Unpredictable Patterns #92: On surprises
On the unsurprised life not worth living, expectations, plot twists and the limits of surprise
Dear reader,
You may be surprised to get this on a Monday. If so, good! This week’s note is about the value of being surprised in small and great ways - and how we can work with surprises to understand the world better and learn faster. But in the interest of honesty I should admit that sending the note out today was not just due to the subject of the note - it was also because I have been travelling in Sweden, enjoying the beautiful autumn weather and some time off with friends and family. That said - I hope you enjoy the note!
On surprises
How often are you surprised? There seem to be two conflicting intuitions about how to answer this question: the first is that if we know a lot and are observant, then we should have adjusted our expectations continuously and not need to be surprised very often. The second, however, suggests that only the very keen mind is surprised often, because it notice nuances and differences that are meaningful all the time in our complex and changing environment - and so surprise is a sign of intellectual curiosity.
I sometimes get the sense that many people I ask this question tend to answer it in a way that they hope conveys that they are on top of things, and not often thrown off course. The exception is the person who confesses to constantly being surprised, and forced to adjust their expectations in different ways - but seems to enjoy it as a great learning experience.
Partly this is in the nature of the concept itself: the ratio of surprises to expectations should be low (a surprising event is by definition somewhat infrequent, since we calibrate our expectations on the frequency with which things happen - someone who is surprised by the sun rising every single morning has a problem with their expectation calibration), but partly I increasingly believe that the inability to be surprised is a severe cognitive handicap.
The unsurprised person has often created a set of expectations that are so coarse-grained that they cannot see meaningful differences anymore. This is almost like the cognitive equivalent of squinting hard so all things so you see seem to have the same shape, and so declaring that you see nothing that is out of the ordinary - and it is easy to end up being that person, especially if you have significant experience in a field - because your experiences structure your expectations and with enough experience you can quickly calibrate your experiences to avoid detecting surprises.
You may also confuse the ignorance that leads to surprise with the ignorance that comes from arrogance or stupidity - and that is a fatal mistake. There are many different kinds of ignorance and it is quite possible to be ignorant of something surprising at the same time as you are a deep expert in a subject.1
There is even an argument that this is a rational thing to do: surprises consume quite a lot of cognitive energy. As we noted in an earlier note, surprise forces us - among other things - to reorder what we think we know, and this is hard work that consumes a lot of our cognitive capacities, so it makes sense that we would have evolved to minimise it. But if we do not check that tendency (just like should not eat sugar in excess just because our brains tell us to), we end up in a state of learned jadedness.
Nothing surprises us anymore, because we have seen it all.
But we never have: our expectations always need adjusting and they should constantly be shifting.2
There is another version of the unsurprised mind, that is tricker to get right, and that is when we have no expectations at all, because things have become so random that it makes no sense to form any predictions. This state is one in which we cannot be surprised, since we know too little and it changes too fast.
We need to be able to form expectations to be surprised.
What this suggests is that the way we curate our expectations is key to develop the ability to be properly and productively surprised. A clear sense of what we would expect sets us up for surprises rich in learning.
A map of surprises
We can reverse-engineer this observation into an interesting tool: a list of possible surprises. If we know what we expect, we also know what would be surprising to us. Writing out and exploring what you think about a project, an issue or the organisation you work in - and then conversely what would surprise you - gives you a great learning device: a map of the space of possible surprises that lie outside of your expectations.
Now, this space of possible surprises is interesting to explore more in detail as well, since there are different kinds of surprises.
Let’s take a toy example: you are trying to predict the advances in artificial intelligence circa 2010. You look at the hardest problems that the field is engaged in, and you look at games. You rank different games by complexity and so start to set out expectations about when a computer might be able to beat a master at a certain game. Maybe you form your expectations in terms of how many years out the event is for games like poker, go and soccer. Say you have assessed the likely time to be, respectively 20, 30 and 100 years. This gives you a chart of different surprises that are well-defined and based on the best effort expectations that you have been able to form.
If artificial intelligence succeeds beating human players before this, you will be surprised, but in a way that sets you up to learn from it in the best possible way: these are structured surprises, they come pre-loaded with meaning and consequences that you can have mapped out beforehand. Now, a list of surprises like this also allows you to start charting the overall frequency of surprises in a field - an intriguing measure that gives you some sense not just of the pace of development in the field in question, but also the pace of shifting expectations.
AI is a good example of a field where the pace of development is high, but also of a field where the frequency of surprises has become quite significant over the last decades. The combination of a high recognised pace of change with a high frequency of surprises is indicative of a field that may have very large social impact over time.
