Decentralisation has been dubbed the quiet revolution. Between 1995 and 2016, in two thirds of OECD countries, decentralisation processes have resulted in an increase in the powers and resources available to subnational governments, in the quest to capture a “devolution dividend” – better outcomes that result from using local information to improve policies. Yet will the enormous processing power of AI enable central governments to match – or even surpass – the information advantage of subnational governments? Can it even replace local decision-making?
Information is power – but for who?
AI can aggregate and analyse huge amounts of diverse data to provide comprehensive insights into local needs. This can help central governments understand and directly respond to the circumstances of each region, delivering better outcomes and service standards. AI’s machine learning capabilities can go further still, and be used to predict future trends and resource needs. For example, in India AI has been piloted for crowd management at the Kumbh Mela religious event, while Sweden’s Public Employment Service has been using AI to improve job matching.
VC Investments in AI for industry
On the other hand, subnational governments are also well positioned to deploy such analysis and predictions. Subnational governments often hold key powers over spending, taxation, infrastructure, economic policies, land use and the environment. In some countries, municipalities and regions are also responsible for education, social services and health care. In all these areas, subnational governments collect different types of data to reflect different local responsibilities and priorities. Local partners can calibrate AI tools to help them meet those challenges, based on the specific data available and the problems they prioritise. For instance, in Pittsburgh, US, the Surtrac system combines artificial intelligence and traffic theory to optimise traffic control. In waste management, cities in Spain, Denmark and Norway have started to use smart sensors and AI to optimise waste collection, leading to considerable efficiency improvements.
AI could also help remove barriers to further decentralisation, which include a lack of local capacity and higher administration costs. The smallest and economically weakest subnational governments tend to suffer most from these issues, meaning that there is often a reluctance to decentralise funds and powers where they could make the most difference to tackling spatial disparities. Here the potential of AI could be transformative, supporting such administrations to design smarter, more responsive policies to turn around their fortunes.
Trusting the black box
Even if AI could substitute for, or improve upon local knowledge, it may lack a crucial element to succeed: legitimacy. Public policies stand or fall on the basis of popular consent. Decisions guided by AI can be difficult to unpick and justify to local communities – particularly when there are limited resources and tough decisions to be made. Legitimacy is also crucial to mobilise efforts and additional resources to support key priorities. That includes efforts to develop effective local partnerships between government and the private or third sector to tackle climate change, support the vulnerable, or align infrastructure investment.
Fundamentally, people may resent AI making decisions for them, and examples of clashes between the civil society and AI have already emerged. In Canada, the City of Toronto´s Waterfront Project plan to develop a “smart neighborhood” provoked extensive public debates about the role of AI and data privacy in urban development and public policy planning that eventually derailed the project. In Boston, US, a geospatial algorithm was used to redesign the school bus system for public schools, leading to considerable cost savings and reduced journey times. But the solution meant new school start times, upsetting parents’ routines and leading them to establish a strong pressure group in opposition. Finally, the plan was altered, leading to less overall benefits. These examples highlight the importance of building in transparency and making sure that AI solutions work with local communities as well as for them.
Despite its promising potential, using AI effectively at any level of government will not be easy. There remain thorny issues of security, privacy, transparency, and trust in addition to the difficulties of competing with the private sector for key digital skills. Governments at central and local levels need to rapidly develop their knowledge and expertise. That is why, for example, Sweden is currently training and educating their public administrators to use AI in their work.
The prize of better decisions will be worth the investment. AI provides the potential to deliver smarter, faster, and better policies, leading to improved welfare of citizens. But ultimately local partners and communities must be in the driving seat to grant it the legitimacy and support it needs to fulfil that promise.
Read more on the OECD work on Decentralisation.
Antti Moisio is Economist and Senior Policy Analyst in CFE/RDG. Before coming to the OECD, he worked in Finnish Prime Minister’s office as Senior Advisor. His duties in the ministry were primarily related to better regulation and legislative impact assessments. Prior to that, Antti Moisio worked for more than two years at CFE/ESG as a Policy Analyst, mainly in projects related to decentralisation and multi-level governance. Antti has also worked several years as Research Director and Principal Economist in VATT Institute for Economic Research in Helsinki. Antti has publications on local public finance, efficiency of public services, municipal mergers and political economy. Antti has taught courses on applied econometrics and local public finance in several universities, including University of Jyväskylä, Université de Rennes 1 and Université de Fribourg. Antti holds a PhD in Economics from University of Jyväskylä, Finland.