Talking about numbers – evaluating pandemic responses (In Swedish)

In the beginning of the pandemic I wrote a blogpost in Swedish, inspired by Tyler Cowen, arguing that we owe it to ourselves to predict a number of fatalities that we believe would within what we expect to reach – because without this we just recede into a position of referenceless criticism. The key I developed then to assessing the Swedish pandemic response had two dimensions – an absolut one where I set success at under 20000 deaths in a year, and a relative one where I set success to not more than 15% more deaths than our country neighbors. On the first dimension, then, the Swedish strategy was a success, and on the other an abysmal failure. In this follow up note I try to explore why that can be, and how to think about it as I keep struggling with how to evaluate the public response to the pandemic – on the 31st of December 2021.

I increasingly think that we need to look at a multifactor analysis that contains both preparedness measures, disease control measures and disease treatment measures to evaluate where we end up here. And it is far to easy to just speak about a pandemic strategy, as if that was a cohesive whole. The whole note here in Swedish.

10 questions for thinking in games (Mental Models VII)

It is fair to say that playing video games have a number of positive cognitive and attentional effects (see this metastudy of 116 different papers), but one thing that is rarely highlighted is the fact that video games in some cases offer mental models that can be applied cross domains and used to think through complex scenarios and problems. It seems almost frivolous to suggest that we “gamify” or engage in “game storming” when it comes to really complex problems like climate change, but the reality is that a game is nothing else than a playable model. There is, I think, some argument for differentiating out a playable model from a simulation if only because the idea of a playable model puts agency at the forefront of what we are doing – we “run a simulation” but “play a game”.

Seeing in games is accepting that meaning is created not given.

And it is not only video games that can be drawn upon to understand and structure complexity better – I firmly believe that the people who grew up playing role playing games like Rolemaster, D&D or Call of Cthulhu have at their fingertips ways of thinking through a problem that includes core parameters like luck, skill, stats needed to accomplish something, the mechanics of conflict or combat as well as the importance of narrative to all of our endeavours.

So what are some interesting ways in which we can think through our problems with the help of game design? Let’s look at 10 really simple questions we can use to structure a challenge in game terms.

  1. How do I keep score? This is a deceptively simple question that is often overlooked, especially in large organisations. Ask your colleagues how they keep score overall and you will find, especially cross-functionally, that different teams keep scores in very different ways.
  2. What is the skill tree here? I have already discussed this in another post, but skill trees are amazingly powerful devices to structure a discussion about the capabilities of your organisation and think through how you need to resource yourself to ensure that you have the capabilities needed to reach your objectives and key results.
  3. Who are the other players? Again a deceptively simple question, but especially in large organisations you tend to lose sight of the fact that almost everything is an n-player game where n>3. If you know what other players are you can start thinking through what moves you would make in their situation.
  4. What character am I playing? This is offered slightly tongue-in-cheek, but it helps to think about who you are in your workplace — you have a professional identity, and that is not far from thinking through how you roll a character in a game. This also helps you to create the distance you need to your work to really be able to seek out and appreciate hard feedback, actually and reminds you that this is just a small part of who you are. Underneath all of our masks we are just ourselves, but we can use the masks to great effect.
  5. What is the quest? If you structure a problem like a quest you will see that there are key points in it where you need to meet someone, perform a hard task, face yourself – etc. And quests are great additions to the core projects that a company needs to run, and they are awesome tools for career development. A simple way to do this is to put up a quest board where folks in the company can post a quest with a short duration, clear outcome and a key problem to solve — and those who want to try their hand at working in another function or just dig in extra can do so.
  6. Resource management. In any sufficiently advanced strategy game you typically die not so much because of your lack of strategic genius as from the fact that you run out of iridium / money / dark mana / any other resources. Which are the resources you are managing and how do you ensure that they are managed right? This approach beats budgeting hands down, because budgeting is based on the amazingly arbitrary idea that a year is a good time to plan for and then breaking it up in quarters makes sense. It often doesn’t – but managing resources dynamically, from round to round, is key!
  7. Research trees. These are related to skill trees, but they are more open ended. What are the open research questions you are working on, where a new concept / technology / model / service / product would allow you to significantly change your game?
  8. Diplomacy. In most advanced 4X games you can win through building alliances and unifying different players. The ability to think creatively and explicitly about diplomacy is key. Who are the key people in the network you are playing in (it helps to think about all games as games on networks) and what is your relationship to them? For many big tech companies the loss of diplomatic relationships with publishers came with a surprisingly high price in the 4X Big Tech game.
  9. Character development and experience points. Are you deploying your learning effectively? Are you helping others develop and get experience? How do you think about levels in your organisation? Many company’s have talent ladders and levels that are hopelessly fuzzy and vague, it is much better to think through how you add experience points and when that means that you are the next level — again it feels frivolous, but is it really more frivolous than the more or less arbitrary performance evaluations we inflict on people in many modern organisations? Would you not prefer to discuss what the experience you have had should be worth in points towards a set goal?
  10. The map. This simple, fundamental tool of any strategy game is almost always missing in modern organisation – they have no real map of the environment they are in, no representation of the fitness landscape they are navigating – that makes very little sense, but again comes from the fear of playing, of being seen to be less serious. And there is something fundamental here – this fear of playing, the idea that play is something reserved for children and something you outgrow is really the mind killer. All art is play, all work is play if you do it right.

