Unpredictable Patterns #27: Thinking in Trees
Causal, genealogical, evolutionary and amazing trees to think in and with
Dear reader,
Today is a hot, long summer day in Sweden - and I hope you are enjoying it too — vacation is important, and we all need to relax. Our friends in the US are celebrating 4th of July, and vacations are approaching - I will be off for a few weeks at the end of July and beginning of August. Spending a lot of time on the island has me think about analogies from nature, and one of the most obvious ones is the tree analogy - so that is what we will explore today!
Skip the forest, look at the trees
Trees are fundamental structures in our everyday reality, and they are very valuable tools to think with. We should, sometimes, actually ignore the forest and concentrate on the trees. There are several different kinds of trees that can be helpful for us, and we would be remiss if we did not look closely at what trees can teach us.
One of the most interesting tree models is the tree of life, and it is an amazing thing. The idea that all things on earth are connected with each-other in a tree like structure, and that we all come from a single common ancestor should give all of us pause, and the model of the tree of life is one of the most interesting conceptual discoveries that we have made. Darwin’s original tree was a simple sketch, but the concept was profound.
Genealogical or evolutionary trees now abound. From a simple family tree to a more advanced conceptual genealogy of tech companies, we can explore tree like structures and understand our world as constantly converging on a single tree-like structure.
And even if the model is simplistic - it is interesting. As philosopher Peter Godfrey Smith has noted in his book on Darwinian populations: ”Life is only roughly a tree, but a great deal follows from its being roughly a tree”. Tracing common ancestors, understanding why some branches stopped and never developed and exploring current traits as paths are all cognitive tactics opened up by applying the tree structure to a problem.
The mapped out evolutionary tree is only one possible way for life to shape, or at least so we think. We could easily imagine worlds in which other species had had more success and so the tree would look different, but the tree we actually live in is a key determinant for our environment and the lives that we can live.
Now, one thing this particular model suggests is strict vertical causality - what happened before determines how options form today. That is no increasingly called into doubt, and the picture we get from the most recent research in evolutionary biology is that of what David Quammen has called ”a tangled tree” - a tree where branches can grow into each-other and the genetic inheritance can be horizontal as well as vertical.
Think of when a new company hires a lot of people from an older and different company in another business - what essentially happens then is that the DNA of that older company in a separate branch is horizontally injected into the new company, and its DNA changes accordingly.
Now, the causal tree like structure still dominates, so this is not a network in the proper sense - it is a tree, and that is worth noting. There is a tendency in thinking about connected structures to reduce them all to graphs and networks and then say that the tree is merely a special case of graphs generally.
Mathematically this absolutely accurate, but conceptually that is like equating a hammer with ”tools” and not look at what it can do specifically. When we think about connected structures we should not get stuck at the network level, but allow ourselves to spend extra time with some more specific structures that deserve extra attention - like trees.
This is obvious from an example. One interesting way to understand companies is to look at the flows of recruiting into them as they grow - they will depend almost as much on where they recruit as they will on their own genealogical history. Evolutionary mental models now need to accommodate horizontal gene transfers in order to represent what we know about evolution today. Any analysis of the tech industry today should start from that tangled tree analysis, and what it can teach us about the future of tech - this is especially true for the second generation tech startups that acquire a lot of genetic traits - like the use of the OKR-model for goal setting - from other tech companies.
Horizontal gene transfers can change a Darwinian population without inheritance, and so can short cut evolution in a sense.
Evolutionary trees and genealogy over all also helps in understanding how political opinions form over time - where do these ideas come from? Ideas of search neutrality come from network neutrality and in turn from public utility theory - creating a powerful conceptual tree structure where we can predict new branches will grow over time as new technologies and architectures are introduced.
Decisions, decisions
Another way to use tree models is through the exploration of decision trees. These models are well-known and quite simple to understand. A complex decision is often easier to understand if we decompose it into its component decisions and then we can draw up a decision tree on the basis of that decomposition. The decision tree then allows us to discuss the overall structure of a decision in much more detail.
