Prompted by the Economist’s latest Bartleby column, on the relationship between humans and AI, I went back and reread something I had written over five years ago.
I was still in the Australian Public Service, then, in the Digital Transformation Agency. I had attended the 2017 end-of-year address by the Secretary of PM&C, Dr Martin Parkinson, to the Institute of Public Administration Australia (IPAA). Martin spoke of the need for ‘bold and imaginative thinking’ to cope with the changing environment, particularly in response to technology.
Given that, I was looking forward to the thoughts of the subsequent panel, particularly when they were asked what disrupted APS looked like. It fell a bit flat. To be fair, they’d been speaking about diversity, and so Chris Moraitis, who was left with the question, picked up on that theme. And in reality, this was a panel that probably was not going to engage with difficult ideas on the fly in such an open forum.
So I went home and wrote up six possible scenarios that night, broadly described including from the perspective of an APS officer. I wanted to illustrate that the future, using technology, was not predetermined–and that there were implications for how the APS operated.
What did I do with the piece? I sent it to some at the top of the service, with no response. The APS was not prepared to deal with speculative futures–and at the time, it was being ground down by the government. A bit over eighteen months later, PM Morrison delivered what I refer to as the ‘downstairs staff’ speech, with the APS to be seen, not heard, simply to do what it was told, firmly put in its place. And we can see how that ended.
Anyway back to the six scenarios. I’ve added the piece below, slightly edited for online readability. The perspective of the APS officer is italicised. Bear in mind that this was written in the world in 2017, well before ChatGPT.
Six scenarios; it’s also long. I’ve labelled the scenarios as ‘enabled APS’, ‘enabled citizens’, ‘government as broker’, ‘panopticon’, ‘packaged services’, and ‘disaggregated flow’. The last doesn’t have a distinct APS officer perspective, deliberately.
Disruption will be driven by a few critical underlying trends.
Remember the days when information was limited, and the British Encyclopedia was a valued resource, and Lexus/Nexus was how we searched for information in university libraries? Remember when overseas calls cost $50-150?
We are so used to infinite information available to us—on our phone, no less, and that phone is used to make not simply free international calls, but free video calls to anywhere in the world.
Beyond infinite information—and cheap communications (though that varies by country, market conditions and infrastructure)—we are moving to a world now of infinite compute and infinite storage. What would we do in such a world? What would that look like in 10-20 years?
For the APS, at a minimum, we need to think in terms of flows and streams, not blocks—block departments, block budget cycles, block budget bids, block ICT systems that sit in a corner, even a ‘block’ FTE (full-time equivalent staff member).
We need to think in terms of outcomes and investments, not inputs and expenditure—which is how our budget and finance framework works—and be held to account for those outcomes.
Imagine.
Enabled APS
Imagine a world in which there were no constraints on compute, on storage, on your ability to pull down data and run scenarios of policy and outcomes, using your choice of a tool set provided by government—though if you wanted cutting edge tools, you subscribed to a university consortium that provided tools in beta.
That information enables you, as a citizen, to shape the services you want, based on your willingness to pay (via what used to be known as ‘tax’). The exchange is undertaken in near-real time, and is not as expensive as may be thought, given much now operates as-a-service or as part of the sharing economy. Even in remote areas, thanks to increasingly sophisticated 3D printing and cheap energy, products and services can be offered reasonably cheaply and quickly.
In the APS…you (the APS officer) may be embedded in a local community, perhaps as part of a government node (GovNode), able to draw on material, advice, interaction with fellow embeds in other communities, and data to help shape outcome, able to respond to immediate community needs.
That community could be in the South Pacific, where you split your time between local aid-related work, and supporting government international economic policy, supporting colleagues in the Canberra and the World Bank.
Your ‘home team’ is a dispersed, cross-disciplinary and largely self-managing group, operating as freelancer policy journeymen and typically centred on a broad policy problem—what used to be referred to a department or agency.
When a problem arises that you can’t solve through a 3D model, scenarios, on-the-ground analysis, or requires either more subject matter depth or highly specific skills you lack, you can forward the problem to a ‘master’ policy agent.
Prioritisation can be an issue, but you can check the relevant masters’ stats on line, including responsiveness, innovation, breadth, strategic vision, etc, which helps, as does the information about the other masters or journeyman in their own network.
Enabled citizens
Imagine a world in which each and every citizen controlled the data that government held on them.