Now there is one thing that I think we should point out very clearly, and that is that the point of playing games is not to win. It is to keep playing. This has been brilliantly captured by James P Carse in his ingenious little book on finite and infinite games, and the definition of infinite games is also Carse’s:

Infinite players cannot say when their game began, nor do they care. They do not care for the reason that their game is not bounded by time. Indeed, the only purpose of the game is to prevent it from coming to an end, to keep everyone in play.

Carse, James P Finite and Infinite Games: A Vision of Life as Play and Possibility

We play to keep playing at the heart of things. We may want to win in individual finite games, but we need to see that they are all played in the context of the infinite game, the great game we call life.

What will the pandemic mean for the composition of our economy?

This paper from the Federal Reserve Bank of Atlanta is an interesting signal for determining what happens to the economy post-pandemic. It shows a sharp recovery in the starting of new businesses – and when compared to the Great recession this really is evidence of resilience – but! – and this may be a marker of something different happening here – that transition rate, the rate at which business transform from non-employer to employer businesses – seems to be dropping. The new companies are non-employer companies (i.e. self-employed in many cases). The graph in the paper hints at this:

So what can we say about this economy? We could form a number of hypotheses. The paper points out that we see a rise in employer startups, i.e. new companies that are formed with employees, but a slow down in the transition of existing business expanding. A way to interpret that is to say that companies that are surviving the pandemic are hesitant to move to expanding and employing people, even though more people start companies and there are more new employer start ups.

This may be fine, and could change as the situation is normalized. The question is what happens to these non-transition companies when there is a contraction of credit down the road, will they stay non-employer businesses? The impact on a coming credit contraction is discussed in another paper from the same bank where the prediction is that we are unlikely to see a “time bomb”-effect in the SME-space. Taken together, however, the papers suggest that there are alternative scenarios where the pandemic may slow down transition rates for a decade or more, leaving new employment not to transitioning business growing, but to new start up employers.

There is probably a lot more here to dig into, but these two papers suggest a few components of scenarios for the post-pandemic composition of the economy.

Can money buy you happiness? On comfort science.

In what has become the accepted wisdom, income does not increase your happiness over some level – usually at around 75000 USD / year. But in a recent paper on PNAS that claim was shown to be false…or was it? The debate that has been raging about the paper since it came out has been really interesting to follow and there are essentially two different positions that people take.

(i) Money does buy you happiness. More money = more happiness.

