Solving Wicked Problems Requires Two-Minds Thinking

 

"Wicked problems" are seemingly intractable societal issues that affect many areas of everyday life -- some common examples are how to educate people, how to handle climate change, and how to respond to a pandemic. The term was coined by systems theorists Horst Wittel and Melvin Webber in 1973. Some important characteristics of a wicked problem are that:

1. they have no single definition, so that different people can see the problem in different ways. For example, the COVID-19 pandemic was viewed by some people as an imminent health threat, by others as an economic threat, and by still others as a fundamental conflict between government regulation and personal liberties. Applying any proposed solution to a wicked problem (e.g., people should wear masks to prevent transmission) often creates other problems (e.g., what about people who have trouble breathing, have anxiety, rely on perceiving others' facial expressions, just don't feel like wearing a mask, use the anonymity of masks to commit crimes, etc.)?
2. they have no single solution, so that different people can declare victory at different times based on different criteria, and there is no clear way to judge between these claims. Again in the case of COVID-19, was the problem solved when lockdowns ended, when a vaccine was developed, when infections or deaths fell below a certain level, when the WHO declared an end to the global pandemic, when the economy recovered (if it did -- another wicked problem that allows for multiple definitions), or has it never been solved at all?
3. they cross traditional disciplinary lines. For example climate change is not a problem solely of the environment, public health, energy policy, or resource management, but has aspects that fall within each of those domains.
4. they are not amenable to experimentation, because many proposed solutions can only be tested through large-scale implementation, which may have unintended consequences and may not be reversible. In the case of climate change, one proposed solution is to fire trillions of tiny reflective particles into Earth's atmosphere to deflect solar rays. It is impossible to know all possible consequences of such an action in advance.
5. have unclear lines of responsibility for both the problem and its solution. In the climate change example, it is not even clear what nation or group would have the right to take an extreme action like geoengineering. In another contemporary example, artificial intelligence, there are real concerns about creating a "superintelligence" that could be harmful to human life, yet governments are rushing ahead in their development of AI systems out of fear that someone else will develop such a technology first.
6. have potentially severe consequences. Pandemic management, geoengineering, and AI superintelligence are all examples in which the range of outcomes includes possibilities from future utopia to species extinction. Policy makers generally get one shot at solving the problem, their options are often constrained by technical or political realities, and yet they are held completely accountable for results that were to a large extent unknowable in advance. Wicked problems therefore demand learning without having enough relevant experience to be good at solving them.

The opposite of a wicked problem is a "kind problem," in which there are clear criteria for success, the range of potential solutions is defined in advance, and people who work hard to develop a known set of skills are generally able to succeed. Kind problems can be handled through specialization and practice, as in the case of virtuoso musical performers, chess masters, or star athletes. Want to get better at solving this type of problem? More practice is the answer. But life is mostly wicked problems. In the words of psychologist Robin Hogarth, most problems we encounter are a form of "Martian tennis," a game where the rules are unknown, the goals are unclear, and the correct way of playing can change without notice. Dr. Hogarth argues that over-learning a particular solution, through the type of 10,000-hours practice patterns that are effective for kind problems, can actually make us less successful in solving wicked problems.

David Epstein's book Range: Why Generalists Triumph in a Specialized World argues that we can best prepare ourselves to solve wicked problems by gaining knowledge and life experience across a variety of different disciplines. Some of his examples include Nintendo's global success with video games by repurposing old technology, science teams that produce breakthroughs by drawing analogies to distant fields, and NASA engineers who were unfortunately in the minority but who correctly predicted the Challenger space shuttle disaster. On the negative side, he critiques over-specialization's effect in limiting people's thought, with research showing for example that patients are more likely to survive a heart attack when treated by less-qualified personnel while all of the top cardiologists are away at a conference. Epstein's examples provide a clear argument for the value of a traditional liberal-arts education in which students learn the basics from many different disciplines, at a time when more students are choosing STEM specializations early and neglecting the arts and humanities. 

