Yet compared with many private sector industries, it鈥檚 fair to say that public sector adoption of AI technology has been more measured. Governments and other public sector organisations face a number of significant challenges, from the availability of skills and investment funding, to demonstrating value and ensuring transparency about how decisions are made.

These challenges are reflected in the 51风流Institute for Digital Government鈥檚 latest report 鈥撀犅犫�� developed in partnership with the University of Queensland. While 80 per cent of public sector organisations are actively working towards data-driven transformation, fewer than 15 per cent have progressed beyond prototypes.

In order to drive greater uptake, the public sector needs to develop best practice frameworks and solutions for the development and use of AI systems that are accurate, robust, and scalable, but also reliable, fair, and transparent.

When building AI systems to meet these high levels of expectation, it鈥檚 vital that public sector workers are able to understand how these systems generate decisions and explain how this impacts results. This is known as AI explainability.

Read more of the article on Government News .

The post Opinion: Building explainability into AI projects appeared first on 51风流Australia & New Zealand News Center.

Right now, in our communities, we are witnessing the consequences of聽limited rationality,听social preferences聽补苍诲听lack of self-control. In their seminal work 鈥淣udge: Improving Decisions on Health, Wealth, and Happiness,鈥� Professors Richard Thaler () and Cass Sunstein postulated that these human traits systematically affect individual decisions and market outcomes. It鈥檚 instructive to explore how these factors might be influencing individual decisions, for example, to stockpile toilet paper:

• Limited rationality: People focus on the narrow impact of individual decisions rather than the overall effect. For example, I鈥檒l buy some extra toilet paper now because I鈥檝e heard that it might be in short supply later. I make this individual decision without realising that I鈥檓 inadvertently contributing to the overall effect of supplies running short, which will ultimately impact me 鈥� along with everyone else 鈥� in the long run.

• Social preferences: People have a social preference for equitable outcomes. For example, I鈥檒l be less accepting of my local supermarket increasing the price of toilet paper in response to a growth in demand than in response to a rise in their cost of supply. Even if the price rise is the same in both cases, my willingness to pay a premium is influenced by my perception of fairness.

• Lack of self-control: People tend to give in to short-term temptation rather than stick to a long-term plan. For example, even though I have more than enough toilet paper at home, I鈥檒l still buy more if I find it somewhere on sale. I know that I don鈥檛 have anywhere to store additional rolls of toilet paper, but when presented with the opportunity to purchase such a sought-after item at a discounted price, I won鈥檛 be able to resist.

As has been demonstrated across the globe, government assurances, pleas, and directives have failed to prevent emotional shoppers from emptying shelves in anticipation of future shortages. Now similar assurances, pleas, and directives are being made in relation to the much more serious issues of self-isolation, social distancing, and personal hygiene. Will citizens heed government rules and regulations now when they haven鈥檛 in the past? Certainly, the Chinese government聽聽in curbing the spread of COVID-19, but most Democratic governments don鈥檛 have the same controls available to them as in Communist China. What then is to be done?

In our aforementioned research, the SIDG and the ANU described how聽digital nudging might be used by governments to drive behavioural change for social good. Empirical evidence told us that certain human actions result in better social outcomes, and digital technology is enabling us to reliably predict those outcomes based on observed behaviours. This caused us to ask: how might we leverage default human nature to positively influence social outcomes, and could we apply technology to influence individual decisions at scale?

Where Thaler and Sunstein (2008) defined a聽nudge as: 鈥淎ny aspect of the choice architecture that alters people鈥檚 behaviour, in a predictable way, without forbidding any options, or significantly changing their economic consequences.鈥� We defined a聽digital nudge聽as: 鈥淚ndividually targeted processes, facilitated by information technology, to achieve social policy outcomes鈥� (Gregor & Lee-Archer, 2016).

Figure 1: At the intersection of agile policy, information technology and behavioural聽science is the digital nudge.

Moreover, we proposed that聽predictive analytics听补苍诲 contextualisation聽capabilities can improve the effectiveness of traditional nudging by enabling the shift from reactive to proactive interventions and by making nudges more targeted to individual circumstances.

