Planning-Centric vs Stigmergy-Centric Workflow Coordination

Discussion

Tiberius Brastaniveau:

"As you can see on the OVN wiki, we use something between central (core group) planning and stigmergy. We don't use pure stigmergy because we don't have the numbers (of agents) to go random in all directions till some development trails get more defined. Like Kurt used to say, evolution takes time and lots of resources, that's why engineering makes sense. Pure stigmergy takes time and resources, best to have some experienced people to at least suggest some initial directions. But if we want OVNs to scale, central planning is not the solution. That's why we need to nurture practices that rely on stigmergy, and we must be ready to rely less on planning once scale increases.

So, if you adopt central planning your central tool is a planning tool, like Trello. The document is where the deliverable is. But if you adopt stigmergy the document becomes the stigmergic environment, the territory of exploration, and everything in that doc becomes a record of all your explorations, leading to the end result, your deliverable. So we cannot dissociate the deliverable from the "territory" of exploration or the stigmergic environment. Ants only have their physical surrounding and an inherited signaling language. They don't work with plans or blueprints. The territory becomes the end product as they fashion it to their needs.

So depending on your practice, planning-centric or stigmergy-centric, your central tool will be a Trello or a Google doc. I don't know how to reconcile these two practices as I see them quite distinct in structure. In one case people go first to the trello and then end up in a google doc (or other content management tool). In the other case you go directly to the google doc and get your cues from there, by picking up these pheromones in their respective context, wherever people placed them.

As a general matter, I think we should always adopt tools that are built with data transport ability as a priority. In other words, we should boycott apps that luck data in one place. By data portability I don't just mean the ability to download my data, but the ability to interact with it from wherever I want and even to move it, or parts of it, wherever I want, without losing its integrity.

Back to the comparison between planning and stigmergy, there may be some bridges that we could build, now that I am thinking about it. You can go from a stigmergic environment and translate all these pheromones, which should have some metadata associated with them, and construct not a plan, but a map of action, as it is at a particular moment. A plan projects far into the future assuming a predictive model. On the other side, stigmergy shows what has been done in the past and how, but also some of the immediate future. Stigmergy assumes that the future is uncertain and the target is unknown or at least moving. So you can't get a full plan from the google doc of Sensorica, the stigmergic environment, the territory of exploration. But if you are a planner you can take that snapshot of all the created trails and project that further into the future. And we actually do that, as we are somewhere in between planning and stigmergy. So you can extract a map of actions from a doc, with a thin layer into the immediate future, and format that as a plan for past and near-future actions.

I don't know if you can do the opposit, go from a plan to stigmergy... perhaps, if you cut slices in time in your planning, and suppose that at every temporal slice you have a wise general suggestion of action, and from there let the ants do their random exploration around. "

(email, November 2021)