Chapter 12. Visions and Challenges – Priorities for Professionals

Underspecified Model Mistakes (UMMs)

Perhaps the most common mistakes in the mental models of foresight practitioners is that they are too simple where it counts. They are missing certain key facts, relationships, and drivers that affect the system they are attempting to forecast. What they are missing may constrain their imagination/possibility space, or it may expand it. Let’s explain this claim.

All good foresighters think in systems. They build mental models of relevant causal relationships, stakeholders, and resources, and look for data on trends, cycles, and patterns in system behavior. But it’s easy for these models, even formal ones with lots of funding behind them, to be underspecified (missing factors, incomplete). They can also be impressively complex, quantitative and equation-driven, but still missing the proper weighting and subtlety among the subset of factors that will dominate the system’s behavior.

If a foresighter’s model is complex but still missing or improperly weighting the dominant factors, that is an Underspecified Model Mistake (UMM). This is a more complicated variant of the MEE. To remember this mistake, think: “UMM… you’re still missing something important from your model, future thinker…” Then try to find out what’s missing and bring it into the model.

For a classic example, think of the famous World3 model, a quantitative systems model of global change built by a well-meaning team of environmental scientists led by Donella Meadows and published in Limits to Growth (1972). This team of futurists used their model to predict mass famines by the end of the century. Their model, though it carefully extrapolated many trends, and so was not in any way guilty of MEE, still neglected to include any factors representing the deflationary, efficiency-gaining, and capacity-building effects of technological innovation itself. Because the factors they missed turned out to be accelerating global developmental trends, it made their model rapidly more incorrect over time, and Limits became a very public UMM.

This is not to say the Sustainability-Limits school of foresight has been wrong any more often than its counterpart, the Globalization-Growth school. For examples of how the Globalization-Growth school has been dangerously wrong by minimizing forecasts of ecosystem damage, see The Bet: Paul Ehrlich, Julian Simon, and Our Gamble Over Earth’s Future (2012). Both sides are equally guilty of UMMs. We must always remember, when forecasting complex systems, how easy it is for any model to be wrong — and being too simple, or improperly weighted, is a very common way to be wrong.

One way to reduce UMMs is to have more broad conceptual and technical knowledge prior to getting involved in foresight. Another way is to use the power of the crowd, to have a cognitively diverse set of colleagues and the urge to survey them with enough depth to help you understand the breadth of the mental or formal model you need to build. Having the statistical knowledge to arrange all that survey or Delphi or expert opinion is another great way to reduce bias and oversimplification.

Most important however is to keep an open mind, and be willing to see subtleties and complexities that you previously missed. For example, consider the important question of whether Earth’s resource needs and environmental problems are likely to get worse or better over the next generation. As is described in The Bet (2014), the sustainability foresighter and biologist Paul Ehrlich made a famous bet with the innovation foresighter and economist Julian Simon (see his excellent, The State of Humanity (1996)) that the price of a small collection of raw materials (copper, chromium, nickel, tin, and tungsten) would be more expensive a generation hence. Ehrlich lost that bet, as he greatly underestimated the variables of scientific advance, human ingenuity, process efficiency, market incentives, and resource and technology substitution (the ability to switch from higher priced to lower priced alternative inputs in manufacturing processes). All of this was valuable learning from a failed forecast. We live on an incredibly resource-rich planet, and a little ingenuity and competitive incentive have continually resolved our periodic resource crises. In those years at least, innovation and automation were far more deflationary than increasing regulations, and environmental remediation was inflationary.

Ehrlich then shrewdly offered a second bet that predicted various environmental indicators would be worse in coming years. Simon didn’t take that bet. He was foresighted not to do so, as those particular indicators (pollution, species loss, global warming, and twelve other trends) all did continue to get worse. Ehrlich had learned to focus on the areas where he could win. His mental models had improved. Environmental problems have turned out to be much more problematic long-term issues than our temporary resource scarcities (including energy, water, and other resources).

Continual learning, model making, prediction, and review of those predictions are at the heart of foresight process, and these efforts constantly lead us to better models. Today for example, the Environmental Kuznets curve model (Chapter 4) tells us environmental quality commonly gets worse in early stages of economic development, when we value jobs and innovation over environmental health, then later gets better when social wealth reaches a certain per capita threshold, and sustainability politics and initiatives gain traction and we decide to clean up our messes. But we’ve also learned this particular model doesn’t hold in situations where wealth and power are concentrated in the hands of pro-growth oligarchs, as in Russia and until very recently, China. This tells us that average per capita wealth isn’t the best way to predict environmental quality. Median citizen wealth and other measures of economic democracy are a more predictable (universal) way to predict the relationship between social wealth and local environmental quality, across cultures.

At present, perhaps the most important system affecting global environmental quality over the next few generations is climate change. Here again, even with all the uncertainties, step by step, our data, experiments, models, strategy, and policies continually improve. It’s obvious this is a key problem. Generating the foresight to address it in the most cost-effective ways, including an accelerated transition to solar and electric power, is the main issue at hand.