Artificial intelligence in the public sector: an engine for innovation in government… if we get it right

Carlos Santiso, head of digital, innovative and open government in the OECD’s public governance directorate, shares his thoughts on promoting the effective, responsible and ethical use of AI in the public sector

This summer, I had the privilege of participating in the conference, ‘Artificial Intelligence: What’s in it for the Public Sector?’, organised in Brussels by the European Commission, the Interoperability Unit of DG DIGIT and the Joint Research Centre. It brought together policymakers and experts to discuss the opportunities that artificial intelligence (AI) brings to accelerate governments’ digital transformation and improve the design and delivery of public services.

The event was a great opportunity for me to reflect back on the work on AI in the public sector that I led at CAF – the Development Bank of Latin America – as well as the work being done by my new team at the OECD, which has been breaking new ground on this subject since 2019 when it released two pioneering reports: Hello, World: Artificial Intelligence and its use in the public sector and The State of the art in the use of emerging technologies in the public sector. It was also a pleasure to discuss this topic with my follow panellists Hilde Hardeman and María Pérez Naranjo, along with our moderator Sven Schade.

The current debate on AI tends to focus on its potential for the digital economy, the importance of regulation that is fit-for-purpose, and the role of governments as regulators. Indeed, governments play a key role in making AI work for innovation, growth, and public value. It is common to think primarily of their role as conveners, building agreement around strategic directions and commitments to foster AI; financiers, providing funding to support research, development and adoption, including through investment funds in AI-driven start-ups; or regulators, ensuring policy coherence nationally with underlying values of society and internationally through regulatory cooperation. However, governments are also prime users of AI to improve the functioning of the machinery of government and the delivery of public service. The focus of our work in the OECD public governance directorate is precisely on the role of governments as direct users and service providers of AI, where governments can and are unleashing many benefits, not just for the internal business of government, but for society as a whole. AI is, as a game changer for governments, likely to reset these.

Read more: Making artificial intelligence fit for a life inside government

As users, governments are increasingly incorporating AI into their arsenal of tools to improve efficiency, through the automation of physical and digital tasks; to increase effectiveness by being able to take better policy decisions and deliver better outcomes, thanks to improved predictive capabilities; and to increase responsiveness to user needs, delivering personalised and human-centred services through the design of human-centric interfaces.

Leveraging AI as a tool for better policies, services and decisions requires governments to think strategically, get the balance right between opportunities and risks and think outside the public sector box to innovate and collaborate with new players.

As I touched on at the conference, there are a few issues in particular that I believe rise to the top for promoting the effective, responsible and ethical use of AI in the public sector.

Think strategically: governance matters

Ad-hoc experiments and initiatives can be great sources of learning, but governments need to have a strategy in place for AI specifically designed for and adapted to the public sector to ensure efforts are scalable. As identified by the OECD.AI Policy Observatory, more than 60 countries have already created a national AI strategy. Importantly, most of these include a dedicated strategy focused on public sector transformation.

Read more: OECD AI policy unit can help bolster national standards, report finds

However, not all strategies are created equal. Some are more like a list of principles or a checklist of projects. To really drive AI in the public sector, they need to be actionable. As discussed in the recent OECD-CAF report on The Strategic and Responsible Use of AI in the Public Sector of LAC, Colombia’s strategy for AI in the public sector is a great example of an actionable strategy driven from the very centre of government. It’s one of a few that includes all the elements for systemic exploration, adoption, and evaluation of AI. It specifically lays out:

  • Objectives and specific actions
  • Measurable goals
  • Responsible actors
  • Time frames
  • Funding mechanisms
  • Monitoring instruments (a rarity in AI strategies)

Having these items in place helps to ensure sustainability and follow-through on commitments and pledges, one of the main challenges we have seen with AI in the public sector. This, of course, will need to go in hand with strong and vocal leadership so that AI in the public sector gains traction. This could take the form of committed leaders making it a priority, but also through new institutional configurations in charge of AI coordination. For example, Spain created a secretary of state for digitalisation and artificial intelligence and the UAE created a minister of state for artificial intelligence. Governing AI is also about making sure that it works for society and democracy. Getting the right balance between its risks and benefits will mean putting citizens at the centre and ensuring that underlying data is inclusive and representative of the society to avoid bias and exclusion. It will also mean setting actionable standards that are fit-for-purpose for the public sector to implement the high-level policy principles being adopted by governments. The OECD AI Principles and the OECD Good Practice Principles for Data Ethics in the Public Sector are two key tools to accompany governments in such efforts. These aspects should be built into the design of public sector strategies from the onset, and they should be leveraged to inform the decisions of leaders and practitioners alike.

Accountability: bring in the auditors and work together on new accountability structures

Carlos Santiso

As underscored in the OECD AI Principles and, more recently, the UNESCO recommendation on the ethics of artificial intelligence, accountability is a critical element for the responsible use of AI in the public sector. While often compared to AI’s use in the private sector, governments face a higher bar in terms of transparency and accountability. Government adoption of AI grows rapidly, as does the public sector’s need to ensure it is being used in a responsible, trustworthy, and ethical manner. 

Key to achieving this within governments are actors in the accountability ecosystem, including ombudspersons, national auditing institutions, internal auditors within organisations and policy-making bodies that establish rules and guide agencies that design and implement AI systems. Often, these different groups have had an uneasy relationship, with limited collaboration, often as part of a broader debate on the right balance between innovation and regulation. However, emerging technologies and governance methods, as well as the increasing wealth of available data, are resulting in a re-thinking about traditional public accountability and the roles of these actors.

Read more: Canada to create top official to police artificial intelligence under new data law

Within the OECD Public Governance Directorate (GOV), we believe that tremendous potential can be unlocked by forging new paths to collaboration and experimentation within the accountability ecosystem, with all actors working together to achieve the collective goal of accountable AI in the public sector.

While there are numerous areas begging for an injection of innovation in how oversight is carried out, one of the most important policy arenas governments are seeking to achieve this through is algorithmic accountability. There is intense debate on how to ensure that the development and deployment of algorithms in the public sector are transparent and accountable. This concern is central to the EU’s AI Act, which is currently under discussion. 

Ensuring that those that build, procure and use algorithms are eventually answerable for their impacts.

– Algorithmic accountability, as defined by the Ada Lovelace Institute, AI Now Institute, and Open Government Partnership.

Some progress is already being made. For example:

However, it’s clear that more efforts are needed in this area. For instance, through applying its framework, The Netherlands Court of Auditors found that six out of nine audited algorithms did not meet basic requirements, exposing the government to major security and bias risks.

Strengthening capacities with the ‘three Ps’: people, partnerships, procurement