Your data is held in perpetuity in a GovVault—a network of secured, resilient, duplicated data centres—and only you can enable who accessed that data under what circumstances, using your personal, encrypted digital signature, provided to you when you turn 18, and your personal key.
You can check any time and see who is accessing that data, for what purpose. You may choose to offer your data, anonymized, to improve public health, for example—and you receive recompense for doing so.
You may choose to include a few personal mementos in your VaultBox, so that you can pass those on to your children when you die—and you put in place a plan for the management of your data on the interface that can be access on your death, changing the default option everyone receives on receiving your encrypted signature.
In the APS….Your task is to implement a new health policy by changing a rules engine. But first, you need to run a series of scenarios to illustrate and test the changes.
You check that available citizen data meets the parameters of the test; it doesn’t have enough diversity in the targeted age group, so you ‘ping’ citizens asking for permission to access their data, suitably anonymized, for a set fee, for a set timeframe and agreed purpose.
When enough responses have been reached, you run the scenarios, then adjust rules and policy parameters, which include the result of the tests you have run, sunset clauses, and desired outcomes so they can be measured against at regular intervals.
You are a little worried, though, as a team nearby has been testing an AI that will shadow their person, and negotiate these calls on personal data autonomously.
Your concerns are around the security of the AI and the integrity of their decisions to release, or not release, their citizen’s data.
Further, you feel that citizens need to be actively engaged in those decisions—government and its decisions should be close and considered by government, not outsourced to (yet another) piece of technology.
The counter-arguments are precisely the opposite—frictionless service and the invisibility of government machinery. You think about raising at the next concerned citizens online forum, which, ironically enough, tends to be monitored by people’s media AIs to alert them when an issues of interest is raised, allowing them to jump online to participate.
Government as broker
Imagine a world in which government existed merely to broker services to its citizens, with a manifesto of citizens rights and economic freedoms (both resting as the basis for national security) as the fulcrum of the exchange.
In such a world, government collects information on citizen needs—typically by region, group or cohort—and organizes what used to be referred to as a ‘coordinated procurement’ of services, provided by the private or not-for-profit sectors, which it made available to its citizens.
Other services are auctioned, with sellers and bidders working in an online exchange are public, with auctions held every week or quarter, depending on the nature of services. Some are media events, such as those bidding to be part of a major art work; others are held closer, with merchant banks and mercenaries in consortia to offer military services. Some are more local, with local or regional communities bidding for—or offering—services and products, while others start kickstarter campaigns.
In the APS…Most public sector functions have been devolved to regions, and your task is to manage the bidding platforms.
In early days, regions, community groups, individuals had to be provided with tools to bid, but then the open source community and private sector companies open to opportunity started offering better options, so the government effort was closed down (placed aside some of the innovation outside, government efforts were a tad embarrassing).
You use advanced analytics to watch for gaming of the system, and your team came up recently with a policy chance, realized through rules engines and implemented across a number of sector platforms, that forces a small delay between bidding and acceptance by consortia.
Your analytics team had noted a trend wherein certain consortia where using some machine learning to pre-empt bids. As outcomes aren’t clear—they may be beneficial for the communities or individuals involved, and may depend on the specific services concerned—you felt it better not to exclude those consortia but to level the playing field.
In the meantime, your team is building a case base on which to establish some better rules for your own AIs, drawing on university research as well as experience on the ground, and focusing on local economic conditions.
Panopticon
Imagine a world in which government controlled all citizens and private sector data, cross referencing it to personality, real-time citizen movement, and economic activity, such that it controlled absolutely society and the economy.
Government builds model of human and economic behavior and uses those to set economic settings, taxes, laws, regulations, employment, education, health. Efficiency is prioritized. Outliers are easily identified and ‘managed’.
Companies providing surveillance, data, machine learning and ‘response’ technologies flock to Australia.
The models, based on sentiment analysis, establish voting patterns—and the language and initiatives most likely to shape voter outcomes.
Further, the models would be able to tell if you deviated in any activity or characteristic by more than a standard deviation from a ‘norm’. When that occurs, government automatically activate ‘remedial’ activities, whether in health, public safety, spending patterns, internet activity, education, or income.
In the APS…The government’s mantra is for a secure, efficient Australia—and technology is the means by which the government will ensure that its citizens pose no threat, are not threatened—and through greater efficiency in all aspects of their lives, they will be able to ensure that occurs at the lowest cost to government.
