Well-framed problems

Balancing freedom and useful constraint.

Hello again friends,

I’m back in the mountains for a too-brief visit. These past few weeks, I’ve been thinking about how hard it is to frame problems well and how little we talk about problem-framing compared to problem-solving. Acknowledging not-knowing is crucial to framing problems well. By coincidence, I’ve also been thinking about the work of Robert Irwin because he died last week. Irwin’s art practice involved framing and re-framing problems with not-knowing in mind.

tl;dr: Well-framed problems are good strategies for coordinating innovative action across diverse and unexpected domains. A well-framed problem balances between freedom (defining the problem broadly to increase the number of possible solutions) and constraint (being usefully clear about un/acceptable tradeoffs potential solutions are allowed to make).

Cows in the mountains.

Well-framed problems

I like programming conferences that are meant to produce action, and occasionally I get to do it. (I’ve written previously about design principles for good conferences.) This July, I worked with a team from a Singapore Government ministry to convene an AI conference to inform Singapore’s AI strategy.

Now I’m working on another conference where the participants will be academics and practitioners from many AI and AI-adjacent disciplines. Driving this conference is the question, “How AI can be developed and used for ‘good’?” — and AI is changing so quickly both in fundamental technology and in application that even defining “good” is a challenge. Right now, the conference is structured as three work days, during which participants will identify and develop a small number of problems intended to stimulate and guide ongoing AI thinking and development worldwide.

Problems can coordinate diverse action

A well-framed problem is a useful way to approach something big like AI for good, because it stimulates work in diverse (and sometimes unexpected) places while still coordinating action toward a shared goal. Examples of articulating problems as a way to broaden and speed up development of solutions include the UN Strategic Development Goals, Hilbert’s 23 problems, and the various prizes set up by the XPRIZE Foundation.

Here’s the important bit: For problems to work as a coordinating mechanism, they must be framed well. But we don’t think or talk about problem-framing enough.

The artist Robert Irwin said, “The thing that really struck me the most as I got into developing my interest in the area of questions is the degree to which as a culture we are geared for just the opposite … We tag our renaissances at the highest level of performance, whereas it’s really fairly clear to me that once the question is raised, the performance is somewhat inevitable, almost just a mopping-up operation, merely a matter of time. Now, I’m not antiperformance, but I find it very precarious for a culture only to be able to measure performance and never to be able to credit the questions themselves.” (That’s from Seeing is Forgetting the Name of the Thing One Sees, Lawrence Weschler’s excellent intellectual biography of Irwin.)

An Irwin scrim installation at Sprüth-Magers in 2018.

What makes a problem framing good?

In my view, a good problem framing explicitly explores the tension between freedom (it defines the problem broadly to increase the number of possible solutions) and useful constraint (it is clear about un/acceptable tradeoffs solutions can make). It’s probably easier to explain these with examples of different framings for the same problem.

Let’s say the problem is that the energy efficiency of transportation is too low — in other words, the problem is: Moving people and objects around consumes too much energy. The broad, lightly-framed question equivalent of this problem is: How can we reduce the amount of energy needed to move people and objects around? (All problems can be rewritten as questions.)

Let’s look at a few ways to frame this problem/question to see how framing conditions the range of possible approaches to solutions.

One way to frame this problem is to ask, “How can we build a 100mpg car?” This framing leads inevitably to solutions that are cars which have better fuel efficiency and run on gas. The framing is inherently narrow and prevents non-car, non-gas solutions from being considered at all.

A different, broader way to frame the problem is to ask, “How can we create transport systems which reduce the amount of energy needed to move people and objects around?” This broader framing makes it possible to consider solutions that range from more fuel-efficient cars, to multimodal transit, to vehicle-sharing systems.

An even broader framing is to ask, “How can we create cities which reduce the amount of energy needed to move people and objects around?” This framing makes it possible to consider solutions like increasing taxes on cars combined with cross-subsidising public transit (to incentivise people to switch from driving to more energy-efficient public transit), or progressively eliminating street parking (to disincentivise routine car use).

The more broadly the problem is framed, the more freedom there is to consider many types of possible solutions — which makes it more likely that unexpected and thus really innovative solutions are found.

To see why making un/acceptable tradeoffs explicit improves problem framing, let’s expand the last problem: “How can we create cities which reduce the amount of energy needed to move people and objects around while prioritising small/independent businesses?”

This new framing highlights the importance of small/independent businesses. Another way to say it is that good solutions to the problem will not sacrifice the interests of small/independent businesses — specifying tradeoffs puts meaningful constraints into the problem. (Any configuration of un/acceptable tradeoffs can be written as a set of priorities and non-priorities.)

This new framing leads us to consider solutions like pedestrianising small streets (to disincentivise driving while increasing footfall to benefit small retailers), changing urban planning policy to encourage mixed-use development (to make it easier for small businesses to establish themselves and allow people to move around less by living/working/playing in the same places), or reducing energy subsidies (to discourage transportation of goods over long distances and encourage small-scale local production and consumption). This framing also makes it clear not to spend much effort considering solutions like cheap, fuel-efficient private cars (which have historically increased urban sprawl and unintentionally made small/independent businesses less viable).

A problem statement that explicitly identifies acceptable and unacceptable tradeoffs usefully limits the search for solutions — it is a useful constraint because it reduces wasted effort by focusing search on solutions that involve only acceptable tradeoffs and not unacceptable ones.

Messy, hard work — but essential

As Irwin points out, we tend to invest a lot of effort in finding solutions to problems and not much effort on figuring out what makes problems good.

Figuring out what makes problems good involves making tradeoffs explicit and interrogating them in detail — this is the only way to balance breadth with useful constraint in any problem framing. (I’ve written previously about why it is overwhelmingly important for organisations to talk explicitly about tradeoffs when they set strategic goals.)

Finding this balance between freedom and constraint requires messy, hard work, but this work is essential for developing problems that do what they’re meant to do: coordinate innovative action to find solutions representing tradeoffs that are subjectively acceptable.

---

Other things

Conversations on not-knowing

Thurs, 16 Nov: Episode 11 of my monthly discussion series is about appropriate mindsets for uncertainty. We’ll talk about common pitfalls in evaluating uncertain situations and how to avoid them, as well as new action strategies that become visible when we explicitly talk about different types of non-risk not-knowing. More information and tickets here.

Talk to me about setting goals

I’m starting some new research on how people set strategic goals in uncertain environments. I’ve now had 11 conversations with different people — preliminary patterns coming soon. But please get in touch (you can reply to this email) if you’re involved in team/org-level goal-setting and are open to having a short conversation (no pre-work!) about it.

Seoul!

I’ll be there November 6-12, to give a talk about using uncertainty to find new business models at the World Coffee Leaders Forum. Please send me lots of suggestions of Cool Things to See/Do/Eat/Drink and Neat People to Meet.

More cows in the mountains.

See you next time,

VT