‘Destination’ and ‘path’ – Photo by lucas Favre on Unsplash
Distillations in this newsletter: ‘Destination’ and ‘path’ – two types of strategic commitment; AI and its relation to strategy.
STRATEGY DISTILLED:
A monthly concoction of insight, learning and things you might have missed for anyone who works on strategy, works with strategy or just loves strategy.
Would you rather listen to this newsletter as a podcast?
_____________________
This month …
- ‘Destination’ and ‘path’ – two types of strategic commitment
- A strategy snippet you might have missed: AI and its relation to strategy.
If you enjoy reading this newsletter, don’t forget to forward it to friends or colleagues who might also find it of interest.
Was this forwarded to you? Sign-up
_____________________
‘Destination’ and ‘path’ – two types of strategic commitment
Okay, bear with me. I’m taking you on a couple of looping detours before we get back on the main path of this post. But I think you’ll find it worthwhile, I’m pretty sure you’ll like the destination and I hope you’ll enjoy the journey. It’s also one I’m particularly looking forward to writing because I read a post recently that forced me to change my mind on something I’ve long believed in. And I love it when this happens. I see it as a robust indicator of genuine progress in my thinking. I’m not just running in the same hamster wheel all the time.
Responsibility for this mind-shift was, perhaps unsurprisingly, Venkatesh Rao – it’s not the first time he’s done this to me (if you haven’t read his work, start with his article on ‘The Calculus of Grit‘, the introduction to his essay series on the nation state or his ‘manifesto’ for protocols). His most recent post opens by summarising an article about how (and possibly why) mice dream. Yuta Senzai and Massimo Scanziani recorded the activity of neurones in two regions of the mouse brain; the superior colliculus (SC) and the anterodorsal nucleus of the thalamus (ADN). The SC is known to be involved in motor control and specifically movements that orient the body in its spatial environment. In the first part of Senzai & Scanziani’s research, mice that were awake and free to move, were found to have specific SC neurones that activated head-turning movements. When clusters of these neurones fired together, the probability of the mouse turning its head in one specific direction increased significantly. Then, once the head has turned, cells in the ADN update their internal representation of the mouse’s head position.
What makes Senzai & Scanziani’s research remarkable is that they recorded what was happening to these same neurones when the mice were (as indicated by other electro-physiological signals) in rapid eye-movement (REM) sleep. Here too, the head-turning SC neurones fired. In response, the ADN neurones showed that the mouse’s representation of head movement also fired … except the mouse’s head stayed motionless. These neurones were evidence of the mouse turning its head in its dream and sensing its head turn in response. The mice seemed to be ‘going through the motions’ of spatial orientation in their dreams. Over twenty years ago, Dave & Margoliash made a similar discovery: Zebra Finches practice their songs in their dreams.
Okay, first detour over, let’s get on with the second one. The conclusion Senzai & Scanziani draw is that dreams are a way of training our brains, whilst asleep, using ‘synthetic data’ – data from our brain’s representation of the world rather than from the real world itself. This is like Jann Mardenborough becoming a professional racing driver from his expertise in the Gran Tourismo computer game (as depicted in the 2023 movie of the same name). Venkatesh Rao goes on to draw the comparison with the use of synthetic data “for making an AI model aware of information it did not encounter during its initial training.”
The advantage of this approach becomes clear in the context of dual control theory. When trying to control a system whose initial characteristics are unknown, according to Wikipedia “the controller’s objectives are twofold:
- Action: To control the system as well as possible based on current system knowledge;
- Investigation: To experiment with the system so as to learn about its behavior and control it better in the future.”
Often, these objectives will be pursued in different ways. To return to a car analogy, we might investigate the handling characteristics of the car in a safe environment (e.g. in an empty car park) whilst acting much more cautiously on a busy road with other cars around.
Right, detours over. What this has to do with strategy is that strategy is all about exercising control in a system whose characteristics are unknown – we are striving to control the future. Consequently, our strategies should have action-oriented elements and investigation-oriented elements. I’ve been advocating this to my clients for several years now, albeit using different language.
- Strategy-as-imperative is a specific and fully-defined instruction – achieve this outcome by this date. It is clear what the desired outcome is and also clear, in broad terms at least, how to get there. The strategy simply says ‘get there! (An imperative is a command, an order or an obligatory act or duty). This is the action element of dual control theory.
- Strategy-as-hypothesis is much less definitive. It clearly defines direction of travel (e.g. increased profitability) but concedes that we don’t know enough right now to define success or the means of achieving it. Our strategy must proceed by proposing and testing hypotheses. This is the investigation element of dual control theory.
In simpler language these approaches to strategy are ‘destination’ (strategy as imperative) and ‘path’ (strategy-as-hypothesis).
So, what’s changed? In what way have I changed my mind on something I’ve long believed in? Well, I used to say to clients that it was important to be clear about whether their strategic goals were imperatives or hypotheses. I would encourage clients to confront the cognitive biases (e.g. Optimism Bias and Over-Confidence Bias) that led them to over-estimate their certainty about the future and their ability to achieve strategic success in predetermined ways. I encouraged them to think whether their strategic imperatives might be better framed as hypotheses.
Now, triggered by Venkatesh’s post, I feel much more bullish about imperatives and hypotheses. So much so that I’d argue a good strategy needs both.
A good strategy depends upon a strong evidence-base, powerful analysis and insight and imaginative construction of the available strategic options. Then, often what is needed is strong strategy leadership: clear decisiveness and lasting commitment to the best of those options. This is powerful strategy-as-imperative.
For strategy to be stretching and inspiring it also needs to extend out into the ‘known unknowns’ and possible even into the ‘unknown unknowns’. It needs to openly acknowledge its own uncertainties. It needs to admit that the destination is, as yet, ill-defined and only the path forward can be discerned right now, to be navigated, during the course of the strategy as a series of hypotheses and hypothesis-testing experiments. In this way, investigation now will lead to much better controlled action in the future – isn’t this what strategy is meant to be all about?
_____________________
A strategy snippet you might have missed
AI and its relation to strategy
For any of you out there who, like me, are trying to make sense of how we integrate AI technologies into organisational strategy, here are three findings that I found highly thought provoking:
- AI seems likely to have a profound effect on how we learn, over the next ten years (see The Metacognition Revolution), in which case it will almost certainly influence how we drive cultural change across organisations in pursuit of strategy.
- AI is expanding its reach into high-speed, highly coordinated, complex actions – like playing table tennis. The ability to learn and execute complex sensory motor tasks has clearly got relevance to putting business strategy into practice.
- On another practical note for us strategists-interested-in-AI, MIT has produced an AI risk repository. I found this interesting for two reasons. Firstly, it provides a great framework for analysing AI risks for any specific organisation or strategy. Secondly, the approach they adopted could apply to other areas of strategic focus – SaaS risk repository? University risk repository? Water utilities risk repository? (to mention three that I am involved with).
_____________________
We, at Goal Atlas, work as strategy facilitators. We provide frameworks, tools and hands-on advice and guidance to support your leaders, managers and front-line teams to develop the strategy that works for you.
_____________________
If you enjoyed reading this newsletter, don’t forget to forward it to friends or colleagues who might also find it of interest.
Was this forwarded to you? Sign-up
Discover past issues of this newsletter