Infinite compute and storage enable 24/7, ubiquitous surveillance and assessment. Your own work is actively monitored—the government is alert to the insider threat, as well. Predictive analytics that focus on behaviours considered to be best aligned with government purpose enable early identification of problems.
Government decision-makers receive weekly reports on social and economic activity; a dashboard available to key decision-makers within agencies monitor in near-real time productivity and delivery within the APS.
Because behaviours have been ‘normalised’, social agencies have been stripped to a core delivery platform. Outliers are directed to the new healing and educative apparatus for special attention; that agency devises programs to ensure citizens can make their full contribution to society and the economy.
‘Packaged’ services
Imagine a world in which you paid consortia for public services, as they proved more responsive to your needs that the basic ‘pleb’ package offered by government.
If you wanted you could upgrade to higher levels of service, though that may come at an additional cost, for switching providers. You used a recommender—a personal AI tuned to your own interests and needs—to work out the best package.
The basic ‘pleb’ package comes at a fixed cost. You figure this is probably used to support other services, such as defence, as the pleb package is less than subsistence and what used to be known as a ‘poor house’—and a basic smart phone that is purely there to access government services and minimal news.
Consortia tend to include a mix of state and local services, so you can get services, if needed, relevant to your area.
That also means that some services can only be accessed by buying into that consortia—for example, the Sydney Easter Show, AFL Grand Final, Burke Races and Gay Mardi Gras are each offered exclusively to specific packages. The pleb package doesn’t have such options (except for Australia Day).
In the APS, you are an account manager for health services—a senior position. That means your teams task is to look for consortia to meet government’s desired outcomes (you still remember the days, sometimes nostalgically, when these were described as policies).
Desired outcomes tend to shift in response to public opinion—and so you keep a close eye on sentiment analysis, not least as sentiment forms part of the performance measures on which your remuneration is based—and health consortia offerings.
But your base ‘map’ is the state of health across the nation, and in your case, focusing on cardio-system outcomes. The map is the result of internet of things collection, near-real time feeds from hospitals, clinics, fitness devices, and gyms, meshed with surveys, supermarket data, ‘city’ data (for example, public transport) and social media.
That data is also available, at a cost, to the private sector, and on that basis, corsortia offer packages to the public and companies who use those packages to attract or reward workers or clients, or on-sell.
You also track the take up those packages and their effects on community health and public opinion.
Disaggregated flow
Imagine a world in which all is a constant flow or stream of information. You can choose to dip into it when you want—for news, education, income, sustenance—limited only by your own personal constraints—attention, time, capability, earning potential, ability to pay.
Because much of the work you had once done—or expected to do—is now automated, you spend more time learning new skills, through online courses, personal tutors (hired online), or co-working as an apprentice or journeyman.
You keep an eye out on various job-boards to which you subscribe—your rating is fairly high, as you’ve done well in earlier stints, and you are confident that the AI you’ve been training on will help you improve some of the design problems you know that Defence is having with fuel storage (a perpetual issue).
If that doesn’t work out, the same smarts may help with energy futures, to help the Energy Directorate. That use to be a Department, once, but with staff dispersed into the regions, devops teams, tiger teams, it works better as an adaptable ecosystem solving problems as they arise.
As resourcing shifted to on-demand and capability to as-a-service, there was less need for large corporate areas. Then Ministers started identifying issues to be solved outside what used to be budget cycles, and given pressures, the natural response was to ‘break and reform’.
But the reforming didn’t go as expected—there was little call to return to old hierarchies. And now some quite junior staff were making considerably more money—and gaining more influence—than older staff.
There remain questions about ensuring a holistic and strategic approach to issues—we remain a nation after all—and we still need to ensure capability and underlying coordination and data is managed and protected. But few miss the old ways.
Summary
All these are not impossible, given the technology available to us, in the near future. Typically, it’s not the technology that’s the issue: it’s people and processes. The same applies to the APS.
Each of these scenarios demand a fundamental change to the APS. In some, government effectively dissolves. In others, a technologically equipped bureaucratic machine is absolute. All bring into question the role of government—and the contract between government and citizens.
Each demands we think hard about what it is to be in a liberal, democratic, market-based capitalist society that is being disrupted by technology and by social upheaval.
These are the matters we should be considering.
Lesley Seebeck 