Becoming Predictive Processors
Using the Predictive Processing Loop to improve your teams
Last time we took a look at Predictive Processing and how it works in one human’s brain. To summarise the hottest parts:
Our minds contain models of the world which we use to make predictions about the world
We then go and have these sensory experiences whilst staying rooted in our prediction
If the sensory experience doesn’t meet the prediction we experience prediction error
Prediction error causes us to update our model and then make different predictions
Cognition is the game of constantly trying to reduce prediction error about the things that matter
Predictive Processing is about the mind making quick and dirty assessments grounded in our existing models. This is generally efficient and allows us to navigate our complex world effectively, learning as we go. Sure, we’ll make errors along the way, but that’s the price we pay for fast assessments which are overall very effective for us. As long as errors can be sensed and lead to updating the model and the next round of predictions, overall this leads to good navigation of the world.
This happens in you even if you consider yourself to be a very slow, deliberative, thoughtful person. It’s unconscious and unavoidable and fundamentally plumbed into how we experience the world in pretty much every way. So claims the Predictive Processing model.
The reason I’m interested in these kinds of ideas is because I think we can use the mechanics of individual intelligence to create better organisations and teams. Predictive Processing allows an individual mind (or a machine-learning mind!) to understand its environment quickly and efficiently to help it act well. I think the logic of Predictive Processing can help teams and groups of people also understand their environment quickly and efficiently to help them act well. This is Organisational Intelligence.
In this post I want to focus on practical implications of this work. As someone in a team, how can this new understanding of Predictive Processing help create more Organisational Intelligence? I’ll propose a broad answer in this post, and go into further detail in the future.
Put in a Predictive Processing Loop
The most basic and first thing to check is whether the broad mechanics of the Predictive Processing Loop are in place at all. There’s no point developing exquisite and fine-tuned mechanisms for sensing with lots of accuracy if you’re in a context where for various reasons the model can only be updated and acted upon at the yearly CEO’s planning gala and budget sign-off.
Here’s a list of things that will be true if Predictive Processing Loop is broadly in place in a team. This might sound basic, but it’s the kind of basic that many many teams - including teams that would describe themselves as agile - do not have in place.
I’m going to share some ideas for tools, meetings, frameworks etc that might help with each stage. Some might help with more than one stage e.g. Refinement I’d say is both model building and action planning. I don’t really know much about project management or Lean 6 Sigma or things outside the agile world - if you have ideas of what else could fit in this list please do get in touch.
The Predictive Processing Loop:
Model and Prediction: For each piece of work your team is involved in, there is a clear idea of what the problem is and what the proximate goals are for the team to solve the problem. This is the model. There is also some kind of explicit or understood prediction of what the team expect to sense if their prediction is correct. (Ideas for Tools: Business Model Canvas, Lean Strategy, OKRs, KPIs, User Story Mapping)
Action: The team is clear on what the actions are that will produce the sensing described in 1, and how those actions will be carried out by the team. (Ideas for Tools: Quarterly Planning, Sprint Planning, Kanplan (who knew that was a thing?), Refinement meetings, Daily Scrum)
Sensing: The team has unambiguous sensors with which they can contact the world (their internal or external customers or technology) and determine whether or not the prediction was accurate. The aim here is to produce Prediction Error if you’re wrong, so if these sensors can’t indicate a negative they aren’t the right ones. (Ideas for Tools: User testing, customer conversations, DORA Metrics, Product Metrics various)
Updating the model: There should be a known and easy method and culture of updating the model and the predictions when prediction error occurs (Ideas for Tools: Sprint Retrospectives, Sprint Planning, essentially the stuff from point 1)
This isn’t necessarily a way of saying ‘look! Do Scrum!’, or some other agile framework, although that is my day job. If you can achieve this process and therefore navigate the world better using Prince2, SAFe, Lean 6 Sigma or whatever then all power to you. I would probably argue that for most of the problems that we face in our different teams and industries, an agile one - if implemented in service of the Predictive Processing Loop above - would be faster and therefore better.
Scrum is one framework that understands this intuitively. It’s contained partially in the Scrum Theory, what I’m claiming is new however is the prediction element. Scrum can be agnostic about how much you use prior knowledge, assumptions and models in your product development and team improvement. The Predictive Processing Loop however has a nice Bayesian quality to it in that it uses past information to inform current assessments, whilst giving you a way of incorporating new information and changing your model when necessary.
In a future post I’ll go into how Precision Weighting can give us a rational way of understanding which of these frameworks we might want to use for different types of problem. Some problems are best solved with agility, others with Lean, others with Prince2. True organisational intelligence is about being able to use the right tool for the right kind of problem.
The upshot - when you’re sitting down with other managers and leaders and scratching your heads over how to work with your team, try checking in the team’s culture and practices with the Predictive Processing Loop. Ask if those elements are true, and therefore if that team has a mechanism in place to allow them to access their Organisational Intelligence. Maybe this will provide some insight that makes an improvement. If this happens (or if you need some help thinking this through), fling me a message.
Drop a like or a comment on this post if you liked it, it’s a small and free thing you can do to help me know that I should keep writing.
Other Stuff:
I’m running an online retrospectives masterclass on March 5th
If you want to talk about Agility and Cognitive Science in person I run a monthly online Meetup that you might like