Nancy and I were eating at Jolly Good in Melbourne, talking about everything, like we tend to. In the background, the music they were playing was incredible. The first song I noticed was Streets of our town by The Go-betweens, followed by Somebody to Love by Jefferson Airplane, Tin Soldiers by Stiff Little Fingers, with a couple of great songs I’d never heard before mixed in — Italy by Rain Dogs and Magic of Meghan by Dry Cleaning.

When the Dry Cleaning song came on, I said to Nancy “There’s no way an algorithm made this playlist!” All the songs were great, but there was no thread between them. No era, no genre, no single nation of origin (a Spotify favourite). I’ve listened to enough algorithm-generated playlists to be pretty confident that this wasn’t one.

As we were paying our bill, I asked the owner what the music was. Her response: “Oh, it’s playlist our friend Craig put together for us.” See? I told you.

What Pulls it Together?

It reminded me of a great mix tape (props to anyone reading this old enough to remember or to have made one of those!) It also reminded me of when I used to do radio shows. By the time I graduated from college, I had done several thousand hours of radio shows on the campus radio station. Some jazz, some reggae, a very tiny amount of classical (I was never let on again after I started a show I was filling in on an emergency basis by saying “Today we’re doing to take a look at 5th symphonies, starting with Dvorak…”), but mostly I played rock — punk, garage, proto-goth, what eventually became “alternative”, that sort of thing.

At the end of each show, I had a sense of whether or not I thought things went well or not, but I couldn’t explain why. One weird thing that I noticed was that on the days that I thought hadn’t worked well, I had just as many people tell me they liked the show as I did on the days when I thought everything was clicking.

How could that be?

I’ve only recently come up with a good hypothesis about this, many years after the fact. For most of the people giving me feedback about the shows, the only criteria by which they judged them was “Did I like the songs?” If they did, they told me it was a good show. Fair enough — that’s the whole point of radio, especially if at least one or two of the things they liked were songs they hadn’t heard before.

But to me, great music was a given. I was always in the booth with the monitors on loud, dancing around, and generally making an idiot of myself in private, until I went on mic to announce the songs, when I would make an idiot of myself in public. So it wasn’t the music that was making it work or not work for me. There must be something else.

At the time, I thought it was a sense of flow. I always had an inherent sense afterwards if things had “fit together” in a way that made sense to me. That’s something that has carried over to a lot of things in my job now — lectures, public talks, workshops, and all the different kinds of writing that I do. Part of it was definitely flow.

But there was something else too.

Wholeness

I now think that it was what Christopher Alexander calls wholeness in his books series The Nature of Order. Alexander is best known for his early books that re-thought architecture. A Pattern Language laid out 253 patterns that make buildings and towns more amenable to living full, rewarding lives. The Timeless Way of Building then explains the actual methods to follow to achieve these patterns. This work had enormous influence not just on architecture, but other fields as well, particularly software engineering.

In The Nature of Order, Alexander is even more ambitious. He tries to go one level up to see if there are general rules that generate the 253 patterns that he found in his earlier work. In doing this, he grapples with the intersections between complexity theory, architecture and design, and other big ideas. Like I said, ambitious. And he does find some general rules — fifteen of them, in fact — that generate what he calls wholeness. These are the rules:

Here’s a good blog post that explains each of them with some examples from nature. In thinking about my shows, I actually think that the tacit knowledge bit that I was trying to explain is exactly this – wholeness. It turns out that a number of Alexander’s rules apply to what I thought of as a good show.

Here’s a small example: Rule 9 Contrast. To me, a great show was one that touched on a wide range of genres and eras, different speeds and moods, some loud songs and some quiet ones, and so on. Contrast is really important. And there’s an art to putting together contrast. It’s relatively simple to put together one set that’s fast and loud, while the next one is quiet and calm. That’s good, but it’s even better if you can find a way to blend between these extremes within a set.

Here’s a great example of how you might do this. It’s from my most-listened to songs list from last year**

I love both these songs, but they’re very different. The Agnes Obel song is quiet, slow, and beautiful. The Propagandhi song is, well, none of those. But it starts quietly before it kicks in. So when you play the songs in this order, you can shift from quiet to loud without it being too jarring. Contrast!

There are probably another 8 or so of the rules that I was unconsciously using to pull together a good show. Since reading The Rules of Order, I think that this explains where the good shows were coming from. The rules that were incorporated created patterns that lead to a good show as an emergent outcome. And I think there’s a LOT of value in working out how we can do this more consciously.

Where Does Wholeness Come From?

This isn’t really a post about AI, but I guess in some ways it is, since I started with machine learning algorithms. Christopher Alexander’s work is a good example of doing things that AI can’t do well (at least currently). Once you define the 253 patterns, AI can identify examples that reflect them. Machine learning can identify patterns, and it can cluster ones that appear together frequently (correlation), but the reason that Alexander’s work was a breakthrough was that he identified the patterns using creative thinking, and, more importantly, he made the distinction between patterns that are good, and generative, and those that are bad, and limiting.

That’s based on values. And on thinking based on values.

This is even more true for the work he did in identifying the fifteen rules that generate these patterns. This is in part because these are examples of abductive reasoning – inferring the best explanation for observed data. When you combine abductive reasoning with values-based thinking, you have something that is still uniquely human.

Values, judgement, and inference lead to wholeness, and so do love and care. These are what made my good shows good, and they were there in the Jolly Good playlist too. These are small examples, but Alexander’s are bigger and more important. These examples show how applying simple rules at a local level can lead to good emergent outcomes at a higher level. We can do this in a range of contexts, though it might be particularly important to keep these ideas in mind for the work we do that actually involves AI.

We can ask of any new idea: would making this real increase wholeness, or break it?

*I don’t think that me playing songs from Eternally Yours by the Saints was foreshadowing a move to Brisbane 15 years later, but I guess you never know…

**This is kind of algorithm-driven, but in theory at least, the list is just based on a stack-ranking of the songs I listened to the most, so both of these were in my top 20 last year.

Note: it’s been a few years since I’ve written anything here (thanks to Jason Fox and Nilofer Merchant for the prompts to get back to writing). I do have a number of ideas I’m trying to work out, so I’ll be showing up here a bit more frequently. Your interests may well have changed since my last post, so if you’re getting this via email and it doesn’t resonate, or if you can’t even remember subscribing in the first place, I won’t feel insulted (or even know!) if you hit the unsubscribe link.