(ii) The old idea of a limit does hold, but with small, marginalen increments at very high prices

The second view is eloquently expressed in this tweet:

So, this is worth a think — not least because it challenges the received wisdom, and we should be very interested in where we have been casually wrong about something. I also think that this qualifies as a mesofact – something you learn once and rarely update – so let’s update our beliefs.

But what should the new belief be?

In this well-reasoned blogpost the author suggests that the new belief should essentially be this:

(iii) Money behaves like any good with decreasing marginal utility, and the decrease is steep after a certain breaking point. High income brackets, however, report less experience well-being even if they report higher life-satisfaction (with a caveat that the data is self-reported).

But is this right? Could we not challenge the entire premise of the experiment? What does it mean to ask people to self-report on their life satisfaction? Why is that judged to be the right question, or a question that will give meaningful answers? Have we not been led astray here by something rather sinister that happens again and again in popularized science – comfort science?

Comfort science, as I think about it, serves a very different purpose than proper science. It gives you support for your current life choices and your own personal narrative – and so you use it not as belief, but as comfort and support. It is like comfort food. Examples include things like:

  • It is healthy to drink alcohol in moderation (probably isn’t, but in the large scheme of things something will kill you and if you like to drink moderately – please do, but without justifying it through “science”).
  • Chocolate is good for you (without distinguishing between the 80% high quality chocolate and the Mars bar you eat with that justification).
  • Children need quality time not quantity time (and the variant that says that we spend more time with our children than any preceding generation – research that is essentially meaningless because it groups a diverse set of individuals – kids – together and makes assumptions about their collective well-being).
  • Just 12 minutes of exercise is enough per day! (It is better than nothing, but an hour long walk will not kill you, you know).

And, of course, the idea that income beyond levels that you can aspire to or just above simply do not make people happier. Let’s, just for the sake of argument, spell out all the methodological flaws in that argument from the outset.

First, Happiness is hardly a helpful measure – and not the only dimension in which we exist. Our age’s focus on happiness is a common psychosis that we have to get over. Back to Aristotle – you can judge no man happy until his life is at an end. Things like learning, relationships, failures all give meaning across a complex set of dimensions – and money is a part of that complex set of meanings, and may matter more for you if you use it to make a difference for others (imagine if the question had been not about the life satisfaction of the individual, but about the question of if they think they can effect meaningful change in the world – then we would connect money and agency, and what do you think the answer would be there?)

Second, people who self-select to respond to research about their income are signaling rather than reporting. They want you to know something about them. As in many surveys, the self-expression trumps the ambition to report valid data. When you ask people if they care about privacy, to take one example, you need to think about what it would signal for them to say that they do not — how uninteresting and worthless would they not have to think themselves to essentially say that they could imagine being fodder for a surveillance state machine or corporate monster? The same holds for highly socially charged things like wealth – across a number of different variables. Dogbert famously said that you should not trust the advice of rich people, since they do not want company. Think about what that does to the research design we are looking at here!

Third, income is in itself a weird focus. If we looked at the power that well invested wealth could give over time – or asked people to think about their life satisfaction when they are younger and directed the question to their future – how happy do you think you will be when you retire? – the answers, again, would probably be different. To look at income, to focus on yearly or monthly income is essentially like asking how happy are you now, when the real question may well be how happy do you think you will be later? Remember Kahneman’s tourist that wants to remember their trip but rather not make it!

And so on. Maybe the best way to update your belief, then, is this:

(iv) This study, like a whole host of other studies in the same style, is nothing but comfort science, and should be discarded for any more serious analysis of the underlying patterns it claims to describe.

Harsh? Yes. But definitely an option as you consider how you want to update your beliefs.

What kind of explainer are you? (Mental Models VI)

This article discusses a subject that increasingly has caught my interest: what is a good explanation? This is an old question in philosophy – and deciding that something is an explanation of some fact or phenomenon is not as straightforward as it seems. Explanations can operate at different levels and we may decide that different explanations are more or less relevant for different kinds of systems.

