One of the abiding problems in working on strategy is that people like to set goals for complex problems. In my experience the hardest part about accepting complexity as a real thing to be wrestled with, is that one really, truly doesn’t get to set goals.

What happens instead, is that one can decide on a direction of travel and then look for the different ways one might move in that direction. It’s deceptively simple, and the hardest part is stopping oneself from measuring progress against goals!

Today, as part of a post that aims to distinguish different kinds of systems thinking from anthro-complexity, Dave Snowden posts a useful summary of this approach, in answer to the question, does anthro-complexity have goals? (I’ve added links to anchor the references):

Yes, but not in the Senge sense. In complexity, you look for the adjacent possible: the next move or moves that are available from where you are, and that requires mapping at some scale first, understanding the terrain as it currently is, not designing what you’d like it to be. [Viable Systems Models] and System Dynamics both seek to understand context, too (Beer’s diagnosis, Forrester’s and Senge’s leverage points), but they do so by imposing their map on the situation rather than surveying it. They arrive with the five-system architecture, or the causal loop diagram, already built, and then locate the organisation’s pathology or leverage point inside categories that were fixed before anyone looked. That’s imposing a map, not doing cartography. [Michael C. Jackson’s] own move with [Critical Systems Thinking] is better: the lens is built from how the people actually in the situation perceive it, not from a pre-specified causal structure. But it’s still a lens rather than a map, assembled once in a workshop and treated as sufficient thereafter, which is exactly where the perception/reality problem and the scale problem both bite.

So there are goals in a complexity account of change, but they are just adjacent possibles, what the map shows is actually reachable from here, rather than targets set independently of, or prior to, the map. The goal is downstream of the cartography, not upstream of it.

That still leaves Jussi Mäkelä’s follow-up question from the comments: if “good” means “preserves future optionality,” isn’t the evaluation circular, optionality judging itself? It would be, if you were running one probe and checking it against one criterion. But that’s not the unit of analysis in a complex adaptive system. You run parallel probes around any coherent hypothesis, several small, safe-to-fail moves at once, not one big bet – which is also where the plural in the title comes from. It’s the next right things, not thing, because in a complex adaptive system you never do this one probe at a time; you do it several at once, deliberately redundant, so that no single move is carrying the whole bet. And in combination, the substrate they’re all running on changes. The landscape isn’t fixed while you sample it; your own probes are altering it as they go. So you’re never judging one probe against a static rule. You’re comparing how a set of variants performs against each other, on a landscape they’re reshaping. That’s not circular; it’s relative and comparative, which is a different thing. Optionality isn’t a yardstick applied from outside the process; it’s a property of the parallel set as a whole, visible only after you’ve run it.

That’s also the answer to the scale problem CST runs into. A workshop needs everyone in one room, at one time, which is exactly what doesn’t scale. Parallel probing doesn’t: mapping becomes continuous and distributed, run through far more human sensors, spread across an organisation, than any workshop could ever seat.

There’s also room for something like a north star: a long-range navigational aid, not a destination coordinate. Kennedy’s “put a man on the moon by the end of the decade” is the standard example, and it’s instructive for what it actually required. It left enormous room; it specified neither the rocket, the trajectory, nor the method of landing, only a direction and a deadline. And it was set only after considerable pre-work by von Braun and others to establish that it was technically feasible, not merely inspiring. A north star that skips that pre-work, or that closes down the room it needs to leave, turns back into a fixed goal wearing a nicer name.

The last piece is path dependency, and it’s where the two accounts diverge the most. Ours takes it as a first-order constraint: where you can go next depends on the specific history of how you got here, and the map has to represent that history rather than abstract away from it. Systems thinking, across the traditions in this exchange, mostly doesn’t. Idealised design brackets history out by assumption. VSM and SD build a causal architecture that’s meant to hold regardless of a particular organisation’s history. Even CST’s workshop, for all its attention to perception, produces a lens for the room on the day, not an account of how that room’s perceptions became path-dependent in the first place.

That’s the argument in outline. How you actually build the map, the mechanics of adjacent possibles, how probes get chosen, how a north star’s pre-work gets done without smuggling a fixed destination back in. I’ve written it up in multiple posts and need to summarise them at some stage in the future.

The post is useful because it traces the distinctions between different kinds of cause and effect assumptions behind different schools of thought that try to address the complexity problem. To most of us, this doesn’t appear obvious on the face of it, but it is really important to understand.

To wit, the important lessons I take from anthro-complexity, and from this post in particular are:

• Causality is not knowable in advance, so we cannot know what will take us from “here” to “there”
• We can find Our way with multiple safe-to-fail probes that help us understand what the “next right thing” to do is.
• Our current state matters, as does the context in which our work is embedded. That will make something more possible in this moment than others (and conversely “we’ve tried that before” might not be an accurate reflection of whether something will work now). This is called “path-dependency” in complexity.
• Data precedes the framework. If I bring a model of change to the work, then everything channels itself into that map or framework. Instead it’s best to start out looking at the current state of things through narratives and to make sense of that current situation together in order to give us a better chance to see different ways of moving in the direction we want to go.
• “Alignment” – traditionally speaking – is not desirable in complexity. Orientation to our direction of travel is useful. When we don’t know where we are going, having everyone pull together in the same direction introduces a tremendous amount of risk. We risk closing off our awareness of possibilities that might emerge along the way.

I’m curious about the ways in which you encounter “goals” in doing truly complex work.