Post

Rationality and Society

/ Adam / 2 Comments

Oh those silly rational agents! What ridiculous duffers we were to ever believe in them. If you’ve read anything about business since the 2007/8 crisis you’ll already have read enough of this sentiment to wallpaper your flat with, but in the unlikely event you missed out, it was all pretty much covered in this op-ed by Joe Stiglitz in 2002 anyway.

And of course he’s right to a point. People aren’t perfectly rational, they are biased in certain ways being fascinatingly explored by behavioural economists, economists can be fixated on nice mathematical models.

And yet, and yet.

A lot of these attacks on rationality are intended as attacks on free markets, that rationality in this sense captures people at our worst. But despite the bad press of late, rationality isn’t just the domain of sociopathic capitalist uber-robot-mensch. It is, for instance, a key premise of Rawls argument in A Theory of Justice:

25. THE RATIONALITY OF THE PARTIES

I have assumed throughout that the persons in the original position are rational. In choosing between principles each tries as best he can to advance his interests.

Rawls fuller argument, in a nutshell, said a just society would be the one chosen by rational but ordinary people who did not know what role they would play in it. They would have to choose the structure of society from behind a veil of ignorance which concealed whether they would be rich and priveleged, or poor, or a redhead. Under these conditions, Rawls argued they would choose two rules; firstly liberty, secondly the prosperity of the weakest class. He then argues that redistribution is just to the extent it benefits the poorest in absolute terms; you can’t sacrifice the wealth of the poorest for pure equality where everyone would be poorer.

Rawls cites Amartya Sen and social choice theory, looking to economics not just for part of his argument but part of his premises. A Theory of Justice effectively revived the idealist or contract tradition as a counterpoint to the utilitarian tradition then dominant. He was also a defender of the welfare state liberalism of his time, and a good one. One marker of the breadth of his contribution in reviving an idealist tradition is the phenomenon of libertarian Rawlsians, like, say, Will Wilkinson; they tend to like the veil of ignorance but not the distributive principle.

Only in a social union is the individual complete. — Rawls, A Theory of Justice, again

((Rawls was too much the careful scholar to go in much for quotable quotes: he is readable but longwinded. He also preferred to caveat his sentences with learned references and restrictions on their scope. The sentence above he only let slip in an unguarded moment at the end of a footnote.))

There are two contrasting uses of rationality here, one from social choice theory, one from efficient markets theory, that are slightly different, but only slightly. The point is that rational agents aren’t just a vision of utility maximising robots chewing each other to pieces. The considered reasoning behind the veil of ignorance also shows rationality as a vision of people at their best.

Flu Seasonally Adjusted

/ Adam / 2 Comments

Permutations points out an elegant paper from Christakis and Fowler (gloriously open access). They exploit a clever result from social network theory called the friends paradox. This is the phenomenon that your friends have more friends than you do – because social networks typically have a few very connected spoke nodes. They use this to track flu within a university student population. By separately tracking the friend cohort they were able to note the evolution of a flu epidemic several weeks before its full arrival in the general population as represented by the random cohort.

Current surveillance methods for the flu, such as those implemented by the CDC that require collection of data from subjects seeking outpatient care or having lab tests, are typically lagging indicators about the timing of the epidemic (information is typically one to two weeks behind the actual course of the epidemic). […] [W]hile potentially instantaneous, the Google Trends and prediction market methods would only, at best, give contemporaneous information about rates of infection. In contrast, we show that the sensor method described here can detect an outbreak of flu two weeks in advance. That is, the sensor network method provides early detection rather than just rapid warning.

Wiring up a distributed computer of neighbourhood gossips to see into the future is presumably a trick with wider applications. For instance, economic data is not only notoriously bad, but notoriously slow. It’s a field where price data from three month old lagging indicators are siezed on with delight at their timeliness, and GDP figures have to be seasonally adjusted a year after the period they apply to. Economic actors also behave as a network for the flow of information and beer.

So, you should be able to systematically exploit this effect in economic surveys to get both more timely results, and information on the velocity of effects throughout an economy. Eg, if you are surveying businesses, get those businesses to also nominate their suppliers and customers, and track that group as well. It’s possible this technique is already used, and I’d be interested to hear about it. I suspect that its main use is though data collection folk wisdom rather than systematically. So it’s well known that health workers are vulnerable, highly connected nodes for disease spread / containment, and that banks and large retailers are hubs of economic activity, but that knowledge is not generalisable in the same way as the friend cohort in the paper. Perhaps you could even use techniques like this to build a network model of critical financial institutions, from the perspective of vulnerability to systemic failure under a catastrophic crisis.

The Napkin Scrawls of Dining Philosophers

/ Adam / 4 Comments

Technologists are, by their vocational commitment to new things, manufacturers and early adopters of language. Our commitment to language is however generally one of casual incompetence. The artifact being built, or fixed, is the focus of our attention, and the language referring to it is an after thought for the tormenting of poets, grammarians and the marketing department. Perhaps that’s as it should be, but it’s still a pleasure to read a book like Java Concurrency In Practice, which has a mastery of both language and its topic.

JCIP has already been well reviewed on its technical merits. [NB: These notes are also from 2006 and written about the first edition.] In summary, it’s a great reference. As the jacket copy points out, concurrency in software is both difficult to deal with and of renewed importance. Problems like these put stress not only on software artifacts being developed, but the context in which that artifact is built and used, like collaboration in its construction, or human and computer interfaces, or eventual maintenance. Java Concurrency In Practice also has some insights into this, but it’s not foregrounded, and seems worth exploring.

