Modellhygien – om du förespråkar en modell måste du bygga den. Fysiskt.

Phillipsmaskiner är hydrauliska modeller av ekonomin. De simulerar och modellerar ekonomin som ett flöde av olika vätskor. Här är en video som demonstrerar hur den ser ut:

Det är, kan vi tycka, helt galet. Hur kan man tro att ett så komplext system som vår ekonomi kan beskrivas som en enkel maskin? Och det är precis där vi borde drabbas av en grundläggande insikt: alla modeller är variationer på detta tema, detta är vad ekonomer gör när de modellerar. Visst – de kanske gör det i datorn, eller med matte, men det är i princip detta de håller på med.

Det ger litet distans till starka uttalanden om vår förmåga att förutsäga och styra ekonomin eller hur? Det ger också ett uppslag till en fin regel för all ekonomisk debatt. Om du vill förespråka en modell, oavsett vilken det är, så måste du också bygga ett ex av den. Med vatten, myror, sand eller vad som helst annars, men du måste ha en fysisk modell så att vi kan se den som en modell, förstå att den är en modell och aldrig någonsin missta den för verklighet.

Det vore en god första regel i modellhygien.

Patent system quality and diversity

One of my pet peeves has always been that we measure innovation in a country by using patents as a proxy. Now, there is a lot to be said for the patent system, and a well working and built out patent system can be a force for good. But the mere number of patents seems a very bad measure. So it is with great interest that I see new research coming out that seems more geared towards understanding the distribution of patents between individuals and the concentration to a few firms. In the paper “Power Law Distributions of Patents as Indicators of Innovation” the authors examine who actually gets the patents. They write:

In conclusion, we have found that distribution of patents amongst applicants within OECD countries generally follow power laws. This provides a new way of looking at the structure of national economies and strengthens the analogy between innovating firms and ecosystems. Indeed, we fi nd that the power law exponents that describe these distributions diff er between countries and are correlated with measures such as national expenditure on research and development, and the ubiquity, or degree of specialisation, of the basket of goods that a country exports. Countries that export more specialised goods tend to have a smaller proportion of companies that hold a larger share of the patents where countries that export more ubiquitous goods tend to have a larger share of patents held in small portfolios. It seems that the innovator of today is more likely to work in the research laboratory of a large multinational company than in the suburban garage or small start-up company.

But this is where I think they go wrong. What the authors have shown is not that the innovators work in research laboratories and large companies, but that the patent holders do. And they are not the same. The power law that the study finds seems to be an indicator of patent concentration in systems, and hence also of lack of patent holder diversity. Both factors that should be used to gauge the health of a patent system.

And this is where I think we need to go. We need to look not at patents as a proxy for innovation, but dig deeper and examine the patent system quality, diversity and concentration to understand if it is serving the purposes it set out to serve, or if it is just creating an economy of transaction costs on top of the small start-ups that do not necessarily patent what they do to the same extent.

It’s the content, stupid…

In this paperExploring Text Virality in Social Networks by Marco Guerini, Carlo Strapparava, Gozde Ozbal, the authors argue that virality is not about the influencers, but about the nature of the content. If this is true, that predicting virality is not about knowing the structure of the social graph, but about understanding what kinds of content goes viral, that seems to undermine much of current thinking in the marketing departements around the world, where lists of “top bloggers” and “top tweeters” are assembled and closely examined.

And it does make some sense. When we predict the virality of, eh, viruses, we look at the viruses, not the social graph. But there would seem to be some threshold conditions that need to hold true for this to be the case (i.e. a basic density et cetera)

From the excerpt (my italics):

This paper aims to shed some light on the concept of virality – especially in social networks – and to provide new insights on its structure. We argue that: (a) virality is a phenomenon strictly connected to the nature of the content being spread, rather than to the influencers who spread it, (b) virality is a phenomenon with many facets, i.e. under this generic term several different effects of persuasive communication are comprised and they only partially overlap. To give ground to our claims, we provide initial experiments in a machine learning framework to show how various aspects of virality can be independently predicted according to content features.

