A new academic paper proposes that three public sector trends — behavioural science, predictive analytics and the investment approach to welfare — can combine powerfully into what the authors term the “digital nudge”.
Combining behavioural science and predictive data analytics through a process called the “digital nudge” offers a way to sharpen efforts to identify who is most at risk of welfare dependency and tailor interventions that work for them, a new academic paper suggests.
Authored by Australian National University professor of information systems Shirley Gregor and Brian Lee-Archer of the SAP Institute for Digital Government, and published in the International Social Security Review, the article speaks directly to the popular “investment approach” to welfare.
Australia’s implementation of the investment approach — recommended by the McLure welfare review and pioneered by New Zealand — involves conducting an “actuarial valuation of the lifetime liability of Australia’s welfare system” based on a set of four longitudinal surveys to identify at-risk groups.
The need to measure and improve the impact of welfare expenditure — rather than just accounting for outputs — has come into sharper focus in a lot of nations since the global financial crisis. In financial terms, lots of governments want a better return on that investment in the form of welfare recipients getting more specialised assistance so they can become taxpayers sooner.
Australia’s current social services policy also revolves around the idea that a lot of welfare expenditure is the result of bludging and rorting, leading some in the social services sector to criticise the Turnbull government’s version of the popular new investment approach. Some experts in New Zealand are also highly critical of how the policy has actually turned out there.
Nonetheless, with behavioural science nudges on the rise and the use of predictive data analytics also becoming more popular in the public sector, the authors suggest governments can take advantage of this confluence of three public administration trends:
“These trends, when considered together, provide the inspiration for a new approach within social security administration, the digital nudge, for improved social outcomes. The digital nudge is based on individual circumstances with the intent of achieving a social outcome. The social outcome must be consistent with the social contract as determined by the political process.”
“At the intersection of those three areas, we’re saying there’s something actually quite unique here because in a social investment model you’re looking for better outcomes at an individual level,” Lee-Archer told The Mandarin. “We have a whole range of social programs … but there’s always cohorts of people where the standard policy or the standard programs just don’t quite work.”
Gregor, one of two directors of the National Centre for Information Systems Research, said governments could do more to use the growing power of computers to learn more about the different kinds of people who receive welfare — as marketers have done with consumers for a long time — as well as to trial and deliver potential nudges that might help break the cycle for those most at risk of long-term dependence.
While their article is focused on welfare, Lee-Archer thinks the concept could be extrapolated for use in other policy areas as well.
Advanced analytics for targeting and tailoring
One of two case studies that show some elements of the digital nudge in action looks at how a regular nudge — “simple education and communication” to outpatients — has reduced hospital re-admission rates in the United States by as much as 30% in some areas.
The concerted multidisciplinary strategies some hospitals employ to keep chronic heart disease patients out of hospital have been helpful to a point, but failed to make much of a dent in re-admission rates. According to the paper:
“One reason is that it is hard to precisely identify target patients, and there is a lack of tailored preventive intervention for patient preferences.
“Against this background, some hospitals have sought a more advanced solution in predicting readmission risks, with the use of health information technology (HIT).”
“There’s more opportunities for that sort of thing — but see, they hadn’t thought of that as a nudge,” said Gregor. “Once we sort of wrap it up under this concept of behavioural insights or nudges, it gives people more ideas about what they can do.”
Big businesses like big banks and retailers are already quite gung-ho about the potential for using predictive analytics to target tailored marketing material to consumers. “Governments should be able to do that — work out when someone’s on the point of doing something, and actually make some intervention,” said Gregor.
The research paper goes through some advances in data collection and predictive analytics — including the rise of real-time data analysis, which allows models to adjust and re-adjust more quickly.
While it mainly refers to the typical nudges made popular by the book of the same name, it also notes another similar theory referred to as “think”, which encourages governments to help citizens “think through challenging issues in innovative ways” by giving them evidence and a wide range of opinions on the matter.
Gregor and Lee-Archer provide a simple framework to explain their proposal that typical nudges or “think strategies” can be enhanced through “the use of advanced information technology, including very large data stores and warehouses, and techniques such as data mining and predictive analytics” that are improving all the time.
Caution: ethical concerns ahead
The authors are also careful to draw attention to “concerns regarding ethics and privacy” that go hand-in-hand with the suggestion of applying nudges to individuals rather than at a population level:
“The use of data and personal information to drive the nudge process has to be managed in such a way that individual rights are protected.”
Lee-Archer says they are not suggesting a computerised system that operates “behind the scenes” or tries to modify citizen behaviour through punitive action. This would take it out of the realm of nudges, which generally aim to improve outcomes under existing policy settings — openly and transparently — rather than modifying the underlying incentives and disincentives.
“The important thing in the social investment model is it’s about consent,” said Lee-Archer. “It’s about actually doing this in a way that the people you are trying to help know you are trying to help them, and that these are the types of techniques and interventions we’re going to use.”
In the social services paper, they argue it is important “the person receives the nudge in the context of a service plan towards achieving a desired outcome they have agreed to” and point out long-term welfare clients are often stuck in a “vicious cycle of ongoing disadvantage” where an event like losing a job can lead to a cascade of struggles:
“People can become trapped inside the social security system and despite best intentions and effort they can become socially and economically excluded. A digital nudge, leveraging nudge theory and predictive analytics, provides an opportunity to assist people make better life choices leading to a virtuous circle scenario where a good outcome, such as securing stable housing, delivers the capacity to hold onto a job leading to a pathway towards social and economic independence.”
Gregor and Lee-Archer contend their digital nudges are a “natural progression” from the way organisations are already using big data. They also urge caution:
“While the core technology to apply a digital-nudge approach is maturing, closer consideration is required at a business level so that when a digital nudge is applied, it addresses social disadvantage, rather than perversely contributing to it.
“Unlike the commercial gain motive of the private sector, a digital nudge in social security administration must have a social purpose where the person subjected to the nudge is the one to gain the most benefit from it.”
They comment that even with a government making its best efforts to deliver citizen-centric social services, a “power imbalance” remains whenever one seeks the support and assistance they are entitled to, and argue it’s even more vital in this context to make sure people know when they are being nudged:
“Power rests heavily on the side of the funding body or service provider who has the power to withhold benefits and services while it determines eligibility and entitlement against legislation, rules and administrative processes. Within this power imbalance, adding a digital nudge capability to the armoury of the funding body may represent a further concentration of power against the socially disadvantaged.
“… There is potential for the ethical and moral risks of nudges to be amplified when applied to people experiencing social disadvantage.”
The ethical and privacy concerns, the authors conclude, will “loom large” over social services if, as they propose, governments increase their use of predictive analytics and nudge-style interventions in hopes of delivering support that works better for more people.
They suggest the inclusion of digital nudges in social welfare is ethical if developed within an informed consent model to contribute to mutually agreed outcomes, but that this conclusion is also likely to be challenged.