An example of this is the use of the concept of “stances” as applied by philosopher DC Dennett. Dennett suggests that any phenomenon or system can predicted – which is not the same as explained, but stay with me – from (at least) three different stances: the physical, design and intentional stance. The kind of predictions you apply will start from very different positions, and these positions are, in a sense explanations of the systems themselves. The physcial stance explains what we are studying as governed by physical laws and phenomena, the design stance looks to the purpose of the system’s design and the intentional stance explains a system as having intentions and emotions.

The way we approach predicting a system is related to how we explain it, albeit not in an entirely straightforward way – and Dennett’s stances are powerful mental models. They apply, among other things, to the analysis of computer systems. Someone who takes an algorithmic stance is doing something that is very close to someone taking the physical stance, whereas someone who takes an architectural stance is much more akin to someone who takes a design stance. The question of AI, then, is essentially a question of when we take an intentional stance to software – very crudely put.

DC Dennett has contributed to plenty to the question of how we understand, predict and – perhaps – explain systems.

When policy makers demand that tech companies explain what their systems do, it would be helpful if they could agree on what stance it is that they are taking – since algorithmic stances may very well be as useless as the physical stance is when explaining how a car works or why an ant hill behaves in a certain way.

Now, explanations are not just interesting because they allow us to predict systems in different ways. They are also interesting because we can optimize them in different ways.

In a recent paper, two researchers looked at what it is that we are valuing in explanations and found that there are (at least) two different dimensions along which we can evaluate an explanation. The parsimony of the explanation and the co-explanatory value that an explanation may have. The first dimension looks at how compressed the explanation is – what is the shortest program I can write that produces this system / pattern – but co-explanation looks at how many other things an explanation can order in one and the same explanatory pattern.

In an interesting application, the authors suggest that conspiracy theories are examples of over-optimizations on the co-explanatory axis, where we seek one underlying cause that explains everything that is happening. Conspiracy theories are attempts – as I have suggested elsewhere (in Swedish) – to compress the program you need to explain the world in a way that generates everything from a single set of initial conditions.

So what kind of explainer are you? Which stance do you prefer and do you seek parsimony or inclusive co-explanations? And more importantly – can you use these different mental models to shift between stances and value dimensions in explanations, and learn something new about what you study?

Exploring the Stagnation Hypothesis

The stagnation hypothesis, the idea that we have seen steady decline across science and technology over the last 50 years or so is increasingly gaining ground and becoming mainstream:

Thiel, along with economists such as Tyler Cowen (The Great Stagnation) and Robert Gordon (The Rise and Fall of American Growth), promotes a “stagnation hypothesis”: that there has been a significant slowdown in scientific, technological, and economic progress in recent decades—say, for a round number, since about 1970, or the last ~50 years.

Jason Crawford, Roots of Progress

This hypothesis is important for everything from how and if tech should be regulated to what kind of economy our grandchildren can expect to live through. But the hypothesis, as it stands, is hard to test and hard to find good metrics for. A generally sinking GDP (as referenced by Jason in the post linked above) is certainly an indicator, but is it enough to show that we have technological stagnation? Tyler Cowen and others have suggested that we may be living through a period of human stagnation (in my reading) and that the challenge is not so much that technology cannot progress anymore for some unknown reason, as that we have ceased to be hungry for the future. One way to phrase this is to say that we have shifted from viewing the world as uncertain to viewing it as manageable risk and this means that we do not value progress as much as we used to. The pandemic is a good reminder that this is a mistake, and the increasing uncertainty in our world (a consequence of, among other things, increasing complexity) should bring us even more pause.

So, maybe, then the stagnation we have to deal with is more of a spiritual stagnation where we have exchanged the spirit of progress for one of security? That may be controversial, but it is worth thinking about – since it may be just as hard to change the spirit of an age as to foster more technological innovation – and maybe we want to be doing both?

