The Big Data Fallacy

The latest issue of Foreign Affairs features the cover article “The Rise of Big Data” by Kenneth Cukier and Viktor Mayer-Schoenburger, which mostly details some of the incredible ways companies like UPS, Google and Apple have come to rely on vast arrays of numbers in order to run their businesses better. But data has always provided a problem in that it gives a substantive assurance of certainty that has a propensity to foster overconfidence in those relying on it. The article attempts to address this:

“[K]nowing the causes behind things is desirable. The problem is that causes are often extremely hard to figure out… Behavioural economics has shown that humans are conditioned to see causes even where none exist. So we need to be particularly on guard to prevent our cognitive biases from deluding us; sometimes, we just have to let the data speak.”

The sentiment here is admirable, and the context perceptive. But the final part of the quotation (my emphasis) assumes wrongly that data can speak objectively, that there is a fundamental ‘truth’ in a number. All too often though the wrong things are measured, or not all variables are measured. What data does not record, or worse, cannot record, can often be overlooked. While ostensibly data is there to provide assistance with building models and predicting future trends and movements, it sometimes leads to a very narrow view of one particular future, and fails to account for possibilities, that, though while unlikely, could potentially be devastating. This is what Nicholas Taleb writes about in his by turns unreadable but seminal work, Black Swan. The fictional, paranoid loner Fox Mulder of the hit series The X-Files had it right fifteen years ago when he lamented “in a universe of infinite possibilities, we may find ourselves at the mercy of anyone or anything that cannot be programmed, categorised or easily referenced”. The financial system before 2008 was a victim of such narrow thinking. 

Hendrik Hertzberg, in his Talk of the Town column “Preventive Measures” in this week’s The New Yorker, made the adroit analogy with the 2002 film Minority Report in our quest to categorise and predict acts of crime. Hertzberg points out that in reality this “turns out to be a good deal more difficult than investigating such an act once it occurs”. Indeed, such prediction methods are being implemented, just with somewhat less efficacy than in the Tom Cruise movie. The stop-and-frisk procedure currently employed by the New York Police Department points to a sustained effort to engage in preventative measures to reduce crime, effectively what Cruise and his myrmidons were doing, albeit without the help of psychic imagery as in the film. While the psychic “Pre-Cogs” turned out to occasionally disagree, the success rate with stop-and-frisk is even less attractive. “In the final months of 2012”, writes the New York Times, only 4% of stops resulted in an arrest. But what is this low figure telling us…?

Hertzberg also alludes to the dilemma of mountains of data, produced without concern for oversight or management; producing more just because it’s possible to produce it, rather than thinking about the implications:

“This fall, the National Security Agency, the largest and most opaque component of the counter-terrorism behemoth, will [open] a billion-dollar facility [analysing] intercepted telecommunications… each of the Utah Data Center’s two hundred (at most) professionals will be responsible for reviewing five hundred billion terabytes of information each year, the equivalent of twenty-three million years’ worth of Blu-ray DVDs… that’s a lot of overtime.”

The other problem this data poses – and increasingly this goes for many industries that are jumping on the Big Data bandwagon – is that intelligence departments and businesses alike are now technically able to put quantifiable targets and figures to what they want to achieve, without considering whether such targets are actually applicable. Police claim the low stop-to-arrest ratio implies that they are preventing crimes by stopping someone before they act. There is nothing to argue otherwise. The New York Times article alludes to the debate over what ratio or percentage the Supreme Court would be comfortable with under the tenet of “reasonable suspicion”. This leads down a dangerous path where we treat data as an answer to a question, rather than as supporting evidence to an answer.

Time for Trouble

Not too long ago, Zeitgeist was wandering home in the wee hours when he had his magnificent watch – a graduation present – stolen from him. The damage was minimal as Zeitgeist was quite blindingly drunk as he staggered by Hyde Park at 4am. Others however do not escape such thefts as unscathed.

As reported by Luxuo, on 25th November, Formula 1 CEO Bernie ‘Hitler was alright, democracy ain’t great’ Ecclestone was mugged. His watch, a Hublot, was stolen from his wrist. The octogenarian sent a picture of himself, severely disfigured by the affair, to fellow CEO Jean-Claude Biver of Hublot, writing “See what people will do for a Hublot”. It wasn’t long before it was agreed that the incident could be turned to the benefit of the company, and on the 8th and 9th of December, print ads appeared in the FT and International Herald Tribune, featuring the undoctored photo and Ecclestone’s quip.

It’s a somewhat tasteless ploy that Hublot, by literally advertising it, are implicitly condoning. In December they also tastelessly illuminated and branded the legendary column of Paris’ Place Vendome. But it also shows an innovative and creative spark in a sector of the ad industry known for its otherwise wholly uninspiring ads.

GMP get busy on Twitter

With the axe of austerity being wielded by the powers that be, all publicly funded bodies are praying that their budgets won’t be slashed too severely in the Spending Review on October 20th.

This will have been part of the motivation for Chief Constable Peter Fahy‘s decision to launch Greater Manchester Police’s social media experiment whereby they tweeted every 999 call they received over a 24 hour period.

Clearly feeling that much of their work passed unappreciated by the public and unrecorded in offical league tables, the exercise in transparency has helped change perceptions of the police more effectively and quickly than any boring spokesman or expensive advertising campaign could have hoped.

By the end of the show they’d tweeted well over 3,000 calls, gathered over 17,000 followers and arrested 341 people of whom 126 remain in custody.

They had also fallen foul of Twitters spamming rules meaning that they had to open several accounts after their original one was frozen due to excessive tweeting.

This forced GMP to set up new accounts to continue their campaign, going from @gmp24_1, @gmp24_2 and @gmp24_3.

Inevitably pranksters were quick to set up false accounts such as @gmp24_7 which included hoax reports in the style of the GMP tweets.

The campaign has been heralded as a success though hasn’t been appreciated by everyone. In what Zeitgeist feels is a slightly shortsighted view that fails to understand both the purpose of the campaign and the effectiveness of the channel, Fiona McEvoy of the TaxPayers’ Alliance complained that,‘The police should be catching criminals, not wasting time on social websites.’

Now that we know how busy the police are, dealing with issues ranging from reported sexual assaults to false alarms regarding four foot dolls near the M56 and burning hedges in Trafford, it could turn out to be an extremely cost effective exercise.

As one follower stated “They did more in the last five minutes than I thought they would in a whole day.”

With thousands of other followers reaching a similar conclusion the authorities may well think twice about making any drastic cuts to Chief Constable Fahy’s budget and dramatically reforming the structure of the Police service.