The power and pitfalls

Use of super computers

A recent book The Finish by Mark Bowden — a product of meticulous research — has brought out the details of tracking Osama bin Laden. The most interesting thing that this story unfolds is the power as well as the limitations of ‘big data’ and high technology embodied in sophisticated surveillance systems, super computers and unmanned and undetected drones/satellites/missiles collecting intelligence and hitting precise targets from the high skies.

The huge mass of apparently unrelated bits of data about things like names, telephone numbers, addresses, travel plans, phone calls and internet messages of suspected or potential terrorists picked up by diverse agencies were being routinely fed into super computers over years and this huge data base was accessible to all the border security, intelligence and  law enforcement agencies.

The name of the courier (Ibrahim Saeed Ahmed aka ‘The Kuwaiti’)  who lived in the Abbottabad compound with Osama came up several times in interrogation from a number of unrelated sources. The supercomputer, using sophisticated data mining softwares, could establish a link between these isolated entries. Finally, a tapped telephone call from one such person (who used to make calls to Kuwaiti only after driving for more than an hour away from Abbottabad) saying that he was still working with the same people, followed by information collected by ground level CIA operatives drawn from the locals in Pakistan and US drones and satellites sending images of moving inmates from a distance, led US authorities to focus on the highly unusual Abbottabad compound and its residents.
According to the author, this would not have been possible without the highly sophisticated technological capability painstakingly built over the years since the 9/11.  At the same time, the author emphasises that, even with all the technological advancement, when president Obama ordered the final assault on the Abbottabad compound, his advisors could not be more than 50 per cent certain that the man in the compound was really Osama.

 It is also interesting to note that instead of using guided missiles or drones to hit the compound they settled for ground troops to land and kill Osama since they wanted to make sure that Osama could not survive by using any underground tunnel or bunker and his capture or death could be conclusively established. Thus, ultimately, despite all the sophisticated arsenal at their disposal, they had to use humans for the final assault, even at the serious (political and military) risk of getting involved in a possible confrontation with the Pakistani army or air force. 

Among other things, the collection and use of such huge data in a routine manner causes the interrogation and harassment of  many travellers to USA, simply because some part of their names matches with a name in that vast data base.

Differential treatment

 At the same time, the brothers charged with bombing at Boston Marathon were not properly investigated by the US security agencies, despite warnings from the Russian government against the suspected terrorist links of the elder brother. Subsequently one of them was granted US citizenship and the other one given the green card status. Possibly this laxity was due to the fact that this family was of Chechen origin whose resentment was thought to be directed against the Russians and not the Americans. Though Muslims by faith, their names were not the typical Muslim ones. They would have been treated differently if they originated from Pakistan or Sudan with standard Muslim names and suspected terrorist association was indicted by a friendly foreign government against them. This again shows how, despite all the super computers and ‘big data’ at the disposal of the authorities, human judgments and prejudices ultimately dictated the outcome.

The power of big data and sophisticated computing technology as well as its pitfalls have been underlined by another recent episode in the totally unrelated field of academics.
Two respected Harvard economists -- Reinhart and Rogoff -- wrote a highly influential paper in 2010, using data over more than two centuries 1790-2009 for a large number of countries, to show that the GDP growth rate faces a sharp decline after the public debt to GDP ratio crosses the threshold of 90 per cent. This finding was used as an intellectual justification by some policymakers, in US and EU, to press for austerity measures even in the midst of slowing economy and rising unemployment as reducing the debt/GDP ratio by cutting government expenditures (specially on social security) for heavily indebted economies (like USA and Greece) would help economic recovery by restoring confidence.
Very recently three researchers at University of Massachusetts, Amherst – using the same data set – have established that the Reinhart-Rogoff result was due to errors in entering some data and that there is only a mild negative correlation between the debt-GDP ratio and the GDP growth rate. Even there, the direction of causation is more likely to be the opposite of what Reinhart and Rogoff  postulated. Instead of high debt leading to slow growth the more likely scenario is that slow growth has led to high debt-GDP ratio for many countries. 

This kind of econometric analysis covering a large number of countries over centuries is being increasingly done by the availability of huge data sets over long time periods and sophisticated computer programs to extract results from big mass of data. But with vast data sets handled by a number of research assistants, the chances of errors (like omitting data or entering some wrong figures or wrong signs) creeping in and giving rise to wrong results have also increased. 

A computer software works out results in the black box based on what data are fed into it by humans. That is why these types of big data exercises need to be checked and rechecked thoroughly before the results are published. 

(The author is a former professor of economics at IIM, Calcutta)

DH Newsletter Privacy Policy Get the top news in your inbox
Comments (+)