What the toy example also gives us is an example of is a capability surprise - a violation of the expectations you have of what someone or something will be able to do at some point in time. Such surprises are of special importance in technology and war - and are explicitly part of the thinking around defence research institutions like DARPA. In fact, DARPA was more or less set up to both create and avoid strategic, technological surprise - which means that the agency needed to start from both a detailed set of expectations and then a map of where the possible surprises could be - and then defence research becomes a question of funding surprises!3
Surprise technological capabilities can fundamentally shift the power balance in a war and so being able to both eliminate surprise by forming accurate expectations, and creating surprise by research funding is key for the military, but it is not the only use of surprise in war.
The other form - the surprise attack - is much more complex and an example of an event surprise. There is a long-standing debate in military science about the value of surprise attack, and if it can create a sustainable advantage or not. What seems to be the consensus is that surprise attack only works if it is then followed up by overwhelming force that ends the war - but if you launch a surprise attack without being able to follow up, you just galvanised the will of your enemy.4
Some writers even argue that US political and military strategy was repeatedly strengthened and articulated by surprise attacks: examples include Pearl Harbor and 9/11, and they both ended with the defeat of the surprise attacker - highlighting that a single attack rarely wins a war, unless it is based on a new, surprising, capability. This kind of surprise, purely tactical surprise, needs to be integrated into a strategic set of surprises to work.5
In war, or any confrontational game, the key to surprise is an in-depth understanding of what the opposing side expects. This practice, of modelling your opponents expectations, is curiously rare in strategy work - since we often end up being quite organisationally self-centric in planning the future. But undertaking a robust study of the expectations of other actors in the field is a great exercise, since modelling those expectations allows you to - yes, craft surprise where you want to.
As noted earlier, surprise is useful especially in external engagement, since this is where we want to change minds. We change our minds when we are surprised, and forced to adjust our expectations, and so if you are interested in influence work you should be interested in how you can craft surprises that open up opinions and allow people to change their minds - and this requires figuring out your audience’s expectations.
This is true for both communications, marketing and policy. A surprising message can shift the swing vote in a meaningful way, especially if it is combined with consistent actions that follow up on the surprise. A company that comes out supporting a piece of legislation it was thought to oppose, forces the swing audiences to adjust their expectations, even if the detractors will work hard to integrate this into the existing expectations. In a way this is a nice description of one dynamic in public relations work: you create surprises that your opponents try to neutralise by forcing them into the expectations they want to maintain (‘they are only doing this because they know they will lose this battle’).
The game of surprise and adjusted expectations is fundamental in our social worlds, and a map of the possible surprises (and expectations) will let us play that game in a much more efficient way - and both avoid and create strategic surprise.
Learning through surprises
We can use surprises to learn efficiently as well. The trick here is to look for disconfirming data and evidence that can help us understand the world better. Imagine that you are about to invest in a company, and you know one of the other investors - would it make sense to ask them about the company and if it is a good idea to invest? At first blush it seems a horrible idea, since they are incentivised to drag you in if they can - but consider the case where they say no, and that they have lost faith in the company. Such a surprising reply would give you plenty of information, and allow you to make a much better decision. In this case you will discard an affirmative answer to the question of if you should invest, but you ask the question anyway because of the off chance that they may actually reveal something surprising about the company.
Another version of this is to design your own reading for surprises. Write down what you think you know about a subject before you read a book about it - and try to state strong opinions. Now, when you read you will be able to read for surprises: ideas or evidence that contradict what you thought you knew. This allows you to read at a different pace as well — if there are no surprises in your reading you can speed up, but if you run into views or facts that contradict your own you should slow down.
This is hard, because we do not like being wrong - and a part of being surprised is realising that our expectations were off, and so, in some sense we were wrong. But it is also enormously valuable: if we read this way we will consciously read to change our minds, and every book that we get through will have adjusted our expectations. This beats reading a book without expectations, because when you do you will subconsciously adjust to whatever is being said, and convince yourself that this was what you thought anyway. When you do the pre-work and state your expectations, you will have to change your expectations and so learn more than if you just read.
Another technique you can use when learning something is to ask questions that consciously are designed to disconfirm what you know. If you think that X is the case, you ask if the opposite is not true - trying to avoid the comfortable lull that we often settle into, confirming each-others misapprehensions.
