Smart systems, smart policy solutions: the cognitive approach

By Tom Burton

Wednesday July 20, 2016

Smart, self-learning computers, powerful algorithms and advanced analytics are offering government major opportunities to help resolve a wide range of public challenges. From children at risk to litter reduction, false fire alarms, public transport planning and the accurate diagnosis of diseases like melanoma, the list of policy applications is long.

Hardware breakthroughs such as the iPhone have spurred massive app innovation. Similarly the development of powerful cloud-based computers is now driving large-scale deployment of intelligent software systems, often termed cognitive computing.

These systems typically digest huge amounts of data — numerical, text, images and video — and, using smart algorithms and applications, turns this data into useful actions. Over time these systems learn from the data they use and react accordingly.

For government the applications are numerous, but a simple example would be using real time public transport usage data from passenger fare cards to manage, say, bus and train fleets on an hour-by-hour basis.

Several of the major tech giants have invested heavily in cognitive computing, with IBM’s Watson platform arguably the most mature in the breadth of applications and depth of its solutions. The speed of development is surprising even seasoned tech observers. Earlier this year Google’s Deep Mind artificial intelligence team famously beat the world’s best Go champion — 10 years earlier than what some thought possible.

“The rapid embrace of cognitive computing is driving profound change …”

The rapid embrace of cognitive computing is driving profound change, underpinning a major shift towards network service delivery, data-driven decision-making and industrial innovation. Commercial services like Uber and Airbnb are just two early examples.

Government has a hugely influential role over the direction of these services, but is also set to be a major beneficiary of the shift towards intelligent systems.

Australia is a relative global leader in the application of cognitive approaches to improved decision-making, infrastructure planning, fraud and security detection, operational efficiencies, project management and large-scale customer care systems.

At the federal level, the merger of National ICT Australia and the digital productivity division of CSIRO into a new agency called Data61 is providing a powerful catalyst for national and state agencies seeking to tap the analytics expertise of the academic and industrial research community. Data61 has around 680 PhDs, including students and contributing staff. This includes about a third of Australia’s ICT PhD students. It’s the largest agency of its type in the southern hemisphere.

New CEO Adrian Turner has recently returned from the United States and is working with federal and state agencies to deploy intelligent systems. These include an engagement with Sydney Water to better detect pipe corrosion, road traffic agencies to improve safety, and with several agencies creating inference modelling using Australia’s first-class locational data system.

The Australian Tax Office is leading the use of advanced analytics among the bigger Canberra agencies. It has applied IBM’s cognitive solution to making better tax administration decisions. The platform has ingested thousands of previous tax rulings and is helping the ATO make better and more consistent decisions based on the precedents from these many rulings.

Tax commissioner Chris Jordan is investing heavily in cognitive and other intelligent systems and the ATO leads the whole-of-government Data Analytics Centre of Excellence.

Fire and data in NSW agencies

A recent McKinsey survey identified leadership as the critical ingredient for the successful deployment of advanced analytics — a strategy and technology tool kit are useful, but the key is direct senior executive involvement and support for the program. State governments, led by New South Wales, are also rolling out specialised data analytic centres, using advanced data analysis tools to help solve problems across a broad sweep of government issues.

One example? The issue of false fire engine callouts. More than 90% of callouts are false. This is a major waste of resources, and a possible danger for fire workers who become conditioned to most callouts being false alarms and therefore not mentally prepared for a “real” fire.

The NSW Data Analytics Centre has been building a model to understand what other data could help improve the accuracy of fire alarms. Data from electrical surges and even lunar periods is proving to be a valuable early indicator of the veracity of fire alarms.

Ian Oppermann
Ian Oppermann

The DAC is headed up by NSW chief data scientist Ian Oppermann, who is encouraging agencies to be ambitious in their use of data-led solutions. A number of lessons — the phase of the moon, for example, is seemingly important for accurate fire alarm detection — suggest agencies should cast a wide net when looking for insights from data.

The DAC focus is on what real-time data will make the biggest difference if it were available to an agency. 10 projects have been identified. Through a rapid development and discovery approach the aim is to have a minimum viable use case settled within three months.

