Digital governance in the age of AI:
evolving policies, regulations, and practices

KEYNOTE SPEAKERS

Title:  The Governance of AI and Big Data: Promises and Pitfalls for Public Policy and Services

This keynote presentation will provide an overview of a range of issues associated with emerging forms of governance surrounding the use of Artificial Intelligence (AI) and Big Data (BD) in public policy and service settings. These digitally mediated practices, which have transferred over from the commercial sector, offer the potential to make public services more efficient, automate public decision-making and provide new insights about societal issues that can be utilised to solve public policy issues. Yet, there are some unique nuanced characteristics of public services and the public service context, which means that their use will always be challenging. This presentation will explore these challenges through a definitional and conceptual assessment of these new technological practices and how they deviate from traditional informational processes in state apparatus. In doing so, the presentation will highlight the importance of different types of ‘data’, the importance of the origins and nature of data, as well as a reliable base of data for AI and BD to function effectively. This in turn has ramifications for administrative, procedural, accountability and oversight norms within the public sector. With this in mind, it is evident that caution is required to accommodate the widespread use of AI and DB in public service settings.

Title:  Determinants of successful adoption and implementation of AI in public organizations

Despite the current popularity of artificial intelligence (AI) and a steady increase in  publications over time, few studies have investigated AI in public contexts. As a result, assumptions about the drivers,  challenges, and impacts of AI in government are far from conclusive. This presentation, which is based on empirical research conducted by the presenter, will focus on the determinants of successful adoption and implementation of AI in public organization. The presentation will distinguish clearly between the drivers of adoption and the determinants of success and will adopt a multidimensional perspective to understand the deployment of AI in public sector organizations. In addition, it will address some organizational routines capable of overcoming adoption and implementation challenges of AI systems in the public sector.

Title:  ‘Artificial Intelligence in the public sector – Hype or hit?’

The topic of Artificial Intelligence (AI) is omni-present these days. Numerous organizations (including public entities) cannot stop talking about AI tools, but it appears that only a limited number report using them. The key question of this keynote presentation is to figure out if AI can be currently considered as a hype or that it has already become mainstream technology (so a hit). If the current AI-adoption can be characterized as a hype, then it is wise that most public organisations do not invest much in AI and leave the investment over to entrepreneurial companies. But, if AI can be characterized as mainstream, then there is not much risk to invest in AI. Based on the scoring on four relevant key evaluation criteria for hype assessment (Expectation, Attention, Involvement of stakeholders, and Benefits), we are able to identify the current situation of AI-adoption in the public sector. The presentation starts with a short introduction about AI including relevant definitions, concepts and principles. The hype/mainstream assessment itself is achieved with the help of all the track participants. The presentation ends with a list of key conclusions.

Title:  ‘The AI and diminished man: the civil servants and citizens in the context of digital governance’

The presentation aims to intersect the reflections from our work L’Homme diminué par l’IA (The Diminished Human by AI) (Bertolucci, 2023), which focuses on AI in all its fields (art, love, democracy, work, etc.), the outcome of our state-of-the-art article on public management research and AI (Bertolucci, 2024), as well as an analysis of scandals and misuses of AI systems in public organizations.

If a civilization needs a religion to take shape, Humanism and the Enlightenment in decline have found their successor in the religion of data, the latest avatar of the sacred transferred to Technique as themed by Jacques Ellul. The question of the exit from the Enlightenment is raised as Kissinger, captivated by the subject of AI from 2015 until his last days, wrote. Let us remember that in What is Enlightenment?, Kant criticized the state of guardianship as the inability to use one's own understanding without the direction of another. Today, is this "other" not AI? The algorithmic accompaniment of life, denounced by philosopher Éric Sadin, is now embodied in Google's project to assign us a personal AI life coach... Our psyches are diminished by the confrontation with algorithms. And we let it happen, while our children now spend more time on recreational screens than at school. Even the French Council of State (Study on Trustworthy AI, August 30, 2022) is concerned about the dependency on algorithmic assistance that could lead to the atrophy of our cognitive abilities. After factory work destroyed bodies, screen work is destroying minds.

From the perspective of public management, we will propose to question the already documented risk of the emergence of what we call a civil servant diminished by AI, dependent on this imposed and inexplicable system, towards which, agents divest themselves of their professional judgment. Drawing on two of the biggest scandals of reckless use of AI in the public sector, Giest and Klievink (2022) demonstrate that the implementation of AI is not without friction, comparing the cases of the State of Michigan's employment agency with the MiDAS AI system and the Toeslagen agency for the childcare allowance system in the Netherlands. In both instances, serious repeated errors over years led to political and media scandals but, above all, to the devastation of the lives of the less fortunate. The Dutch government had to resign in 2021 following revelations about the 26,000 families who were indebted to the point that a thousand children were removed from their families. The Michigan agency, for its part, dismissed a third of its staff due to the complete automation of its fraud verification process with the MIDAS system (2013-2015), whose 93% error rate did not prevent it from operating for more than 2 years. In total, its implementation, reimbursements, and legal fees cost over 80 million dollars, which did not prevent the State of Michigan from hiring Deloitte in 2023 to implement another AI system for a cost of 78 million dollars.

On the level of agents, laboratory research on decision-making shows that individuals are less conscientious when assisted by AI and make more mistakes without it being visible from the outside (Cymek, Truckenbrodt & Onnasch, 2023), and that biased AI influences individuals' judgments even when it is withdrawn and, worse still, biases the decision of individuals who have nevertheless gained experience in a task by themselves (Vicente & Matute, 2023). Numerous empirical cases of AI systems in the public sector corroborate these findings. Unfortunately, what many surveys show, coming from associations (e.g., La Quadrature du Net in France) and journalists (e.g., Disclose in France), is that dozens of public actors in Europe misuse discriminatory and biased algorithms. We advocate a definition of AI in the public sector that embraces its integration at the societal and organizational level, taking into account the risks to citizens and the environment:

AI is a technical apparatus that translates its environment into data in order to infer structures and/or generate content that holds meaning, based on a predetermined objective. Within a public organization, this objective must be pertinent from a citizen, societal, and environmental perspective, while ensuring that oversight is possible by both organizations and society at large.

Title:  AI regulation in the European union: Taming the tiger through dialogues and trialogues

In principle, the digital age is characterized by the ACQU determination: Access to data, Control of data, Quality of data, and Use of data. In the digital arena, the traditional states are no longer competing amongst each other alone, but are joined by non-state, private actors whose power and influence often overshadows many contemporary nations. To that end, the regulation of digital arena is becoming increasingly important for the security and well-being of the society at large in the global context.

The upcoming European Act on Artificial Intelligence deals with a number of fundamental security threats to the society at large and fundamental European values, including the protection of human rights and the right to privacy. The Act, in its many iterations and amendments in a relative short legislative period, address a plethora of sensitive subjects such as the use of automated decision-making utilizing sensitive personal data, biometrical identification in real-time and post-remote operations of law enforcement bodies, use of generative AI tools to create deep fakes and enhance disinformation campaigns, utilization of various scoring systems powered by algorithms, and others. The European Commission, European Parliament and the Council have recently reached a political agreement concerning a number of items that were heavily debated during the last several years. The presentation will reflect some of the items on discussion agenda during the noted trialogue.