We should always be looking to maximise the information we can get out of an interaction in some way. Information and surprise are connected: one way to understand Claude Shannon’s information theory is that the measure of information in a message is the level of surprise it creates: if you are not surprised at all there was no new information, but if the message blew you away and forced you to reassess everything you thought you knew, well, then it was rich with information.6
The number of surprises you have in a day is a great measure of how much new information you have been able to gather and process, and so if we are interested in learning we can tweak Socrates admonition, and say: the unsurprised life is not worth living.
Plot twists
We order the world in narratives, and think in stories - and this also means that the purest surprises are the ones that we encounter in trying to predict and organise our narratives. We recognise this easily in literature, since we all love the plot twist that we encounter when we read a detective story or a horror tale. The plot twist is a special form of surprise, a surprise that forces is to restructure the entire narrative understanding that we have built out during the telling of the story.7
Plot twists also teach us something interesting about the limits of surprise. When we start to craft our own stories, we use blunt and clumsy plot twists, like “and then he awoke and realised it had all been a dream!”. This is a surprise, but it does not force a new understanding of the story as much as a rejection of the story. We grow irritated with plot twists that are too simple, that resemble a “deus ex machina” - a god from the machine, that turns up and sorts everything at the end of the story.
Why is that? One reason is that a good plot twist is built on information that is already given in the narrative - it needs to be in place for the new pattern to fit clearly and for us to understand that we could have interpreted the signs all along the way if we wanted to. This creates a very special kind of unnerving surprise that suggests that our entire narrative pattern recognition skills maybe compromised in some way.
Examples include the end of The Usual Suspects and The Sixth Sense, where the information about the twists are embedded in the narrative, available to us if we had just read with a little more suspicion. Such plot twists are powerful because they build on the realisation that we could have formed other expectations than those that we had - and that failing to do so set us up for a massive re-interpretation of the narrative we were engaged in.
The dream-ending simply feels like cheating, since the twist did not follow from the narrative at all, but was just bolted on at the end.
Studying narrative surprise is also important to understand how we really predict the future. The first thing we notice is that we do not predict events, but the continued story - and this is why the idea that rationality is about evaluating propositions for probability or predicting distinct events is so broken. The world is not propositional or just a set of events - it is a narrative, and surprise, as well as information and expectations, are embedded in stories.
Now, plot twists are interesting in themselves, but they also provide us with some helpful hints on how to work in public relations work of different kinds. If you do want to change the interpretation of a set of events, your new interpretation needs to be possible to overlay the dominant narrative of the moment. You will never win with a message that is bolted on at the end like the dream plot twist, your new message needs to build on the information in the narrative and allow for an alternative understanding of the very same events that led to the dominant view.
Great public relations work reminds us of plot twists in that it allows for a different interpretation of the very same story that led us to the view we want to challenge.
The plot twist also gives us another failure to study, however, and that is the conspiracy theory. A conspiracy theory is often designed to be surprising, but it over-integrates all the element in the narrative in its twist. A great plot twist settles on a set of narrative elements and shifts the meaning of them - the conspiracy theory seeks to change the meaning of not just the narrative, but all narratives. It is an attempt to design a surprise that is so great that it shifts your entire understanding of the world.
For most of us that fails, because we have narrative horizons that do not allow for a single story world. The world is made of many narratives, and they are not just disconnected but even contradictory - and that is how the world is. The idea that the entire would could be subsumed in a single narrative, with a great plot twist, is an example of a religious kind of thinking. We see it in the idea that everything is the will of God and that even cruel events can be justified in some way because it was all for the better. But the same global plot twist is present in more current ideas, like the simulation hypothesis - the idea that we are all living in a computer simulation.
These plot twists fail for most of us because they are too grand, too all-encompassing and they attract only those that hope that there is a single narrative behind the messiness of the world. This need for a single narrative and a final surprise is the essence of the religious impulse - a will to understand everything as a single story. But for this to work you need a surprise that is too big, to disconnected from the world - and we are back in the equivalent of the child’s ending, where everyone wakes up and the story was just a dream.
Maybe that is a comforting thought.
So what?
Surprises matter, and they are powerful tools for individual learning as well as for public relations. They matter in military strategy as well as in literature, and a deeper understanding of surprises is key to understanding how we change people’s minds. What, then, does this mean practically? Here are a few things.
First, we should expect that the ability to be surprised and to build surprise into our self-development is going to become more important in a world that is becoming ever better at predicting patterns and narratives. To be surprised may be a key skill, and something that we will start to look for in recruiting and skill assessments. The person who is never surprised is likely to be replaced with a machine.