Projects include modelling the hour-by-hour movement of people into and out of the rapidly expanding area to the south of the city so facilities can be better planned. The DAC is also focussed on how best to design a system that successfully integrates children into home care.

Other projects include identifying buildings at risk from non-compliant cladding and identifying property overcrowding.

The NSW government is now investing heavily in more whole-of-government data analytics projects. The DAC has received $17 million over the next four years to continue its work, with 17 more projects being considered.

Importantly, this is being strongly supported at the political level. “In a world rich with data, the DAC can diagnose the problems confronting many in government and allow agencies to provide the solutions,” said Victor Dominello, Minister for Innovation and Better Regulation.

Premier Mike Baird is personally engaged in the program and the DAC is collaborating with the Premier’s Implementation Unit on a series of projects to support the Premier’s 12 priorities. The PIU is being led by former Service NSW chief Glenn King and is bringing a much needed whole-of-government approach to solving some of the toughest problems of government.

These priorities include material cuts in domestic violence, obesity, youth homelessness and litter. The priorities have clear public outcomes and lend themselves to the use of advanced analytics to best understand what are the key drivers that create the issue in the first place.

For example, identifying bottled beverages as the biggest source of litter enables a focussed effort to remedy the litter problem. Observing that males aged around 20 are the main offending group also offers the option of a tailored education campaign.

Data to inform program design

Using data to inform and optimise program design is a huge opportunity for government, but does require a much sharper analysis of the outcomes required. Vague statements that focus on outputs, such as number of cases completed, will inevitably divert focus from the real drivers that will inform the best solution. While there has been much progress to switch to outcomes-based programs and contracting, it is critical the outcomes are well considered and do not lead to perverse results.

The use of data to inform policy and program design dovetails well with some of the behavioural economics thinking now being explored by governments to obtain better and more cost-effective results. Using data insights to better prevent disease and accidents is an obvious way to improve the overall health system. At a time of major fiscal pressures, only a minuscule amount of the national health budget is currently focussed on preventive measures.

Another major opportunity for government is the use of data to drive service design. The DAC has already created a model to better understand what drives off-peak usage of Sydney’s very large public transport system, with temperature emerging as a key determinant.

Another significant benefit is the reduction in costs by delivering intelligent services to citizens. Martin Hoffman, secretary of the NSW Department of Finance, Services and Innovation, says governments are facing a very real medium-term challenge of growing deficits. Hoffman says this means the quality of the services and productivity of those services (the output per dollar) becomes very important in any vision of a government embracing intelligent service delivery.

Government also has the challenge to lift the level of service to those citizens are experiencing for non-government services. NSW has led the way in service reform with its Service NSW initiative, a centralised portal and customer care network. That network is now providing real-time data back to officials — including the Premier — and informing ongoing service improvements.

Customer service commissioner Mike Pratt says the Premier now can answer the simple but powerful question: how did the government go yesterday?

In short time many of these services will be automated, meaning eligible citizens will simply get the underlying service without having to go through a manual application process.

Many government services are based on predictable lifecycle and life events and are ripe for automation. When a baby is born, they will soon be automatically registered for healthcare and education services. Ditto for an 18-year-old ahead of their first election.

“For government this means a rapid improvement in user experience.”

In this scenario data is critical to ensuring services are tailored and relevant to citizens and enterprises. Oppermann argues advances in search and recommendation engines have changed user expectations, with consumers less loyal and much more willing to switch.

For government this means a rapid improvement in user experience. It this means a joining up of services that typically have been delivered through multiple agencies. An early example is the work Service NSW and the federal Digital Transformation Office are doing to develop a single registration process for business start-ups. An early prototype is available at

These joined-up services require data to move across traditional portfolio boundaries. A major early challenge for government is to develop deep collaboration between the major dataset holders.

NSW passed legislation late last year to enable the sharing of anonymised data between agencies, which is a prerequisite if government is to take advantage of large-scale automation. This is removing the legal barriers, but the bigger challenge is the systemic silos and cultural impediments which are hindering the sharing of data across agencies.

This is a worldwide issue for government. The NSW Public Service Commissioner, Graeme Head, commissioned Nous Group to review the preconditions for better collaboration. The report details a very useful blueprint for agencies looking to promote deep industrial collaboration.