The imperative of openness for data society

Aficionados of Isaac Asimov’s Foundation series will like this article from MIT Technology Review. In the article we can read bout how physicists have developed methods that allow forecasting, much like weather forecasting, of purchasing decisions. And the way they do this should sound familiar to Asimov fans:

Here’s how they do it. These guys think of humans as if they were atoms interacting with each other via three different forces. The first is advertising, which they think of as a general external force, like a magnetic field. The second is a word-of -mouth effect, which they model as a two-body interaction. Finally, they think of rumours as an interaction between three bodies.

The possibility of predicting the systems as the number of forces working on the atoms increase is of course the really tricky question. Asimov also used the analogy of a gas and establishes a number of important theorems that form the basis of psychohistory. One of the most important ones, but one that I believe is often excluded from enthusiastic discussions of how we could use data sets to predict different things, is the second (or third, accounts vary) theorem of psychohistory. It says:

that the population should remain in ignorance of the results of the application of psychohistorical analyses

I have seen people react differently to this idea, that the usefulness of a predictive system declines with how well the predictions are known, but most reactions seem to want to trivialize the problem. Yes, people will say, but let’s try anyway. I think that is a good attitude, but it is theoretically interesting to examine what it would mean if we move down the road that research like the one referenced in the MIT Tech Review opens up.

In one sense psychohistory is a manipulative tool. A society predicted loses some, perhaps all, of its democratic nature, and we move from a free market to a more planned economy. Hayek famously argued that the knowledge coordination problem is not only best solved, but only solvable, through markets, but there are those who believe that modern technology proves him wrong — that we could predict and plan society much better now with the large data sets that are accumulating everywhere in our society. They are not necessarily wrong, but their conclusion, per the second theorem, depends on those data sets and predictions being kept secret from others.

But, they argue, that is fine. Because the value of living in a carefully predicted and planned society is larger than the value of living in a society where everyone has access to the data used for the predictions and where these predictions are thus rendered useless. Their arguments will range from national security, health, economics to social equality.

Let’s assume for a moment that the complexity facing any attempt at psycho-history is not such that it renders all attempts futile. That we could in fact develop social predictions with high accuracy through the use of large data sets, and that we could act on them and plan our societies accordingly. The question we then have to ask ourselves is this:

Does the value of predictability trump the value of openness?

There is an assumption here that is worth highlighting. And that is that for a democracy to remain open it can not be predictable by only a few. That is a complex and perhaps provocative assumption that I think we should examine. I believe this to be true, but others will say that our democracy already is predictable, in some sections and instances, only to a few and that they build their power base on that information asymmetry, but that it is reasonably open still. Maybe. But I think that those asymmetries are not systematic to our democracy, but confined to those phenomena, like stock markets, where they are certain to be important, but where they also do not threat the nature of democracy as such.

In summary, if we share the data and allow everyone to use it, then predictability goes down. And I think it decreases fairly fast. So in a simplistic graph:

This leads to the guess that in the information society, open access to data is not only a great way to encourage new entrepreneurship, it is actually also a safe-guard for an open democracy. Governments are still, by far, the largest data holders, I think, and the data they collect is very suitable for social predictions. (And yes, companies should do this too. We do, and others do too, through tools like trends and insights, and we believe that the world is better for it, with lots of research flowing from that shared data).

I remember reading Asimov, and re-reading him, with awe. I loved the idea of psycho-history, and though that if only the math-geeks were in control then things would turn out just fine. I thought Asimov himself was a prophet. Turns out that he was quite sceptical, and often talked about both the advantages and dis-advantages of being able to predict societies and plan them.

In fact, in Asimov’s later writings it turns out to be a robot that encourages Seldon to develop his science into a social means of control. As we all know, the robots in Asimov operate under the three laws, and they want to reduce harm to humans. A predictable society would allow for that to be done as carefully as possible, but it would also curtail some of the human spirit, creativity and holy insanity that make us human.

If there is a conclusion here it seems to be to explore the amazing value of data under the imperative of openness to the the full extent possible to ensure that our societies gain from this new, fantastic age of data innovation, discovery and exploration that we are entering into, but never compromise on that openness in the pursuit of macro-social predictability.

