Unpredictable Patterns #9: Styles of Knowing
Axioms, experiments, analogies, taxonomies, probabilities and genealogies - and why you should be enjoy being a dilettante using them all for horizontal depth.
It has been a really interesting week. At work I have been fully immersed in thinking through the risk/fitness landscapes of technology issues and what capabilities we should build to navigate that landscape (as well as finally getting into the day-to-day of our issues, which is where the real learning is to be found), and outside of work I had the pleasure to give a video remote talk about ways of valuing AI, at Carnegie Bank [1] as well as finish an essay that will published next week on how the pandemic has revealed that Ferdinand Tönnies Gemeinschaft is alive and well underneath a thin veneer of Gesellschaft.
Dear Reader,
Spring is here, and even though the shadow of the pandemic lies heavy over all of us, the sun reminds at least me that there is a time after. That we will return to a life we now remember in alarmingly vague terms - with public gatherings, concerts, restaurant visits and travel, or, well - not return perhaps as much as rebuild it.
A lot will have been destroyed by the pandemic and our response to it and we never really return (that is a nostalgic notion) but we will rather have to reconstruct the world anew. There is a lot to be said about the way that we will do that - if we will launch in the an exuberant re-enactment of the 20s or if we will re-build in a way that is more cautious and restrained. Will handshakes come back? What about the European habit of cheek kissing? Some argue that it is a Christian habit that can be traced back to St Paul and there are even studies on exactly how you are supposed to do it. As a stiff Swede I have always messed this up, and find some comfort in the fact that there seems to be as little European consensus about how to deliver the air kiss as about any other topic. Seems Parisians prefer two kisses, those in Provence three and in Italy they start on the left! Joking aside - a habit like this can really disappear, and if it does we should think about what it means and how it will change our relationships overall. Handshakes, air kissing, hugs - physical closeness overall creates a social texture that permeates everything around us. Touch is profound and essential to human beings and one of the silent tragedies of the pandemic is the fading of human touch.
Things change in complex causal networks - as we lose touch - literally - we also lose other social cues and behaviors, and our entire way of greeting each-other changes. How does this affect the way we then converse? Will we become more distant from each-other, will the distance to the Other expand in a more philosophical sense? If it does we know that the Other is fundamental in the way we view ourselves, so we become more distant also to ourselves - the ripple effects can be surprisingly powerful, and it may be worth some time thinking about how we reconstruct the social responsorium, to borrow a term from Swedish sociologist Johan Asplund, as the pandemic lifts.
Styles of Knowing
Now, onto this week’s theme. What I have tried to do in these notes in to find different ways of thinking and then applying them to issues related to technology and policy, and in this note we will look at what historians of Science sometimes call ”Styles of knowing”. The idea that there are different scientific styles is not new and exists in a number of different versions, but the one we will look at here is Alistair Crombie’s. Crombie is, if not one of the founders of the modern version of history of science, at least a pioneer in it, and in his work he tried to catalogue the different scientific styles of thinking that we can engage in. They are:
Mathematical postulation and proof, the axiomatic approach.
Experiment, and the testing of hypotheses.
Modelling or analogical thinking.
Taxonomy (the method of classification into natural kinds), arguably the foundation of many sciences.
Statistical or probabilistic thinking, often traced back to medicine.
Historical - genealogical methods, including evolutionary explanations.
Crombie views these different styles of knowing as the ways in which we can find out about what is going in in the world. Now, this list is helpful to anyone who wants to think through a problem, since it suggests six different ways of thinking about anything. I would almost recommend that knowledge workers print the list and put it next to their work space, as to ensure that they look at the issue they are working on through all of the different perspectives offered by these styles - and suggest that doing so routinely is likely to yield important insights.
If we look a bit more closely at the different styles we find important nuances, and it would of course take diligent study and practice to master any of them. But mastering is not always necessary, sometimes the dilettant application of ideas is as helpful as, or even more helpful than, mastery since you end up moving between perspectives and getting horizontal depth rather than vertical depth in your analysis.