The other challenge with the hypothesis is that it measures progress overall linearly. Here we run into an interesting question of horizontal vs vertical innovation speeds. The best example is a coarse grained analogy with the Cambrian explosion, and it is easy to explain in a picture:

Measuring the pace of progress

So, if A in the figure is smaller than the sum of the B vectors, it seems that progress might actually be faster in the horizontal case than in the vertical case. That in turn seems to suggest that any general metric risks missing a hidden, silent Cambrian evolution – like connectivity becoming the norm in all artifacts or so.

Now, this is just a rough model, and more needs to be done here. But I think that the heart of the problem is that we may be misunderstanding the word progress – we may be thinking of progress in a single dimension or across a very small domain. Still — it is worthwhile working out what we should do if we believe that we are caught in a stagnation. is it more crazy projects hoping for break-throughs? Or is it, like Stanislaw Lem suggested a problem of the combinatorial growth of space of possible invention where the haystack is expanding but the amount of needles remains constant – where the answer is more scientific search capacity (through, for example AI)?

And this leads to a final point – maybe we ought to assume technological stagnation. Because everything we do to accelerate innovation will benefit us anyway!

Regulate Tech #2: The Arab Spring Revisited

This time, Richard Allan discusses the Arab Spring, what we have learned as a society and how the Arab Spring might play out today. It is an interesting discussion, and Richard was close to the whole thing in a way that makes it really worthwhile to hear him think through the issues and challenges.

All ideas and thoughts about other subjects, ideas or new things to discuss are warmly welcomed.

Regulate Tech 2023:1 The Online Safety Bill – Nearly There! Regulate Tech

In this first episode of the year we discuss the Online Safety Bill and the latest changes, new challenges and everything from criminal liability to age verification in practice. Tune in and send us your questions and ideas.    Participants: Richard Allan, Nicklas Berild Lundblad
  1. Regulate Tech 2023:1 The Online Safety Bill – Nearly There!
  2. Regulate Tech 2022 ep 20 – What a year it was!
  3. Regulate Tech 2022 ep 19 :The who, why and what of content moderation
  4. Regulate Tech 2022 ep 18: The DSA Halloween edition – trick or treat?
  5. Regulate Tech 2022 ep 17: The ABCDE model of what is driving regulation

Three pieces – Friday reading & links

Looking for new perspectives? Here are a few articles to check out.

  • “The New National American Elite” by Michael Lind. In this article, Lind suggests that the US has first now established a national elite – and risen from the fragmented state aristocracies that used to compete with each-other nationally: “In short, a historical narrative which describes a fall from the yeoman democracy of an imagined American past to the plutocracy and technocracy of today is fundamentally wrong. While American society was not formally aristocratic it was hierarchical and class-ridden from the beginning—not to mention racist and ethnically biased. What’s new today is that these highly exclusive local urban patriciates are in the process of being absorbed into the first truly national ruling class in American history—which is a good thing in some ways, and a bad thing in others.” I have written elsewhere in the newsletter about how elites provide an interesting analytical prism for us to think through, and found this article thought-provoking, even though I may disagree with some of it.
  • “The Unauthorized Story of Andreessen Horowitz” by Eric Newcomer. An interesting analysis of how more and more companies are becoming media companies, they need to have a media component to them to really be able to attract attention, capital, deals — your narrative is a key strategic asset. “Benedict Evans, Andreessen Horowitz’s former in-house analyst, has mused over the years that ‘A16Z is a media company that monetizes through VC.’ That observation becomes truer by the day.” It is fascinating to me that not more Swedish VCs adopt this strategy.
  • “Self-organized biotectonics of termite nests” by Alexander Heyde, Lijie Guo, Christian Jost, Guy Theraulaz, and L. Mahadevan. This is a fascinating model that suggests how termite nests are built and how their shape is dependent on a specific multi-agent model. “Termite nests are a remarkable example of functional self-organization that show how structure and function emerge on multiple length and time scales in ecophysiology. To understand the process by which this arises, we document the labyrinthine architecture within the subterranean nests of the African termite Apicotermes lamani and develop a simple mathematical model that relies on the physical and biological interactions between termites, pheromones, and mud in the nest. Our model explains the formation of parallel floors connected by linear and helical ramps, consistent with observations of real nests. In describing this multiagent system, we elucidate principles of physical and behavioral coupling with relevance to swarm intelligence and architectural design.” It seems far from impossible that these architectural designs may become more commonplace in the future.

