New South Wales Government Chief Data Scientist Ian Oppermann PhD is playing a key role in guiding the Australian government’s digital transformation, driving better analytics into numerous sectors to improve outcomes for citizens
What would you say have been your greatest professional achievements of the past 12-24 months, and why?
In the past couple of years, we’ve launched the New South Wales NS Data Strategy (2021), and the NSW AI Strategy (2020). We’ve also released a smart cities strategy (2020), so there are three major initiatives that are the result of a lot of work and wide consultation.
In terms of the COVID-19 challenge, in NSW it was around the middle of March 2020 that things really started to ramp up. When talking to colleagues about the sort of scenarios they were preparing for, they were focused on the rates of people able to attend work and not so much on systems starting to fail. So, I started to explore what we could do in terms of considering scenarios related to ‘continuity of service’.
I started exploring a whole range of different scenarios and ‘outcomes frameworks’ to think through the non-health implications of the rising impact of COVID-19. For instance, continuity of services for vulnerable individuals, such as children with elderly carers, and try to look at how you might prioritize actions to cope with an increasing disease burden.
These outcomes frameworks also provided a way to slot in partners who were rushing to offer help to do useful work to either better understand the changing situation, or create what-if scenarios. Improving situational awareness was really useful last year, as everyone wanted a better understanding of what was happening around them and to have that in as real-time as possible.
We also saw the ramp-up of use of a tool we developed called the Personal Information Factor (PIF) tool, which helps us determine how much information is in a linked dataset and so how identifiable an individual might be based on a de-identified dataset.
When COVID-19 hit, the NSW government was determined to release data on confirmed cases on a daily basis. We had a look at what kinds of data New Zealand, South Korea and Singapore were releasing. From this, we determined the type of data we wanted to release in NSW. We then used the PIF tool to work out how to protect this data and keep the level of personal information to an acceptable level. The PIF tool is used every day to ensure the level of personal information released as open data does not exceed that agreed threshold. That initiative has been really well received.
How has the data and analytics maturity in your organization evolved in 2021? What new challenges is this giving rise to? And how are you working to overcome them?
Most governments around the world take the approach that services are delivered in silos. Education does education, health does health. It’s becoming increasingly apparent that building an understanding of life journeys is really important to be able to drill through these agency silos of agencies to deliver more effective, joined-up services focused on individual needs rather than organizational structure.
People’s life journeys actually touch many different parts of government. Over the years, NSW has built some really powerful data assets to help with some of those very complex and challenging life journeys including to help reform home care which involves children at risk of significant harm, taken from their families and put in a protective environment.
We built life journeys of those children and their families as seen across many different parts of government. We’ve been working on a bigger version of that for the national disability insurance scheme, focused again on reform of the system to improve outcomes for people with disabilities. The Commonwealth has recently announced an additional $40 million AUD funding to take up the pilot programs that we’ve been building and move them up to being an enduring data asset.
These data assets are incredibly powerful. They are fine-grain detail. They follow the life journeys of children or families or vulnerable individuals. They also shine a light on complex systems.
Once you’ve got a really powerful data asset once you can say, ‘We’re not talking about 50% of people or 75% of people or 5% of people. We’re saying – while still de-identified – this is the experience of these exact people, these individuals’.
That really can be problematic when you shine the light on a system with that level of detail, for people who are really trying to do their best and trying to do the best with a lot of complex systems.
When you get that far out ahead of a problem, the analytical abilities can quickly get out of step with a responsible agency’s ability to respond, and so you can get this ‘rubber band’ snap back which can really hurt. Sometimes managing the information that paints a picture beyond the actual capability to respond creates a substantial ethical dilemma.
In your experience, what does it take to be a successful data or analytics leader? And what characteristics or skills should aspiring data leaders focus on cultivating?
I have really enjoyed being the Chief Data Scientist for NSW, rather than the Chief Data Engineer. A difference between scientists and engineers is this: If you give a problem to engineers, they will analyze the problem, pull it apart, work on the pieces and then bring back a solution and say, ‘There you go. There’s the answer’. With scientists, you give them a problem and the first thing they say is: ‘I don’t think you’re asking the right questions; I think the problem looks like something else.’ They will go away and then come back with a better problem which you can then start to deconstruct.
For example, smart cities are a really big, complex problem and unless you’ve got as many of the pieces as you need to come together, then you don’t make meaningful progress or you don’t deliver results that really impact the real world. Similarly, with AI, there are issues of ethics, and data quality, and appropriateness of use, and data bias, and governance. It’s not any one of those things on its own.
If you don’t consider them all together, you don’t make progress. So, data leaders should deliberately be re-complicating problems to consider all of these factors but then help people understand that we can address these different factors in steps. Communicating that we don’t need to address everything at once, but we do need to address all of these important factors in order for the barrel to hold water. If you miss any part of the barrel, it does not hold water.
What will your priorities be in 2022? And what are the key things you hope to achieve in the coming 12 months?
Smart cities, AI and data sharing are the three major areas. In the smart cities case, the New South Wales Government has set up the Smart Places Acceleration Fund. We’re about halfway through allocating that fund and the projects we are seeing are becoming increasingly innovative and ambitious.
These projects are all about the outcomes, getting meaningful indicators in place, and building those journeys of community and even journeys of infrastructure. New South Wales really is moving the conversation forward with smart cities and smart places.
In the AI space, the New South Wales cabinet is soon to endorse our AI assurance framework, the first of its kind in Australia. When it does, we then need to build it into an actual tool, which interestingly will have AI in it, which means it needs to be reviewed by our AI committee. Bringing that to life and making sure that we have a useful yet generic assurance framework to apply to all AI projects in the state will be important.
In data sharing, the big one is to get the standard finalized within the standards body JTC1. With the working group that’s on it, we got really close, but we aren’t likely to get it over the line this year. It’s a consensus-building process.
We want this to encapsulate as much data usage as possible, and we have spent a lot of time trying to put our arms around everything, which takes time, but we were determined to do it. Building consensus on every use of data is a big job.