Decision trees can be designed in increasingly complex ways - from adding weights, probabilities and pay-offs to working with random forests of decision trees in machine learning. Not only is this a useful tool, it is also a quickly expanding research area with any number of different applications.
Decision trees are also used in game theory to help us visualize a game in a tree of different possible tests and outcomes. Mapping out a decision in a decision tree is not helpful because it ”solves” the decision as much as it helps you visualize and think through the way that you have structured the decision: what are the tests, the component choices? How do you think about possible outcomes?
It is also useful to be able to discuss which branches are simply so unlikely that preparing for them is as close to wasting time as we can come.
Decision trees can also be used very efficiently for classifying things, and these classification trees simply have leaves that are conceptual categories. Such classification schemes help organizations sort through incoming issues and classify them accurately in a way that relies on much more than just intuition.
Should you try to engage with a particular legislative proposal or a particular ask to come talk at a conferences? These recurring questions almost deserve classification trees so that you can start behaving consistently and surface what the classification should be based on!
Many times the value of the decision tree lies in articulating it - getting everyone to agree on what the choices and classes and categories actually are in a given set of type situations. The fear that this will render you inflexible should be tempered by the guarantee of consistency that you can acquire, and the speed with which you will be able to make new decisions - rather than agonizing endlessly over what to do next.
Capability trees or tech trees
Capability trees are different mental models, and we have spoken about them before. Anyone who has ever played a computer game will know these as tech trees, where you unlock complex technologies and sometimes social policies by unlocking simpler technologies and social policies. These then connect in a tree that slowly allows you to move from agricultural to information societies.
A capability tree is also a useful way of analyzing a competitor. If we understand what they can do routinely and well, we can start asking what this capability is built on and how they have achieved those earlier capabilities - are there any points of learning for us here?
A company that can get their customers to speak up on their behalf in Washington has almost certainly built that capability on strong customer support, genuine knowledge of what the customers care about and an ability to engage in customer communication on range of other issues - building loyalty and momentum. As we decompose that capability we also understand the basis of their strength and can learn from it (or, perhaps undermine it).
Capability trees are extremely important complements to planning exercise - if you are not honest about what it is you are able to do routinely and well, your plans will be decoupled from reality in a particularly nasty way; you will have objectives but no means to reach them! We have discussed this in earlier notes, but it is worth mentioning again: investing in, and training to enhance, capabilities is something far too few companies do well today.
The root of it all
Causal trees are another kind of model that is worthwhile exploring. In scenario analysis one of the key questions is usually to try to build out the set of ”driving forces” that we want to understand more in detail. One of the best ways of discovering this is through causal mapping - building backward from the phenomenon we want to understand and then map out the causes of the causes of the causes of this phenomenon.
This also allows us to find interesting points of intervention and branches of the tree that we can change and those we are unlikely to impact with any of the options that we can generate.
Causal trees also allow us to decompose prediction problems into a set of factors that will determine a decision or an event and then look more closely at them - a key technique know as fermization and helpful for anyone who wants to emulate Tetlock’s Superforecaster methodology.
Causal trees are different from evolutionary or genealogical trees in that the causes are not just historical, they can be freed from a time perspective and just be seen as the forces that hold the structure together.
They are, in a sense, an exploration of the roots of a phenomenon.
Real trees and real lessons
But trees are also useful to understand more complex problems - like hypergrowth in an organization. Now, most organizations grow vertically in functions, rather than horizontally across functions. The result, if we frame it as a tree problem is that the branches are growing and becoming heavier, but the trunk is not thickening at a rate that allows the tree to be stable and succeed.
Consciously investing in the trunk - building cross-functional groups and mechanisms - then becomes absolutely essential.
Trees are amazing organisms, and they rely on a specific relationship between branches and the trunk - and not just for simple stability, but also for efficient flow of energy within the tree as a structure. The some of the thickness of the branches is the thickness of the trunk, and so as the branches grow, they will be limited by the earlier branches that they rely on for energy sources. This means that the tree structure depends on, as Geoffrey West has suggested, area preserving branching. A fact that Leonardo da Vinci seems to have been among the first to realize.