Let’s look closer at what applying them means, with a few scattered examples around tech policy and politics.
Axiomatics and geometrical thinking
The axiomatic style of thinking at first seems very remote from what we do in policy, but in fact is very useful in any analysis that looks to what the law says. Law is a kind of axiomatic system, where conclusions are reached by logical manipulation of basic concepts. The basic concepts then decide what outcomes you can get and so by clarifying the axiomatic structure of a piece of legislation we can see both weaknesses in it and alternatives to it.
Take privacy law. Our current data protection regimes are based on an axiomatic structure that starts from the concept of data as fundamental, and then from that concept we build data subjects, processors, controllers etc and we add the notion of data processing. That the world is made of data is as axiomatic to privacy law as points and lines are to Euclid.
But just as we can construct non-Euclidean geometries by changing basic axioms (the one about parallel lines) we can construct non-data privacy regimes by starting from other concepts. It is, for example, likely that you can develop a consistent, coherent and more sustainable privacy regime by starting from the concept of identity and deriving concepts of autonomy, confidentiality and harm.
To think of people as data or think of them as having identities is a choice that is far from necessary or obvious.
In praise of experiments
If we move on to the experimental style of knowing it seems increasingly obvious that experiments are key to any organization that wants to grow or learn. Hal Varian, chief economist at Google, has argued that the experimental infrastructure a company commands is the key competitive advantage that it has - and surprisingly many companies have no experimental infrastructure at all! The conscious construction of that infrastructure - of a laboratorium - should be a key priority for any CEO.
But the idea of experimentation is not limited to products or services. It should also be applied to areas like public policy. A policy team should have its own experimental infrastructure, test different debating techniques, takes on issues and different forms of outreach. And this is easier than you think - let’s look at two really simple examples of experiments in policy work:
First, paper vs email. In this experiment you first send an email to a list of network contacts offering to brief them on a subject and then you send a letter - a physical letter, signed - to them and offer the same thing. Which one has the most take up? I remember doing this in one of the countries I was working with, and the result was stunning. The response rate to the letter was in the high 40 percent, and no-one responded to the email.
Second, developing a product to respond to a legitimate concern. The concern around large companies collaborating in opaque ways with government led companies to try to devise a product response: the transparency report. In these reports companies would list the number of government requests per country and category and bind themselves to share with users how their governments used the legal powers they had to make these requests. The product was an experiment and was developed experimentally, but today these transparency reports are close to - or should at least be - industry standards. (And, they should be embedded in institutions, but that is another story).
Experiments require organizational curiosity and that is not always easy to protect. Organizations that look to minimize errors rather than generate insights will not be willing to experiment, and they will, ultimately, fail. An interesting exercise here can be left for you, the reader: how would you grade the organizational curiosity of the organization you work at between 1-10?
Analogies and models
The analogical style of knowing is perhaps the most powerful form of thinking we can engage in, and developing models of problems is absolutely essential to be able to discuss them. We rarely take them to articulate our models and that often leads to a situation in which everyone around the table has a slightly different model of reality and so optimizes for different things. A clear, explicit and shared model of reality is a pre-requisite for any important work.
Now, models are abstract analogies, where you build something that can be used to study a problem you are interested in. Analogies are somewhat more concrete and often really useful for starting to think about a problem.
The analogical question is this: how is A like B?
It seems almost ridiculously trivial, but when we start using it we find that it is surprisingly good at two things: to show us an aspect of a problem we have missed and to highlight mechanisms that we may have neglected to study.
Here I have a recommendation. I think you should have a stock set of generative analogies that you apply to any problem that you are interested in, and by this I mean that you should pick a few especially powerful analogies and make it a habit to apply them to your thinking.
How do we pick or choose these stock analogies? One way is to read classical literature, like Shakespeare, Homer or Plato. Look for the analogies that he returns to, and list them - they are likely to be deeply rooted in the human mind, not least through these authors using them! My short list of stock analogies are the following - but you will find others:
How is X like
…a game.