Weil och yttrandefriheten som skyldighet

Simone Weil ställde i sin anmärkningsvärda bok Att slå rot frågan om rättigheter eller skyldigheter kommer först i samhället. Hennes svar är tankeväckande, och säkert provocerande för någon:

Begreppet skyldighet kommer före begreppet rättighet, som är underordnat och relaterat till det. En rättighet träder inte i kraft av sig själv, utan som en följd av den skyldighet som den motsvaras av. När vi utövar en rättighet gör vi det i ett sammanhang där andra anser sig skyldiga att respektera vårt utövande av den.

Rättigheter är något andra har, skyldigheter är något du själv har. Och rättigheterna andra har följer av att dessa andra också inser sina skyldigheter. När vi bygger upp ett samhälle, en konstitution eller en stat bör vi alltså inte börja med att lägga fast våra rättigheter, utan våra skyldigheter.

I ljuset av Weils analys tycks det mig allt mer klart att yttrandefriheten inte är en rättighet, utan just en skyldighet. Det är skyldigheten att låta andra yttra sig fritt som sedan också garanterar mig rättigheten att uttrycka mig utan begränsningar.

Det är, förvisso, en subtil distinktion, men det finns en nyansskillnad mellan att tala om någons rätt att yttra sig fritt, eller vår skyldighet att låta någon yttra sig fritt. Det kanske viktigaste i den skillnaden är insikten att yttranden aldrig sker i ett vakuum. Yttranden är yttranden för att de sker i ett språkligt sammanhang, i en kontext. Yttrandefriheten är inte atomär, den är en del av ett komplext språkspel, en av de grundläggande reglerna vi accepterar i det demokratiska språkspelet som just en skyldighet för oss själva. Det finns, för att parafrasera Wittgenstein, inga privata yttranden. Det är i våra delade livsformer som vi yttrar oss och formar en offentlig dialog tillsammans.

Av det följer kanske också några av de gränsdragningar som vi gör. Om vår yttrandefrihet är något andra är skyldiga oss, då bör vi också utnyttja den med just den relationen i åtanke, med sammanhanget i åtanke. Inte så att vi skräder orden, men med den gemensamma dialogen i fokus. Vi bör värdesätta yttrandefriheten som en del av vår relation med andra.

Simone Weil igen:

En ensam människa kvarlämnad i ett tomt universum skulle inte ha några rättigheter, utan bara skyldigheter.

Vad vore väl yttrandefriheten utan de Andra? Det är när vi ser yttrandefriheten som delad mellan oss och andra som vi också ser som tydligast att den nog är en skyldighet.

Det är möjligt att gå fel här, tror jag. Att läsa in fel saker i begreppet skyldighet. Att tro att respekt och kulturell relativism ger oss skäl att ursäkta de som sviker sin skyldighet att låta andra yttra sig fritt.

Men det är tvärtom.

Yttrandefriheten är en skyldighet. En skyldighet vi alla har att låta andra yttra sig, inte bara i vårt eget land, utan över gränser och över åsiktsavgrunder.

+++

Den här posten är en del i en bloggstafett. Fredrik Wass inledde och jag hoppas att Johan Norberg kanske kan fortsätta. 

SyFy-channel och den post-apokalyptiska vändningen

De mest populära serierna nu på SyFy handlar om “världen efter…” – efter zombies, miljökatastrofer så svåra att mänskligheten måste fly planeten, utomjordiska invasioner…Det är intressant att se hur livet skildras i dessa världar. Det är enklare, råare, mer fokuserat på överlevnad och sammanhållning. Det är ett anmärkningsvärt okomplicerat liv med starka värdegrunder.

Jag har sagt det förut, men det gäller fortfarande: längtan efter det post-apokalyptiska är i grunden värdekonservativ.

A note on biking and probability

There is probably a good explanation for this somewhere, but I have noted that cars never come alone. When you are passed by a car, another car or two more cars always pass too. So the best tip to avoid accidents is actually to n o t turn your head when the first car comes, but wait. I have seen multiple instances of folks looking back after having been passed by o n e car just to wobble and almost be crushed by the next.

I guess the probability analysis is one of bursty patterns and not too hard to work out, but still. The other thing about biking and probability is that the very place where you bike, side of the road, is statistically more likely to have cracks or damage, so attention to the road is well invested.

Third note in probability. The likelihood of fellow bikers asking if you are ok if you stop in California? 100 percent. The community – gemeinschaft – of road bikers is awesome.