps expose error, but also can enable groups to pose questions that the experts never would have thought to include. It also helps us focus on the actions we might actually take, rather than on the outcome of events we cannot control.

Forecasting has been dogged by lack of precision in predicting the future, foiled by the very complexity it tried to contain. But there is great social value in the very process of pulling people together, and engaging in a joint discussion of what actions taken today might increase positive outcomes in the future. Given the challenges facing 21^st century societies, attaching greater value to the wisdom of the crowd may be a prudent direction to take. This could take place in a series of town meetings that combine human reasoning with computer technology to ask what we hope the world will look like in the future, and to recommend policy steps that could create a world we would want to live in. Visions of positive outcomes in the future are critical to social development. As the old saying goes, if you don’t know where you’re going, you’ll probably get there.