…a piece of music.
…a play.
…the weather.
…a machine.
…an ecosystem.
…a duckrabbit.
…an immune system.
Analogical thinking is then developed into models by increasing the resolution of the analogies. There are a few intermediate steps that are powerful and should be studied, I think, and an interesting variation is: What is to A as X is to Y?
The application of this to technology and policy is enormously broad, of course, but using different analogies to understand the overall tech lash and then seeking a finer model of it is an obvious example. How is the tech lash like a duckrabbit? Like an immune system?
The more weird an analogy looks, the better! If it feels forced, force it - push through and see what you end up with. The challenge is not to find an easy analogy, but to push until you find a useful one.
What kind of things are there?
Taxonomy is underrated as a style of knowing. Just figuring out the taxonomy of a particular field of study is key to starting more serious exploration of that field. What kinds of things are there? How can they be grouped and by what principles? The apparent simplicity here deceives us, and makes us think that taxonomies are arbitrary, but they rarely are.
The brilliant Jose Luis Borges plays with the idea of the taxonomy in a beautiful way when he confronts us with the taxonomy of animals he claims is to be found in a certain Chinese encyclopedia:
those that belong to the Emperor,
embalmed ones,
those that are trained,
suckling pigs,
mermaids,
fabulous ones,
stray dogs,
those included in the present classification,
those that tremble as if they were mad,
innumerable ones,
those drawn with a very fine camelhair brush,
others,
those that have just broken a flower vase,
those that from a long way off look like flies.
And the thing is that not even this taxonomy is arbitrary, it exhibits a non-pattern that makes us laugh. Indeed, Michel Foucault credits this taxonomy with making him laugh and then write The Order of Things about how taxonomies structure all of human knowledge and the societies we build upon it - so taxonomies can have real impact!
We rarely take the time to think taxonomies through and too often apply coarse-grained categories ill-thought through to even our most important problems. But asking what kinds of things there are is surprisingly powerful.
In politics we should think hard about what kinds of advisors, advocates and detractors we face and how we deal with them - to take a very simple example.
Now, for the two last categories I will just say nothing here. The idea of genealogies and history has been dealt with in a previous note, and I do think it is absolutely essential. The statistical style of knowing is the one we apply most in policy - in surveys and polls and similar devices - so little needs to be said about that here, but the perhaps obvious point that what we ask here should be informed by the other styles of knowing! Here taxonomy can help us - especially in sorting out what the baseline probabilities are of the things we are studying.
So what?
What can you do with this? How does Crombie’s styles of scientific reasoning help us in our day-to-day work? I think there are a few take aways that we can look at, but first I would argue that the perhaps most important conclusion is that we should ensure that we can shift between the different styles of knowing, and when we do we can concentrate on the following:
Know the axioms of any problem field you discuss, and understand the logic it operates by.
Develop, sustain and use an experimental infrastructure in your organization and business. Foster a laboratory mindset.
Ensure you have a shared model of reality by making it explicit, and use analogies to generate new perspectives.
Know what kinds of things there are in any problem you are trying to solve.
Learn the baseline probabilities of events you are interested in.
Understand the story of your problem, and its genealogy.
So, the so-what here is about developing thinking habits that help you work through problems methodically and in a diversified way.
On the blog this week
This week the a few things that may interest you on the blog:
A note on limiting factors as a useful mental model.
A tip about an episode of the Mindscape podcasts about the classics. Very recommended!
Thank you for reading and keep the comments coming. I would love feedback and questions and suggestions for other topics! If you know someone else who might want to read these notes drop me an email in reply to this and I will add them.
Take care!
Nicklas
—-
[1] I give talks when the inviting organisation agrees to pay my speaker’s fee to the Children’s Cancer Fund. If you can contribute to them do so here https://www.barncancerfonden.se/ —
Wonderfully inspiring. I will try to use your advise in my next project.