Big data: big benefits to public service, big risks to public?

By Stephen Easton

Tuesday November 3, 2015

Public service leaders are increasingly enamoured with big data, and the big possibilities of capturing, storing, combining and analysing increasing quantities of it. Now they must build the workforce capability and infrastructure to realise them.

Department of Employment secretary Renee Leon commented that public servants can’t all become data scientists but they do need to get “much more savvy” about the possibilities.

Richard Collis
Richard Collis

The big data era offers government agencies some enticing prospects. In the United States, predictive algorithms are being used to map out future crime levels in different areas, or predict the fallout from interruptions to public transport services. More research is being commercialised all the time, making new tools and techniques available and potentially making risks easier to mitigate.

Vast and growing stocks of increasingly diverse data can provide powerful insights, but the act of obtaining them is made difficult by the sheer volume of information. Richard Colliss, an assistant commissioner at the Australian Taxation Office who also chairs the Commonwealth Data Analytics Centre of Excellence, thinks of this as the crux of what big data means.

A review of public service data management is currently before cabinet and should be released within weeks, Colliss told the recent IPAA national conference in Sydney.

“… not only are we improving policy debate, but we’re actually doing it more quickly.”

The DACoE is currently surveying analytics capability in the Australian Public Service and working with the APS Commission on development models and pathways to improve it. It’s also talking to universities about talent identification programs and course content and has a prototype online “analytics space”, some of which is only open to people with a email address, which presumably includes state and territory agencies.

But Australian public sector agencies face the same challenge as businesses across the rest of the world in building capability: a skills shortage.

Colliss doesn’t think this will be a problem in Australia for much longer. He says the data management skills that are already common in the public sector can be “scaled up” as a growing number of skilled data scientists amble out of universities. Data science will probably even become a big export earner for our higher education sector, he suggests.

And just as his own programming skills are now eclipsed by those of clever 12-year-olds, Colliss also predicts the mystique of the profession to dissipate rapidly. He expects tools that make big data more accessible to “improve in a way that will actually not negate, but diminish our demand on data scientists into the future”.

What data can do

The opportunities of big data are not confined to any one area. Colliss says the ability to “analyse rich layers of data from a variety of sources in a variety of formats to inform policy decisions” more quickly than in the past could help departments become even more responsive to the needs of governments.

“So, not only are we improving policy debate, but we’re actually doing it more quickly,” he explained, adding it’s “self-evident” that data analytics could also provide a clearer view of economic, social and environmental conditions.

Another opportunity is better, more objective policy evaluation, giving deeper insight into why programs succeed or fail. Analytics can also improve service delivery through “tailoring and targeting” and identifying areas that are being over-serviced, says Colliss.

“Things like increased transparency and accountability are [also] opportunities for us that arise out of big data,” he added, endorsing forecasts of major cuts to administrative overheads, and increased revenue for governments to arise as well.

New tools and technologies, as mountains of data-focused research is commercialised, would help mitigate the attendant risks.

Colliss gives four simple examples of data-driven activities already underway in the federal sphere. The Department of Health is forecasting hospital admissions years ahead by linking with the Australian Bureau of Statistics projections of population growth, which is more “joined-up data” than big data, he says.

“Approaches using big data allows analysis to be moved to the point of a transaction, so we can influence a citizen at the point of them interacting with government.”

The Department of Immigration and Border Protection has an “online, real-time system” that allows it to screen in-bound passengers and stop some of them before they even board a flight to Australia, based on an individual risk assessment. After limited trials in South-East Asia, it’s ready to go global.

DIBP is also exploring the abilities of IBM’s suite of Watson artificial intelligence products to help search through masses of information it hoovers up from digital sources. The Australian Crime Commission’s five-year-old Fusion Capability is creating clearer pictures of criminal organisations using “multiple intel sources” — particularly interpersonal communications like intercepted phone calls and metadata about who is calling whom, Colliss says.

The ATO is using advanced “nearest neighbour” analytics to hone in on suspicious tax returns, Colliss explains. At this point, his phrasing began to conjure up images of the sort of technologically controlled society often imagined by sci-fi writers.

“We have the ability now to compare you as an individual with 30 of your nearest neighbours,” Colliss said. The “algorithmically derived information” tells the Tax Office if you are “in alignment with everybody else or out of alignment”.

It used to take two weeks to run that algorithm, for only 3 million taxpayers. It now takes 18 minutes and is applied to all 13 million taxpayers. Next year, the e-tax and myTax applications will automatically ask pointed questions if one submits eyebrow-raising claims for work-related expenses that vary wildly from one’s peers.

The alerts that come up before the tax return is submitted give the data-driven system the equivalent of warning signs a few hundred metres before speed cameras.

“Approaches using big data allows analysis to be moved to the point of a transaction, so we can influence a citizen at the point of them interacting with government,” said Colliss. “That’s one of the lessons we should learn and one of the directions we’ve been taking everything.”

Big data should be a boon for citizens, he said, by making it quicker and easier for well-behaved citizens to interact with government: “[If] I want to take out a payment arrangement and if I’ve been a good taxpayer the Tax Office would say, ‘yes’. If I’ve been a bad taxpayer and I wanted a payment arrangement, the Tax Office would say, ‘you need to talk to one of our operators’ — that’s where we’re heading.”

Privacy must be protected

As public servants work to link up data silos across and between governments, the importance of “managing public perceptions” — and particularly “this notion of Big Brother” — cannot be ignored, Colliss argues.

The DACoE chair suggests the long-standing general rule that information should only be collected and held by government agencies for a specifically defined purpose will need to be done away with. Public servants sometimes claim they are legally prevented from sharing certain data when it isn’t strictly true, just out of convenience.

Legislation could also be re-interpreted by public servants in a more liberal way, if they believed linking up certain data aligned with the true “intention” behind the law.

On the other hand, some of those who are enthusiastic about sharing data across government focus mainly on the benefits, the inevitability of digital transformation and the need to keep up. But the inexorable tide of change is no reason to ignore or play down the clear risks to the public, just as the convenience of a particular app might not be worth the personal information it demands from users.

To that end, the DACoE is testing out ways of determining “whether there are real problems or [only] perceived problems about sharing of information” in particular cases. The pilot project involves comparing the social implications of business failure, to the risks of sharing data between agencies to stop it happening as often.

Timothy Persons
Timothy Persons

It pays to remember that data is dangerous in the wrong hands and that it is difficult to keep it secure and make it useful at the same time. As noted directly before Colliss spoke by Timothy Persons, chief scientist with the United States Government Accountability Office, anonymous data can be linked back to real people using only a few other attributes. And when analysed, even a relatively small amount of data about a person’s online activity can easily reveal a lot about them — often more than they would confide in friends and family.

Persons suggested executive government — which his employer holds to account on behalf of Congress in the US — should be extremely transparent about what sort of data it holds and why, how it secures it and even how the algorithms work. He also urged public servants not to copy commercial apps and websites and bury that information in long, legalistic agreements.

“I don’t think anybody reads that; you click through, you want to use that thing,” said Persons. “The public isn’t getting that and I think the public sector would do well to not go there … I think the private sector’s going to have to change dramatically in that regard.”

About the author
Inline Feedbacks
View all comments