The issues of privacy and data security loom large in this space for both government and business, especially where there is highly sensitive data about health and children.

Adrian Turner says it is for this reason much of the technology is now focussing on bringing the analytics to the data, rather than moving the data to the analytic systems.

Turner also flags how blockchain is also promising to help. Blockchain provides a highly secure single registry which organisations can access, but not tamper with. This is vital to ensure the integrity of the underlying data — both to stop theft, but also the covert manipulation of data. This is especially so where data is being applied in real time to support security and other lifesaving systems.

The applications are diverse. Data61 has been commissioned by the federal Treasury to review the implications and applications of using blockchain technology in the financial services sector. Western Australian start-up is using blockchain technology to offer secure and instant electronic voting.

Building a strong analytics practice

Rapid improvements in data analytics applications are also making it much more straightforward for agencies to build a strong analytic practice within their agencies. These desktop applications enable policy officials to easily interrogate large data sets and are available now. The IBM Watson platform, for example, offers ready-to-use tools that can read and literally digest massive amounts of structured and unstructured data.

The McKinsey survey suggests the best gains will come from creating a dedicated data team to build capability around key projects. These can be externally focussed service upgrades or on internal operational projects.

Woodside Petroleum devotes huge resources to exploration and development. Woodside has collected all the data from previous projects and built a powerful library of experience. An algorithm interprets this data and guides better decision-making around its exploration and resource development program. Over time the system learns and improves, leading to even better decision-making.

This is similar to the work Sydney Water and Data61 have been doing to better predict which pipes are most at risk, and to focus maintenance resources accordingly.

Dr Joanna Batstone
Dr Joanna Batstone

Another area of huge benefit to government is the use of cognitive techniques in the health diagnostics arena. A team led by Dr Joanna Batstone, a Melbourne-based vice president for research at IBM, has brought together the huge amount of data around melanoma disease to enable doctors to make more accurate diagnosis and suggested remedies.

Batstone’s work is a collaborative exercise with the University of Melbourne and several leading oncology clinics who have joined forces to try and reduce the incidence of what Batstone calls “Australia’s cancer”.

Australia’s climate and outdoor lifestyle means we have some of the highest incidents of sun-fuelled skin cancer. Australia has some of the best research capability around melanoma. Batstone’s team has been able to work with these researchers to build a world-leading model to aid doctor diagnosis of melanoma.

Cognitive computers can be taught to read and understand human text. The Watson platform Batstone is using parses all the research data from the large collection of journals and other sources that are constantly being updated with the latest research breakthroughs. An algorithm has been trained using past cases to offer the best advice. This algorithm continues to be improved as more and more cases are fed into the system.

Advanced cognitive systems can also read images and video. This means the vast libraries of images many hospitals collect can now be interpreted. Batstone’s team is using this approach to improve the early detection of melanoma. These cancers exhibit quite distinct lesions. These can be compared to thousands of other images and Watson can make a reasoned insight as to the likelihood it is melanoma. Early detection significantly reduce the chances of the cancer causing death.

In a large, sparsely populated country like Australia there is an opportunity to send these images remotely to a central processing area, offering regional and remote communities access to similar services to those available in the city.

Batstone’s work underlines the need for a wide set of quality data sets for cognitive work to be most effective. These data sets almost always will be a mix of public and private data so there will need to be real collaboration between various institutions and jurisdictions for the wider benefits to flow.

Ensuring quality is a challenge. The federal government’s GNAF address database is a good example of a high-quality dataset and the efforts needed to ensure its integrity. This represents a decade of support and fostering by Drew Clarke, a former surveyor, but now Prime Minister Malcolm Turnbull’s chief of staff.

His commitment and determination is what practitioners, early pioneers and the research suggests is the key ingredient for success in the cognitive era: a strong leadership vision for what can be achieved using powerful intelligent systems.

The Mandarin recently hosted a colloquium with the NSW government and IBM on the topic of intelligent government and the opportunity for the application of cognitive systems for the public sector.

About the author
Inline Feedbacks
View all comments

The essential resource for effective
public sector professionals