Continue reading →

Scribed On Demand

/ Adam / Leave a comment

Two opposite poles of art creation in Paul Clark’s acclaimed art history of the Cultural Revolution, the slightly misleadingly titled The Cultural Revolution: A History, illustrate his thesis marvellously. Clark argues, with voluminous examples, that the Cultural Revolution was an artistically productive time, even if it was also a politically terrifying one. He takes aim particularly at the received wisdom in a common joke; 800 million people watching 8 performances (八亿人看八个系).

方海珍（李丽芳饰）是《海港》里的女主角； Fang Haizhen (played by Li Lifang) from The Harbour

The first pole of art creation is the model operas (样板戏). These were highly professional productions closely supervised by the cultural leadership including Jiang Qing (江青). These have a reputation as clumsy kitsch. Clark points out, with their production values, long lead time, and close executive supervision they are actually the very pinnacle of high kitsch. They made a number of technical innovations that moved Chinese opera smack in the middle of the twentieth century. Their focus on clearly delineated roles, modern settings, post-war language and ideological agenda – arias about revolutionaries and Mao Zedong are commonplace – make them distinctively modernist projects. Also, far from cultural troupes being entirely disbanded, certain parts of the culture industry were kept very busy on large productions of this kind. The striking film still above is from one of these model operas, 《海港》 (The Harbour).

To me, the Antarctic counter-pole to the model operas are the hand-published books. Amongst metropolitan youth sent into the countryside when schools and universities closed, there developed a subculture of letters, poems and entire novels written by hand, copied by transcript or mimeograph, and distributed by being passed from person to person. 《第二次握手》(The Second Handshake) was the most successful of these, eventually landing its author in jail for two years, followed by being published and filmed in the eighties. Works written in this fashion were enhanced and edited by their readers and transcribers, in a　cultural movement part schoolroom note, part monastic rescribe and part wikipedia collective publication.

Clark was a student in Beijing at the end of the seventies, so he’s able to add personal anecdote to careful scholarly description in a tone reminiscent of antiquarian journal Arts of Asia. Though a fair number of pictures are included, it’s best annotated by Google to get a fuller impression of the many works he mentions.

Finally a minor quibble – why do academic books on China, written in English, use (non diacritical) pinyin rather than Chinese character translations? I doubt much of the audience is able to use the pinyin that is not able to use the characters themselves, which are also far less ambiguous. Is it a printing cost limitation? It seems archaic in a time of Unicode and print on demand PDFs.

Portal

/ Adam / 3 Comments

But, soft! what light through yonder window breaks?
Laser guided, a gun turret aimeth. — Romeo and Juliet, Act 2, Scene 2

Portal is as much a poem as a game. The creators take the vernacular elements of the first person shooter and make a discliplined selection from them. For instance, the only weapon you equip doesn’t actually hurt anything, because that option is simply not available, and doesn’t make in game sense. Falling damage has been removed. They then add one new one element – the famous portals. In the words of Jeep Barnett, one of the developers,

We try to understand our limitations and work within those and in some ways embrace them.

And embrace them they did. Once the game’s chosen elements and their relationships are established, it recombines them with wit and grace. And it tries to kill you a few times too, but what is art without struggle?

Though analogies from one form of art to another are not perfect, and I doubt the team set out to write a poem, the description as poetry is apt because the formal constraints of Portal make it possible to describe it in a kind of ludic meter. The remainder may spoil. (I also owe a debt to work from the Rocketboom Institute of Internet Studies on image macro haikus).

Like Goethe’s Faust, Portal is divided into two distinct parts, which differ in structure and content. Part 1 has a whimsical tone, but a formal structure reminiscent of a sonnet. Each level can be treated as a stanza. The structure of each stanza is then

G (O B|B O|S)* G

G is dialogue from the AI Glados, half sexy librarian, half HAL from 2001. Each level starts and ends with AI dialogue, in a great performance from Ellen McLain. The repetition is not as strict as a villanelle, but is similar in linking the beginning and end of a stanza. The AI returns to themes throughout the game (eg, cake).

B and O are the blue and orange portals respectively. These form rhyming couplets. It is the chain of portal couplets that progress you through a stanza. When you go in one portal and out another, you move one or both portals and repeat the process. In terms of rhyme scheme, going in the orange portal, and coming out the blue is O B. Then, for example, moving the orange portal and returning into the blue portal is O B B O. A Shakespearean sonnet has the form (A B A B)^3 C C, that is, three sets of A B rhymes followed by a C C rhyming couplet to finish.

The structure of each line is not restricted by time meter as it would be in the iambic pentameter of an English sonnet. However, the deliberately limited vocabulary of Part I, using floor buttons, boxes, doors, lifts, energy balls and switches, and not much else, still gives a sense of formal restraint.

S is a substanza. As Portal proceeds, the levels get more complex, and the AI will bookend sections of a particular level with more dialogue.

What distinguishes this game from a conventional puzzler is the retention of a sense of flow. This seems to have come from Valve’s process of rigourous playtesting (mentioned in the interview above). This in turn fed back to the reduced vocabulary and minimalist setting, which reinforced the formal structure. It means you don’t get stuck on an impossible or frankly tedious puzzle. This also shows the limitations of playing a game in poetic mode. It’s as if you couldn’t go past page three of Paradise Lost unless you read out the fifth line exactly right.

At the beginning of Part II, the AI famously and hilariously asks you to assume the party acceptance position. When you, as the protagonist, reject this and scoot off into the factory innards of the laboratory, the game itself leaves its formal structure for something more anarchic. It’s actually a more conventional FPS setting, with more jumping onto lifts and such according to timing, though you still don’t shoot anything except walls. The climax, though, despite Eric Wolpaw’s hatred of plays, really does have a wonderful setpiece, a coup de theatre.

Valve, this is a triumph. I’m making a note here: Huge Success.