TOC Justice and Allais' paradox Your Turn

October 19, 2010

Popular Justice, swift and bloodthirsty, is not the same as Justice. Bound to uphold the law, judges are not supposed to be influenced by the desire of making themselves popular. Yet one should be ill advised to ignore people's reactions on principle, lest injustice creep under the cloak of Justice.

Two weeks ago French judges condemned Jérôme Kerviel not only to a jail term but to repay the $6.8bn Société Générale lost through his unauthorized trading bets. More than a few found the punishment erred on the heavy side, as if meant to exorcise the risk investment banks ever become liable when their traders lose their moral compass, the way mining companies are found to be when their miners come down with silicosis.

Last week a sentence handed down by Norwegian judges in another financial affair had the same effect. According to Andrew Ward and Jeremy Grant, "the reaction among Norway's amateur trading community has been largely sympathetic towards the [defendants]" (*). Yet they "were found guilty of exploiting flaws in the electronic trading platform of a US broker to send "false and misleading signals" to the market".

"The enterprising traders, Svend Egil Larsen and Peder Veiby" have now received the full moral support of the editorial page of the Financial Times, who suggests "surely programmers of a robot that ratchets up prices undiscriminately whenever someone buys from it share the blame" (**).

There is indeed something uncanny at foot here. Require both judgments to be consistent and logic seems to imply that, had Kerviel been a machine without a soul, the counterparts on his badly programmed bets would have had to make his employer whole, unless, that is, they too were robots.

Still, bent by its ideology, the Financial Times wonders whether not let "the market root out stupid algorithms in its own way: by bankrupting their owners". The editorial staff should first check feasibility with its stable of able reporters. "Two years after the collapse of Lehman Brothers, regulators are working on ways to prevent it happening again. [...] Don't hold your breath", John Gapper advised us the day before (***).

The fact is markets can be and are successfully manipulated by forces stronger than their underlying mechanisms, themselves not exempt from design flaws. Whatever the fault of the US broker, our "enterprising traders" were too clever by half and the Norwegian judges saw through their guile.

How is it then human judgment can appear to be both rational and irrational at the same time? This very question inspired the late Maurice Allais (1). In his obituary (****), Phil Davison recalls this Nobel Prize in economics "challenged the dominant theory of how people made choices in the face of uncertainty. In [...] the Allais Paradox (2) [...] most people's natural inclination would be to act in ways the theory deemed irrational".

A philosopher would say that social sciences do not enhance their prestige by calling irrational a human being who fails to follow a model rather than the other way around. A mathematician would remark that the flaw Maurice Allais spotted in the theory was its blindly assuming the linearity of human choices, as if the solution when "adding" cases A and B together had to be the same as the "addition" of each individual solution.

To better appreciate his paradox, one should read how Maurice Allais assessed the cost of a train ride from Calais to Paris. It all depends on the scale at which one examines a problem. What is rational at one scale may be found singularly short-sighted within a larger context, which may include people bearing costs the lower scale decision makers generate but do not account for or, as in the paradox, a dominant, certain outcome.

At their own scale, the Norwegian judges have served Justice. They punished two market manipulators. The popular reaction however has assessed the penalty in a larger context, the investment banking industry. At this global scale they know Goldman Sachs never faced a judge for having avowedly fooled its clients, selling them products it knew would generate large losses for them and, with appropriate hedging, further gains for itself.

When the scales of Justice tilt in favor of the biggest, creating a spurious economy of scale (3), Might may make Right but Justice it is not.

I have claimed our Information Age itself is flawed because it rests on technologies which have torn apart the tacit agreements upon which our society relies. This point is worth revisiting in the light of the previous discussion. Most such tacit agreements depended on a certain scale. It is by changing this scale that apparently neutral technologies made the breakdown of these agreements predictable. Take privacy for instance.

Randal Picker is right to write "back in the dawn of sales, my local merchant knew me and knew what I had purchased". Indeed no one living in the typical village of the golden past could hope to keep much from one's neighbors' idle curiosity and everlasting gossip. Small scale privacy simply did not exist but people were not without defense mechanisms. Locally they knew about their local merchants and neighbors as much as the latter knew about them. And they could always leave their village and disappear into a larger scale, be it in the city, the army or even the New World.

Fast forward to our Information Age and both of these defenses have been defeated. First the village has scaled up to cover the entire world. Second local merchants have scaled up to become an Amazon, local pubs or tea houses have scaled up to become a Facebook, but individuals have not scaled up in proportion. As a result they experience a total loss of both the relative privacy and the global privacy they enjoyed before.

"Facebook lay[s] claim to leadership in [...] something it dubbed "social design"". Richard Waters detects there a stiff dose of marketing Newspeak (*****). But nobody can deny getting 500M users to give a profile to a single company is a genuine large scale social experiment and nobody said it sprung from an intelligent design. So Richard Waters is right to drill down and wonder if "it understands that it must [...] translate social norms on to the web, not risk subverting them" and whether "it will believe too much in its own ability to shape how large groups of people should interact".

I am however more pessimistic. Given its implicit business model must satisfy its foreign investors, its very scale makes Facebook intrinsically subversive. What is rational on a local basis becomes irrational on a global one, which dwarfs even the Stasi. In his self-interested myopia, Mark Zuckerberg differs from the well intentioned Norwegian judges only in that most people still overlook what scale does to sharing without eprivacy.

I also have a warning for society at large. Mark Zuckerberg may act as a rogue much as Jérôme Kerviel did but so is fully entitled to the Kerviel defense. They let me do it as long as I was winning. The French judge rejected it en bloc. But wasn't silicosis blamed on the miners a century ago?

HTML 5 for instance is a new technology which for "Pam Dixon, the executive director of the World Privacy Forum, [...] opens Pandora's box of tracking in the Internet", as quoted by Tanzina Vega (******). If Facebook avails itself of it, should Mark Zuckerberg shoulder the entire blame?

When the bill for retrofitting eprivacy to the Web comes due, will Justice assess responsibilies across all scales as Maurice Allais computed costs?

Philippe Coueignoux

  • (*) ........... A tale of man versus algo in Norway, by Andrew Ward and Jeremy Grant(Financial Times) - October 15, 2010
  • (**) ......... Man vs machine, editorial column (Financial Times) - October 15, 2010
  • (***) ....... The best bet to curb too big to fail, by John Gapper (Financial Times) - October 14, 2010
  • (****) ..... Visionary who warned of bank practices that brought crises, by Phil Davison (Financial Times) - October 14, 2010
  • (*****) ... Facebook's 'social design' risks turning unwelcome, by Richard Waters (Financial Times) - October 14, 2010
  • (******) . Web Code Offers New Ways To See What Users Do Online, by Tanzina Vega (New York Times) - October 11, 2010
  • (1) In the spirit of full disclosure, I happen to be an alumnus of the same school and of the same branch of the French civil service as Maurice Allais.
  • (2) for more details see the Allais Paradox in the wikipedia. The French version also includes a section on cost assessment.
  • (3) I am aware that product liability lawsuits also create a diseconomy of scale, but trial-based safety regulation is outside of the scope of this fillip.
October 2010
Copyright © 2010 ePrio Inc. All rights reserved.