Over the last two years, a group of scholars from disciplines including political science, political theory, cognitive psychology, information science, statistics and computer science have met under the auspices of the MacArthur Foundation Research Network on Opening Governance. The goal of these meetings has been to bring the insights of different disciplines to bear on fundamental problems of collective problem solving. How do we best solve collective problems? How should we study and think about collective intelligence? How can we apply insights to real world problems? A wide body of work leads us to believe that complex problems are most likely to be solved when people with different viewpoints and sets of skills come together. This means that we can expect that the science of collective problem solving too will be improved when people from diverse disciplinary perspectives work together to generate new insights on shared problems.
Complexity theorists have devoted enormous energy and attention to thinking about how complex problems, in which different factors interact in ways that are hard to predict, can best be solved. One key challenge is categorizing problems, so as to understand which approaches are best suited to addressing them.
Scott Page is the Leonid Hurwicz Collegiate Professor of Complex Systems at the University of Michigan, Ann Arbor, and one of the world’s foremost experts on diversity and problem-solving. I asked him a series of questions about how we might use insights from academic research to think better about how problem solving works.
Henry: One of the key issues of collective problem-solving is what you call the 'problem of problems' – the question of identifying which problems we need to solve. This is often politically controversial – e.g., it may be hard to get agreement that global warming, or inequality, or long prison sentences are a problem. How do we best go about identifying problems, given that people may disagree?
Scott: In a recent big think paper on the potential of diversity for collective problem solving in Scientific American, Katherine Phillips writes that group members must feel validated, that they must share a commitment to the group, and they must have a common goal if they are going to contribute. This implies that you won't succeed in getting people to collaborate by setting an agenda from on high and then seeking to attract diverse people to further that agenda.
One way of starting to tackle the problem of problems is to steal a rule of thumb from Getting to Yes, by getting to think people about their broad interests rather than the position that they're starting from. People often agree on their fundamental desires but disagree on how they can be achieved. For example, nearly everyone wants less crime, but they may disagree over whether they think the solution to crime involves tackling poverty or imposing longer prison sentences. If you can get them to focus on their common interest in solving crime rather than their disagreements, you're more likely to get them to collaborate usefully.
Segregation amplifies the problem of problems. We live in towns and neighborhoods segregated by race, income, ideology, and human capital. Democrats live near Democrats and Republicans near Republicans. Consensus requires integration. We must work across ideologies. Relatedly, opportunity requires more than access. Many people grow up not knowing any engineers, dentists, doctors, lawyers, and statisticians. This isolation narrows the set of careers they consider and it reduces the diversity of many professions. We cannot imagine lives we do not know.
Henry: Once you get past the problem of problems, you still need to identify which kind of problem you are dealing with. You identify three standard types of problems: solution problems, selection problems and optimization problems. What – very briefly – are the key differences between these kinds of problems?
Scott: I'm constantly pondering the potential set of categories in which collective intelligence can emerge. I'm teaching a course on collective intelligence this semester and the undergraduates and I developed an acronym SCARCE PIGS to describe the different types of domains. Here's the brief summary:
- Predict: when individuals combine information, models, or measurements to estimate a future event, guess an answer, or classify an event. Examples might involve betting markets, or combined efforts to guess a quantity, such as Francis Galton's example of people at a fair trying to guess the weight of a steer.
- Identify: when individuals have local, partial, or possibly erroneous knowledge and collectively can find an object. Here, an example is DARPA's Red Balloon project.
- Solve: when individuals apply and possibly combine higher order cognitive processes and analytic tools for the purpose of finding or improving a solution to a task. Innocentive and similar organizations provide examples of this.
- Generate: when individuals apply diverse representations, heuristics, and knowledge to produce something new. An everyday example is creating a new building.
- Coordinate: when individuals adopt similar actions, behaviors, beliefs, or mental frameworks by learning through local interactions. Ordinary social conventions such as people greeting each other are good examples.
- Cooperate: when individuals take actions, not necessarily in their self interest, that collectively produce a desirable outcome. Here, think of managing common pool resources (e.g. fishing boats not overfishing an area that they collectively control).
- Arrange: when individuals manipulate items in a physical or virtual environment for their own purposes resulting in an organization of that environment. As an example, imagine a student co-op which keeps twenty types of hot sauce in its pantry. If each student puts whichever hot sauce she uses in the front of the pantry, then on average, the hot sauces will be arranged according to popularity, with the most favored hot sauces in the front and the least favored lost in the back.
- Respond: when individuals react to external or internal stimuli creating collective responses that maintains system level functioning. For example, when yellow jackets attack a predator to maintain the colony, they are displaying this kind of problem solving.
- Emerge: when individual parts create a whole that has categorically distinct and new functionalities. The most obvious example of this is the human brain.
Henry: You argue that even though collective problem solving is crucial, it is not a core research topic for most academic disciplines. Why is this so?
Scott: Collective problem solving crosses too many domains. It appears in economics, organizational behavior, engineering, psychology, sociology, neuroscience, computer science, engineering, and even ecology. Yet, it is not the main focus of any of these disciplines. No discipline has any reason to take leadership and none can ignore it. I think this is beneficial. It broadens the study of collective problem solving.
Henry: What would a true science of collective problem-solving look like?
Scott: I think a science of collective problem solving is beginning to take shape. Ideally, there would be a collection of core models that organize our thinking and have some empirical purchase. There would also be multiple categorical distinctions – disjunctive and conjunctive tasks – that help us to differentiate types of problems. Good social science will always be messy. There needs to be a mixture of 'lumping' and 'splitting.' Sometimes, we will make scientific progress by lumping several different kinds of problems together, and emphasizing what they have in common. Other times, we will do better by splitting, separating different kinds of problems from each other and focusing on their unique features.