• Predictive analytics is a specific field of data mining in which large stores of data are analysed to detect patterns and to predict future outcomes and trends. While predictive algorithms have been used for many years, they have typically been restricted to operating on pre-existing data. Real-time computing platforms have changed this by allowing data to be analysed as it鈥檚 created. This means that analytical discoveries can be applied to adjust government action dynamically, thereby influencing trends as they emerge.

• Contextualisation聽is the next evolution of personalisation: blending together information about past interactions and anticipated behaviours with present motivations and intent. Where personalisation attempts to anticipate future behaviours based on past activities, it lacks the in-the-moment context of the citizen鈥檚 current circumstance. This is important because it鈥檚 precisely that current context that鈥檚 most relevant and useful for predicting future behaviour.

Figure 2:聽Our framework for the design and application of digital nudges.

Of course, our thinking has evolved since 2016, and so we would now add聽experience management聽into the mix.

• Experience management聽brings together operational data (O-data) about聽what聽is happening, with experience data (X-data) that tells us聽why聽it鈥檚 happening. This fusing of X+O data can enable governments to better understand citizen sentiments and motivations, and thereby take effective action. Importantly, since sentiments and motivations are constantly changing, governments need to embed feedback and analysis throughout their business processes and at every point of citizen interaction.

With this in mind, let鈥檚 return to our example of stockpiling toilet paper and see how governments might apply digital nudging to curb this behaviour鈥�

An online聽聽suggests that to last 14 days in isolation, each person requires only four rolls of toilet paper. So, the average American household (2.6 people) should be able to get by with just a single pack (10 rolls). Most likely, very few consumers did this calculation prior to purchasing, so a simple SMS informing citizens about how much toilet paper they actually need could be quite effective. It might even be possible to target the digital nudge by advising the required number of rolls for a given household.

Another approach would be to leverage the behavioural science influencer of .听聽of over 6,000 Australians indicated that only 9% had purchased more than 20 rolls of toilet paper due to COVID-19. This sort of statistic could be promoted via digital channels, especially in geographic areas where a small percentage of people have been observed to be buying in bulk. To further improve effectiveness, the poll could be extended to understand what鈥檚 motivating consumer purchasing decisions (e.g.,听Why聽did you decide to purchase X rolls of toilet paper?).

Figure 3:聽A conceptual architecture for digital nudges.

These same capabilities could be applied by governments to nudge citizens towards cooperation with rules and regulations relating to self-isolation, social distancing, and personal hygiene. The Behavioural Insights Team鈥檚 provides nine of the most robust (non-coercive) influences on human behaviour, including:

• Messenger:聽We are heavily influenced by who communicates information.听 suggests that 鈥淪cientists and physicians are the most trusted authorities [on COVID-19], along with officials from the World Health Organisation and the U.S. Centre for Disease Control.鈥�

• Norms:聽We are strongly influenced by what others do. Governments, researchers, public health authorities, and the general public are聽聽successful responses to COVID-19 and to avoid repeating the missteps of others.

• Affect:聽Our emotional associations can powerfully shape our actions. The CDC has dedicated聽聽to managing anxiety and stress related to COVID-19.

Finally, it鈥檚 important to be mindful of the iterative nature of our digital nudge framework. Under normal circumstances, nudges are tested with focus groups in聽. While there鈥檚 a need to change certain behaviours relating to COVID-19 immediately, the potential for unintended consequences is heightened as a result of panic, so it鈥檚 important not to skip this important step. 聽approaches can assist in expediting the test-and-improve cycle, both prior to disseminating the initial nudge and to inform adaptation of the nudge as circumstances change.

While digital nudging is not a silver bullet for containing COVID-19, it is part of the overall toolkit available to governments today. As we鈥檝e shown by way of examples, digital technologies can be used to both scale and personalise traditional nudges to improve outcomes for mass cohorts. Specifically, the combination of predictive analytics, experience management, and contextualisation capabilities can enable governments to predict social outcomes, understand what鈥檚 motivating those outcomes, and take effective action to avoid today鈥檚 emerging trends from becoming tomorrow鈥檚 next crisis.

This article first appeared on

The post Dealing With Disruption: A Digital Nudge appeared first on 51风流Australia & New Zealand News Center.