I,Science Issue 35 (Winter 2016)

Page 30

Meet the Data Scientists

Artur Donaldson meets Professor Yves-Alexandre de Montjoye and Professor Michael Huth to discuss all things data.

I

it. It is data.

t is at our fingers but cannot be touched. We leave a trail of it wherever we tread. It drives our daily lives. Yet there are relatively few who understand

So who are the scientists with the power to make sense of it? Imperial College London is home to the Data Science Institute, and a renowned computer science department, who are delving into the depths of data science. We have put questions to two of its residents: Professor Yves-Alexandre de Montjoye, recently arrived from MIT, and Professor Michael Huth, who specialises in cybersecurity and trust. Professor de Montjoye works with metadata – unseen data providing information such as your location and your IP address. Metadata is crucial for making webpages work, but is invariably used to track consumers and generate targeted ads. Responsible companies try to anonymise metadata, for instance by splitting it into separate categories for recordal. But this is no longer enough for the scale and diversity of current datasets. For example, de Montjoye has shown in his research that locating four geographical positions of a mobile phone is enough to identify 95% of people. This isn’t just theoretical. Recently, a German newspaper revealed that millions of people had been identified from a database of URLs collected by browser add-ons, including the security extension “Web of Trust” (WoT). Is there a way to stop such invasions of our privacy? For de Montjoye, the answer is to think like the enemy. “Good data privacy is like hacking… You can’t be a good hacker if you’re inside and all you do is build up walls. To be a good hacker you need to know how it works … A lot of [my] work is to develop the tools that would be used to re-identify individuals”. Can we trust data? On the other side of the spectrum is Professor Huth, who works with trust. As

30

I, Science

he welcomed me into his office, I asked him how he would describe himself. “Not a data scientist … a trained mathematician who now works on security, verification, and optimization”. Huth is particularly interested in handling nuclear proliferation. For example, imagine that a country agrees to an arms reduction treaty. Another nation is called upon to inspect whether the arms stockpile is actually being reduced. However, the host country insists that weaponry is not visible during inspection in order to protect design secrets. How can the world know with confidence whether the treaty is being upheld?

technology has evolved, it has begun to challenge this unwritten right to privacy. For instance, with the invention of photography, and then the internet, it is easy and quick to capture and share an image of a complete stranger with millions of other people. The stranger has no choice or control over such a situation. This might not bother most people, but Prof de Montjoye explains it is an important issue to consider. “We could live in different times… it is essential to think of what the rules are in times when we trust our government, but also when we distrust the government or if we were to live in a different country.”

The answer is not big data, but just enough data, and the method used to collect it. Inspectors are given multiple identical devices. Inside the device sophisticated sensors assess whether a weapon is present, using factors such as the ratio of isotopes detected. From this, a simple yes or no is reported, which provides enough information to the inspector whilst protecting design secrets. Nothing to hide Huth has also used artificial intelligence to visualise the internal structure of organizations in terms of electronic interactions, such as emails. “In the future, HR departments will be digitised”, he says confidently, “[organisations will be able to see the activity of individuals] at any level of [detail] – this is a great opportunity for more effective HR - but a balance with privacy should be struck.” Some are terrified by such a prospect, citing an apparent disregard for workers’ rights. But, why should honest people be worried? Underlying the anger which Trump stirred up over Clinton’s emails was transparency. So I put the question to Professor de Montjoye, “why should people worry about privacy if they don’t have anything to hide?” “I see [privacy] as a right to be left alone, … think freely … develop an opinion”. In the past, he explains, privacy was protected by informal agreements but, as

Picture: Courtesy of the Data Science Institute. Back at Imperial… Before leaving, I was keen to find out what a typical day in the life of a data scientist involved: “Conferences, whiteboard work, and lots and lots of coding”. It may not sound glamorous, but among the academics I have met there is a spirit of moral imperative. “It is a question of values in society,” says de Montjoye. “Which kinds of societies do we want to live in?” Huth draws on another point, giving the example of drivers being required to always keep a hand on the steering wheel in selfdriving vehicles - “We will do well to always have humans in the loop in decision making that affects living things.” What is data science about then? Certainly not playing with, or restricting, the powers of data; if anything, it is about understanding their relevance to society.

www.isciencemag.co.uk


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.