Enterprise risk management strategy: A planning guide for CIOs
A comprehensive collection of articles, videos and more, hand-picked by our editors
It's hard to believe today's CIOs could learn a lesson from Louis XVI, but that's precisely what Eric Brynjolfsson advises. "It's said Louis XVI didn't entirely understand he was in the midst of the revolution until he was standing up on the scaffolding of the guillotine," Brynjolfsson, director of the MIT Center for Digital Business and professor of management, said at the 10th annual MIT Sloan CIO Symposium.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
As Louis XVI perhaps finally understood, leaders can become so insulated from what's happening with the groundlings that they don't realize the status quo is no longer quo, Brynjolfsson said. Today's companies face a similar fate, as the big data revolution pushes the boundaries of how finely data is measured. That, in turn, is generating a new set of elite businesses -- those that are giving greater priority to data and less to instinct or gut feeling. Call them "data-driven."
The insight came from Brynjolfsson's introduction to an academic panel discussion at the symposium, where he and three MIT professors discussed how big data could potentially transform just about every industry, from finance, public health, transportation and IT services to -- you name it. Together, they delivered a striking message to CIOs: Getting on the big data bus isn't enough; CIOs need to take the wheel and drive (with caution).
Big data, big picture
What does being in the driver's seat entail? If they haven't already, CIOs should start by accepting that the big data revolution is real. New tools are uncovering intricacies and patterns in data -- lots of it -- that have never been seen before. Take financial services. Andrew Lo, professor of finance at the MIT Sloan School of Management, provided graphical illustrations showing how the industry has become increasingly interconnected since the Great Recession. The "before" and "after" slides showed a massive uptick of statistically significant correlations among insurance companies, sovereign countries and banks -- an unprecedented interdependency that has never been fully mapped out before, according to Lo.
"It's only recently that we've begun to take seriously the notion of financial services as a system," Lo said. "In the past, we focused on banks or hedge funds or insurance companies pretty much separately and in isolation."
Are people 'of interest' simply because they text each other, independent of what they have to say?
professor, MIT Media Lab
Big data is enabling the financial industry to literally visualize the system and its relationships, but for the data to predict the next financial crisis will depend on how it is analyzed and utilized. That's where the real value of big data lies, panelists said, and it implies that companies need to change how they make decisions. For years, companies have relied on what Brynjolfsson referred to as the corporate HiPPO -- the highest-paid person's opinion -- to be the official decision maker, but no more. "We need to put them on a diet," he said, "and turn our HiPPOs into geeks."
Correlation does not imply causation
That's easier said than done, it turns out, because not all data relationships are equal and that reality can lead CIOs down the wrong path. Cancer data, for example, shows a "shocking correlation" between lung cancer and ashtrays in the home, Lo said. "That would suggest if you eliminate all of the ashtrays, you can cure lung cancer," he said.
It's yet another data relationship CIOs need to be aware of, because correlation doesn't necessarily mean causality. To avoid making this mistake, panelists suggested striking a balance between traditional analytics and techniques that take advantage of some of big data's unique characteristics. High-frequency traders, for example, can dispense with building certain data models in favor of doing real-time experiments to test and grade a strategy, because they have lots of data and it's moving quickly, Lo said, but modeling isn't in danger of becoming extinct because those same experiments won't work for every scenario, such as trying to understand systemic risk.
"You can look at consumer credit data, home prices, mortgage applications, and you can look at that on a massive scale to get a sense of where the economy's moving," Lo said. "But nevertheless, you need to have some kind of filter to be able to interpret that, so they can turn correlation into causation."
Correlation invades privacy?
There certainly is power in correlating big data, but there are concerns as well. As one attendee said to the panelists, "Data at rest is benign, but when you start correlating that data, you run into security issues."
Some of the scenarios shared by panelists walked the Big Brother line -- none more so than MIT Media Lab Professor Alex "Sandy" Pentland's Afghanistan example. Intelligence agencies there are using prison release information to help keep tabs on certain individuals and predict potential criminal activity.
Erik Brynjolfsson, MIT Center for Digital Business and the moderator
Andrew Lo, MIT Sloan School of Management
Dimitris Bertsimas, MIT Sloan School of Management
Alex "Sandy" Pentland, MIT Media Lab
"They have certain signals which they believe statistically are significantly predictive of crimes that are about to happen, and that includes things like patterns of communication among criminals," Pentland said. "But it does raise the question: Are people 'of interest' simply because they text each other, independent of what they have to say?"
It might be an extreme example, but it's one that underscores the privacy concerns of law-abiding citizens and the hesitation from the corporate community to use that data. Those concerns weren't ignored by panelists, who described privacy as a "huge issue," and one that plays a role in their own research.
Lo, for example, uses secure multi-party computation, which gives multiple parties the ability to encrypt data; then researchers can poll that encrypted data and compute accurate statistics -- but with no way to uncover who's who and jeopardize personal privacy. The privacy safeguard comes at a cost: "The problem in all of these methods is that it increases the computational burden, and so you want to look for efficient algorithms," Lo said.
The Media Lab's Pentland, though, is taking a different approach. He is working on a project with Orange, a telecommunications company based in France. By allowing researchers to work on (anonymized and aggregated) cell phone data collected from its customers in Ivory Coast, Orange discovered how to reduce commute time by 10% and even how to help reduce by 20% the spread of AIDS and malaria and other infectious diseases. Pentland said Orange holds out hope that its findings will offset public concerns about the use of cell phone data.
2015 MIT Sloan CIO Symposium
Explore this ultimate guide for SearchCIO's 2015 MIT CIO Symposium coverage, including video interviews, panel recaps and a thorough investigation into the impact of digital disruption.
"The idea from Orange's point of view is if they can show the data they have creates public good, people will get used to the fact that there's this [big data] resource for society," he said.