I certainly agree that there is value in breadth of knowledge. Where I disagree with Epstein's perspective is in his belief that only the abstract Narrative Mind is nourished by this type of cross-disciplinary fertilization. Epstein specifically critiques naturalistic decision-making research that shows most experts make decisions using their Intuitive Mind, such as experienced firefighters who report that there was only one possible solution in their evaluation of how to stop a blaze. He prefers Kahneman's more abstract approach to problem-solving using the Narrative Mind, arguing that abstract conceptual reasoning will get us farther than Intuitive problem-solving that is likely to be based on over-learned but non-generalizable patterns. For instance, he gives the example of a "Fermi problem" such as "how many piano tuners are there in New York City?" The problem is not easily solvable by guessing, because most of us have no experience with the business of piano-tuning. But it can be reasoned out with back-of-the-envelope estimates, for example by starting with the number of people who live in NYC (8.58 million, an easily findable statistic), thinking about how many might live in each household (maybe 2-4), what percentage of households have a piano (5%?), how often a piano might need tuning (once every 2-3 years?), and how many pianos a tuner might service in a year (2 per day x 4 days a week x 50 weeks per year?). My rough estimates here suggest that the number of piano tuners in NYC might be in the range of 100 to 150 people. This is surely not completely correct, but the estimation process tells us that a guess of 5,000 piano tuners is certainly an order of magnitude too large, and a guess of 10 is much too few. A quick check of public advertising sites suggests that there are several hundred NYC piano tuners, but less than 1,000, so the math seems close.

I suggest that Epstein has under-valued Intuitive-level thinking in the process of solving wicked problems. Some of the steps in his Fermi-problem reasoning are not purely functions of the Narrative mind: For example, I started by thinking about how many pianos a tuner could maximally get to in a year, but then realized that most self-employed people will take some days off, not work a full 8 hours each day, etc. That realization led me to estimate 2 pianos per day for 4 days a week, rather than 4 pianos times 5 days a week. I also realized that there was likely to be a significant amount of travel time that wouldn't count towards the actual number of pianos tuned. Again, it's only a rough estimate, but my guesses were based on my experience as a human being who has done some traveling work, and not purely on abstract logic. It is a form of "embodied cognition," as the AI researchers might say. And it's also reminiscent of Klein's naturalistic decision-making findings, in which people drew on experience to arrive at conclusions that were pretty accurate, but that couldn't always be easily put into words. 

Research also supports the idea of creativity as a function of analogical reasoning, just as Epstein suggests. But analogies are also more strongly based in the Intuitive Mind than Epstein gives them credit for. Famous examples like Keukle's visualization of the benzene ring as a snake biting its own tail illustrate how strongly Intuitive many creative insights can be, even though these "aha" moments have the greatest effect when they then present themselves to a properly-prepared Narrative Mind. I have previously argued that creativity in fact requires both Narrative and Intuitive thinking, and I believe that Epstein's examples support that assertion. In the Nintendo example, the Intuitive insight suggested a new use of old technology, but careful Narrative planning was needed to operationalize the idea. In the Challenger disaster, part of the novel insight about the fragility of O-rings under cold conditions was based on logical analogy, but part was also a "gut feeling" that something was just wrong. In another contemporary wicked problem, online misinformation, I have argued that excessive Narrative thinking is partly what leads people astray, and that Intuitive-level trust-building can help to reduce the damage. Intuitive thinking can be a problem, for certain, but it can also lead us to novel solutions. In the case of wicked problems especially, Narrative thinking is unlikely to get us to a solution on its own.

I will end with another example from Epstein's book that shows the superiority of Two-Minds thinking in solving kind problems as well as wicked ones. Research on chess shows not only that AI models can now consistently beat human grand masters, but also that human-AI teams (referred to as "centaurs") can beat the unaided AI models more times than not. The winning approach is to let the human team members hash out an overall strategy, and then let the AI model suggest specific tactics to achieve it. This leverages the AI's ability to consider a broader range of options than a human might see, and to project more moves into the future using brute-force computation. AI expert Gary Marcus suggests that humans will eventually lose the race to AI in almost every kind problem, but not in wicked ones: "In narrow enough worlds, humans may not have much to contribute much longer ... [but] in open-ended real-world problems we're still crushing the machines" (Epstein, p. 29). That's because humans have Intuitive Minds in addition to our Narrative ones.