This allows energy to circulate efficiently and optimizes the tree structure. If any branch would grow beyond that it might not just break off, but could also severely harm the energy circulation systems in the tree as whole. It is not hard to see the analogy to information circulation in an organization when one part of the organization becomes too dominant at the ”leaves”-level and so consumes more energy and information and creates an imbalance in the organization.
The trunk has to grow before the company can expand into too many different branches.
Now, interestingly, this same structure that seems to maximize through-put also seems to be a structure that allows trees to better withstand heavy winds - thus creating a much more stable structure! This extra benefit suggests that organizations that manage - by analogy - to recreate are-preserving branching as they grow could even be more resilient to external stresses!
Tree time
Finally, we would be remiss not to discuss the age of trees. Trees live for a very long time, and some of the oldest organisms on Earth are trees. This is another reason to really think about trees and explore the mental models related to them - we can think in ”tree time” and use that to suggest that there are changes around us that matter more and matter less — trees can be catastrophically ended (cut down), but they survive many other minor events and challenges in their environment.
A tree organizes itself to care about /erosion/ and captures carbon - as well as creates an amazing biotope for other creates to live in, and organized in forests they manage to co-exist with other trees in ways that are mutually supportive.
If we define the tree as business model it is a long term, plus sum and mutually beneficial business model - and it seems to work perfectly well. Trees are successful evolutionary adaptations and their age is testament to the sustainability of their business model!
Ok, so this may be a bit far out - admittedly it is not easy to ask all companies to be ”tree business models”. But there is something in the tree business model-idea that deserves to be fleshed out, and it is the combination of symbiosis and long termism that trees seem expert at.
It is intriguing that some of the oldest trees in the world exist in California, just as some of the fastest business models in world does — one cannot help but think that there is something to be learnt there.
In his masterful The Hidden Life of Trees, Peter Wohlleben suggests that a deeper study of trees is indeed worthwhile, not least because they communicate and collaborate a lot more than we usually suspect:
One reason that many of us fail to understand trees is that they live on a different time scale than us. One of the oldest trees on Earth, a spruce in Sweden, is more than 9,500 years old. That’s 115 times longer than the average human lifetime. Creatures with such a luxury of time on their hands can afford to take things at a leisurely pace. The electrical impulses that pass through the roots of trees, for example, move at the slow rate of one third of an inch per second. But why, you might ask, do trees pass electrical impulses through their tissues at all? The answer is that trees need to communicate, and electrical impulses are just one of their many means of communication. Trees also use the senses of smell and taste for communication. If a giraffe starts eating an African acacia, the tree releases a chemical into the air that signals that a threat is at hand. As the chemical drifts through the air and reaches other trees, they “smell” it and are warned of the danger. Even before the giraffe reaches them, they begin producing toxic chemicals. Insect pests are dealt with slightly differently. The saliva of leaf-eating insects can be “tasted” by the leaf being eaten. In response, the tree sends out a chemical signal that attracts predators that feed on that particular leaf-eating insect. Life in the slow lane is clearly not always dull.
Wohlleben, Peter. The Hidden Life of Trees: The International Bestseller (p. 7). HarperCollins Publishers. Kindle Edition.
So what?
Thinking in trees is about structuring decisions, exploring genealogies and understanding resilience. It can help us think about ecosystem collaboration and plus sum games in ways that much of the usual evolutionary analogies seem to miss.
Where the tech industry once blustered that old industries were dinosaurs soon to be extinct, or that software will eat everything, we may want to explore alternative narratives around trees, slowly communicating and adapting in a close knit ecosystem run in treetime where mutual dependence is more important than zero sum competition.
California has a lot to teach us about speed, technology and trees.
Thanks for reading, and hope to hear from you - as always send any thoughts or questions my way!
Nicklas