Treat yourself to a fly-through of a termite nest below:

President Biden and the attention-deficit

As President Joe Biden now takes on the presidency he is facing a lot of substantive challenges, but he also faces an interesting stylistic one – and that is how he communicates. We seem to be back to press-briefings and less of the tweeting that the former president excelled in until his account was banned. That has been hailed as a relief, but there is a problem with that view: how much attention will the US and the world pay to president Biden compared to the enormous amount of attention paid to his predecessor?

The way we pay attention shapes our context, society and futures.

Think about it. How many hours will the average US-citizen spend thinking about Biden in a week, vs how many was spent thinking about the 45th president? How much of the media mindshare will Biden be able to own? Is there not a chance / risk that Biden gets but a fraction of the attention?

Why, then, is this interesting? For at least two reasons.

First, the mission of unifying the country is one that requires that citizens pay sustained attention to really developing their democracy and participating in it – something that will be hard if attention is paid to other things and led away from politics.

Second, the attention gap that emerges may well be filled with new politicians that adopt the former president’s attention strategies, seeking conflict and outrage as means to build mindshare and absorb all the attention that is suddenly up for grabs.

It may well be that the new administration’s most important task is to build an attention strategy – outlining how they aim to elicit and use citizen attention constructively, and that may mean that they have to adopt communication strategies that can be as engaging as those of conflict and outrage – or the new administration will just be a pause in which something else will eat all of the left-over attention and grow over time.

The midterms will, as always, provide more data on the way the political attention patterns are changing, used and developed by different politicians.

Wikipedia and the future of the Internet

Wikipedia celebrates its 20th birthday and there are a lot of interesting articles out there about the wonderful and weird phenomenon that wikipedians have created and are curating. Here are a few of the perhaps less well-known

“No Rest for the Wiki: The free encyclopedia is one of the last vestiges of an earlier internet” by Rebecca Panovka. In this article the author focuses not just on the accomplishments, but also some of the remaining challenges Wikipedia is facing – such as gender and language diversity, micro-aggressions and increased complexity. There is an interesting question here — how big can Wikipedia become? When will bit rot set in and start creating problems? There is a bet to be had here on how old the Wikipedia will become; most assume that the success of the project so far guarantee that it will continue to evolve in the same way over the coming 20 years, but there is an alternative interpretation (I am not taking sides) according to which we have already passed peak Wikipedia. That is not a reason to despair, rather the opposite — it is a reason to think about how the centralized collaboration of the Wikipedia could transition into a more decentralized, perhaps modularized, collaboration.

“An Oral History of Wikipedia, the Web’s Encyclopedia” by Tom Roston. This is a long, but worthwhile read with interviews tracking the history and origin story of the Wikipedia. It contains one nugget of quoted, that I think is interesting – and that is that original plan was to get to 100-200k articles, but the site grew well beyond that. That it could grow into millions of articles was not within the conceivable in the beginning. There is something about this that should remind us that we have built technologies that scale incredibly efficiently, and the flip side of that is that we have built things whose growth we often cannot control. The resulting size is interesting, and what determine the future growth curve of the Wikipedia will be interesting to find out. In fact, growth seems to be slowing — suggesting, again, that Wikipedia may be peaking.

The future of Wikipedia, no matter what, also says something about the future of the web, and modeling and discussing different scenarios for Wikipedia may give us some ideas about what is really going on with the web.