Monday, January 23, 2017

BIG (BAD) DATA?

The term Big Data has well and truly muscled into our daily life. Not only data pundits but marketers, psychologists, academics and policy makers try to exploit Big Data for their own ends, which according to the IT industry, is the most important “natural resource” of our time.

The problem with data is that it can be presented in many different ways. The danger lies in the fact that while individual elements of data are immutable, inaggregate, a database can be manipulated to mean different things to different people. This also increases the possibility of drawing false conclusions from data sets.

One of the major areas where Big Data can be manipulated, whether knowingly or otherwise, is in crime prevention in the form of “predictive policing”– using Big Data to identify potential criminals before the commit the crimes. The pitfall here is the historical biases inherent in our criminal databases. For instance, the Pardhis, a notified criminal tribe are rounded up routinely by the police in many northern states whenever there is a crime in the area. Similarly, there are groups of people in other states who fall into such a category. If our computers were to blindly rely on these historical databases, this would reinforce historical biases and force the community into machine – determined discriminations.

A model’s blind spots reflect the judgments and priorities of its creators. Big Data requires fresh discussion of the nature of decision making, destiny, Justice. There is a great danger of mistaking correlations with decision making, if managed wisely. Big Data can offer key insights. Otherwise these weapons of statistics will turn into “weapons of math destruction”. Society needs to tame and disarm them before they mislead the policy makers.

***

[Source: articles in Mint dt. January 11, 2017 and The Hindu Business Line dt. 23, 2017]


-

Prof. K. K. Krishnan
Chairperson - CCR &
Prof. Centre for Insurance & Risk Management
Birla Institute of Management Technology

No comments:

Post a Comment