Second, we should expect that we will get some kind of formalisation of surprise in machine learning and artificial intelligence. The ability to be surprised and change our expectations at different levels of granularity seems to be key to human level intelligence.
Third, politicians and companies that master the art of the plot twist are going to do really well in public relations and influencing, finding the right level of plot twist and the right size surprise to convince their audiences that their interpretation of the elements of our shared narrative is the salient one.
Fourth, the frequency of surprises in a given field will indicate how fast it is moving and where there are strategic advantages to be had. We should expect military research to double down on the idea of creating and preventing strategic surprise.
Finally, then, we should actively seek out surprise in our everyday world, and carefully avoid the jaded stance that finds nothing surprising. Sometimes we mistake this for wisdom, when it is really nothing else than mental stagnation.
Thanks for reading,
Nicklas
See eg. “Ignorance and surprise belong together. Surprises can make people aware of their own ignorance. A surprise is normally rendered surprising when it occurs unexpectedly and also runs counter to accepted knowledge. A surprise thus cannot be fully understood independently of a person’s or a group’s ignorance. In fact, novel things always include elements of surprise, uncertainty, and the unknown, all of which are located outside the sphere of prediction. Consequently, surprising events in research are often something to which scientists aspire in their activities since it means a window to new and unexpected knowledge. Scientific methods thus should allow researchers to surprise themselves as well as their peers."in Gross, M Ignorance and surprise : science, society, and ecological design.
Surprises build on expectations, e.g. “There can be no surprise without expectations. Fortunately or un-fortunately, human cognition runs on a steady diet of them. As we use language, interact with others, and navigate the physical world, we continually extrapolate from incomplete information. We depend on prior experience to fill in our understanding of what we’re seeing, hearing, encountering, and doing. We rely on our expectations of how physical objects should behave as we move our bodies through space, avoid obstacles, and manipulate objects. Even our most apparently unmediated perceptual experiences are strongly influenced by prior expectations about what is possible or likely in a given sensory environment (Summerfield and Egner 2009)." in Tobin, V. (2018) Elements of Surprise. Harvard University Press. Available at: https://www.perlego.com/book/3119900/elements-of-surprise-pdf (Accessed: 7 October 2022).
Their mission as laid out on their home page explains this: “The genesis of that mission and of DARPA itself dates to the launch of Sputnik in 1957, and a commitment by the United States that, from that time forward, it would be the initiator and not the victim of strategic technological surprises. Working with innovators inside and outside of government, DARPA has repeatedly delivered on that mission, transforming revolutionary concepts and even seeming impossibilities into practical capabilities. The ultimate results have included not only game-changing military capabilities such as precision weapons and stealth technology, but also such icons of modern civilian society such as the Internet, automated voice recognition and language translation, and Global Positioning System receivers small enough to embed in myriad consumer devices.” The surprise of Sputnik led to a profound change in the US strategy. See https://www.darpa.mil/about-us/mission
See eg: “"From a military operational standpoint and for the purposes of this report, surprise is an event or capability that could affect the outcome of a mission or campaign for which preparations are not in place. By definition, it is not pos- sible to truly anticipate surprise. It is only possible to prevent it (in the sense of minimizing the number of possible surprises by appropriate planning), to create systems that are resilient to an adversary’s unexpected actions, or to rapidly and effectively respond when surprised.."
Responding to Capability Surprise: A Strategy for U.S. Naval Forces. (2013) The National Academies Press. Available at: https://www.perlego.com/book/531342/responding-to-capability-surprise-a-strategy-for-us-naval-forces-pdf (Accessed: 7 October 2022)"
See especially Gaddis, J. L. (2005) Surprise, Security, and the American Experience. Harvard University Press. Available at: https://www.perlego.com/book/2634070/surprise-security-and-the-american-experience-pdf (Accessed: 10 October 2022).
See eg "“He had this great intuition that information is maximized when you’re most surprised about learning about something,” said Tara Javidi, an information theorist at the University of California, San Diego." "The logarithmic formula for Shannon entropy belies the simplicity of what it captures — because another way to think about Shannon entropy is as the number of yes-or-no questions needed, on average, to ascertain the content of a message." https://www.quantamagazine.org/how-claude-shannons-concept-of-entropy-quantifies-information-20220906/
This is nicely explored in aforementioned Tobin, V. (2018) Elements of Surprise. Harvard University Press. Available at: https://www.perlego.com/book/3119900/elements-of-surprise-pdf (Accessed: 10 October 2022).