Improv for Ideas

TLDR: Our responses to early, exploratory ideas should be improvisational rather than argumentative.

In The Beginning of Infinity, David Deutsch proposes that the creation of knowledge encoded in genes is similar to the creation of knowledge by people. Knowledge creation involves two phases. It starts with mutations, changes or errors in the replication of the DNA sequence, which are similar to our creative leaps or conjectures. This resembles the creative or brainstorming phase. Next, via natural selection, mutations or variations are selected for in order to survive particular environments. This resembles the competitive or argumentative phase. It involves selection of specific ideas among all the conjectures we have come up with. Let’s call the first phase hypothesis generation (mutations/variations) and the second phase truth filtration (natural selection).

The first phase requires a high velocity of creative, surprising ideas. The second phase requires a careful, reasonable breakdown of these ideas. Contributions to the first phase involve creativity at best and fraud at worst. Contributions to the second phase involve thoroughness at best and pessimism at worst. We need both.

This essay talks about the first phase, where what matters is that which is surprising. A mutation is a mutation only if it is different from the original, expected sequence; in other words, a surprise. One of the ideas in information theory, which underlies modern computing, is that information is surprise.

What disseminates through Twitter is what is surprising. You scroll through the mundane and pause at the surprising. Billion dollar start-ups were surprising ideas. Investments that generated high returns were surprising. If they weren’t surprising, other people would have started these companies or invested in these ideas and then they would not have had outlier outcomes. Economic growth is surprising because it primarily relies on the flow of new ideas. Stand-up comedy that works is surprising. One of the theories on laughter is that we laugh when we are surprised.

All these also need to be “true” in some sense, but they have to be surprising to begin with because if they weren’t, they would be arbitraged away, either emotionally (e.g. Twitter, stand-up) or financially (e.g. start-ups, investing). We crave surprise, however, we want it in appropriate doses. Sociologist Murray Davis said in one of his papers that if an idea questions one of your weakly held beliefs, you accept it as interesting. If it questions one of your strongly held beliefs, you dismiss it as insane. (I think that if an idea is completely new and doesn’t wrestle with any of your beliefs, you might not notice it.)

What helps the first phase, hypothesis generation, is typically different from what helps the second phase, truth filtration. Truth filtration comes from peer review, markets, experiments, and democracy. It requires criticism, devil’s advocacy, error correction, and honesty. On the other hand, hypothesis generation requires coddling, encouragement, and a suspension of belief. It requires an improv-like “yes and” rather than “no but”. It tends to come from solitude or small groups (e.g. Einstein, Satoshi Nakomoto, start-ups) because most large groups devolve into devil’s advocacy.

We have many institutions, incentive structures, and emotional responses for truth filtration. We can instinctually argue the other side when someone shares an idea. Twitter, Reddit, and Hacker News are all suited for truth filtration; users are quick to call out BS or offer counterarguments. Cunningham’s Law is a good articulation of this – “the best way to get the right answer on the Internet is not to ask a question; it’s to post the wrong answer.” Mutations are “wrong” in some sense and most will never be selected for, however, they need to be encouraged to reach their potential before weeding them out. So while society has a plethora of mechanisms in place for truth filtration, it has poor institutions, incentive structures, and emotional responses for hypothesis generation.

Some significant, especially new, ideas may begin as tentative hesitations. In Jony Ive’s eulogy of Steve Jobs, he said, “He (Jobs) better than anyone understood that while ideas ultimately can be so powerful, they begin as fragile, barely formed thoughts, so easily missed, so easily compromised, so easily squished.” Solitude is a good starting point, but discussions with people add creative value – they make your mind race in interesting directions. Improv is a good framework to think about what helps hypothesis generation. The key elements of improv are listening, trying to find ways to make an idea work, and being present. We should have a few institutions and emotional responses that help respond to a new hypothesis by assuming that this is our new reality and running with it. These ideas are like newborn babies; they are too young to be killed with criticism, even if that is honestly what we think about doing at that moment.

In the first phase, anyway, truth does not matter as much as surprise. Instead of, “why could this be wrong?” a good response would be, “with what worldview could this be right?”. This involves suspending your belief for a moment and, like an improv artist, uncompromisingly building on what the other person is thinking and feeling. It is a pure engagement with the idea without bringing in how many other people have tried it before and failed. It is a fantasy world in which this idea is new, original, and deserving of being explored to its fullest potential. You want the hypothesis generator to leave the discussion feeling energized and with a fuller version of her idea.

Improv isn’t about mere agreement. Many times, agreement is as boring as disagreement. When someone says A, the typical response is “yes, A!” or “not A!”. A better response would be “if A is true, then B” or “why A?” or “A adds to C, which this other person proposed”. Improv involves deeply empathizing with an idea, identifying its premise, and then taking that premise in an interesting direction. A good response to A can’t just be B if B is unrelated, but B is a good response if you use the premise in A to get to B. When someone proposes A, we often think that the purpose of the conversation is to decide between A and not A. But A can lead to B, C, D, E, and an infinite number of other options, so merely agreeing or disagreeing with something exploratory isn’t as interesting.

In practice, it may be difficult in some cases to separate hypothesis generation and truth filtration. The process of creating knowledge is fluid; we are making creative leaps as we are discovering what is true. However, there are ideas for which this is clearer. The improv approach is suitable for early, exploratory ideas; for example, “what if everything is priced in life years?” or “maybe we can replace cars with …”. It isn’t suitable for long-drawn, argumentative ideas like “abortion should be legal”, “free will doesn’t exist”, “Seinfeld is better than Friends”, “intelligence isn’t inherited”, etc.

What’s misleading about the phrase “truth filtration” is that while we might think that we are filtering for truth, sometimes we are not creative enough to understand that something is true. Stewart Brand put it well – “Reality is routinely stranger than fiction because reality has no obligation to sound plausible.” Also, sometimes we might just be focused on winning an argument instead of finding the truth. Winning an argument often involves disproving the other side’s weakest point, while finding the truth involves trying to disprove your strongest point. It would be nice to have a platform with the thoughtfulness of Hacker News but that thoughtfulness put to build on rather than reject exploratory ideas.

Hypothesis generation is important because without mutations or creative leaps, there would be nothing for evolution to select for. We need more institutions, incentive structures, and emotional responses that encourage creative leaps; we need some version of improv for ideas.

…….

Thanks to Sarah Friday, Shwetha Hariharan, and Daniel Schlabach for reading drafts of this essay.