No one will win the AI race if we don’t take our time

AI is transforming society. It has been presented to us as a technology with the promise of endless efficiency. But at what cost? Especially when it comes to government services, how risky and impactful is it when we place too much value on efficiency? Cecília Barreto de Almeida shares her work on the alignment and misalignment of AI with public, individual, organizational, and professional values.

Cecília is a socio-legal researcher and lecturer working at the intersection of law, ethics, public policy, and digital governance. Her work combines legal analysis with empirical research. “In my postdoctoral research at Radboud University, I focus on conflicting values in the digitalisation processes within government. The project examines value alignment and misalignment between AI systems, organisational goals, and employees’ values in the workplace.” She got this role after obtaining a double PhD degree for comparative research on Inequality and Inclusion in Online Public Services: Digital Government Policies and Digital Divides in Brazil and the Netherlands, conducted at the University of São Paulo and the University of Groningen.

The disruptive nature of AI

Needless to say, Cecília is a passionate researcher on the topic of inclusion and digital society, in which AI plays an increasingly important role. She shares the conflict AI might create with well-established values. “People and societies have some shared interpretations of values like legality, impersonality, equity and equality, morality, publicity, legitimacy and efficiency. When it comes to efficiency of government, we expect elected representatives of a country or region to be as efficient with our tax money as possible.”

Recently however, efficiency has taken on a new meaning. “The recent disruption with AI is unlike anything we’ve seen before. It’s very different from other technologies. The rate at which AI is transforming everything is very fast. It has been sold with the idea that AI can solve everything. But we can’t possibly know that. What’s also different with AI is that it mimics human behaviour. It’s becoming harder to distinguish reality and something that was produced by generative AI.”

The extreme disruptive nature of AI can have significant impact on our value-system. Traditional values might clash, mutate or be transformed in unexpected ways, changing their meaning. “A value that once had a specific meaning in the public domain could change, or its weight and relevance in the public debate may change. Previously considered important values might become less important than the ones promised by AI implementation. For example, valuing morality could become less important than the efficiency an AI system could potentially bring.”

Respecting well-established values

But Cecília stresses these values exist for a reason. They have their own integrity and have an important role to play. They accommodate different interests, priorities and ways of operating as a society and as people. They reflect the needs and desires of people. The ease with which some of those well-thought-out values can be disregarded, changed or decrease in importance, is one of Cecília’s worries when it comes to misalignment with AI. “The risk is also that we may think everything is still the same, but that reality tells a different story.”

Things get even more complicated for governments and their workers. “When you decide to apply to a job as a civil servant, you’re driven to contribute to society. You followed all the steps needed to help you get the job you’re applying for. But now, all of a sudden, you’re also expected to have sufficient knowledge of AI, being able to flag risks and other implications of the technology in your line of work, or you may simply be afraid of AI replacing you. Now, you might become the ‘human-in-the-loop’ in an AI-oriented system to deliver public services.” These implications can change people’s relationship with their work. It can cause misalignment in people’s motivations to work at government organisations. Professional values may start to clash with values highlighted by AI implementation.

AI literacy for civil servants

A way to address these issues comes in the form of AI literacy, an idea pushed in for instance the EU AI Act to help civil servants work with AI. The Act states: “Providers and deployers of AI systems shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used.”

But this raises two concerns, says Cecília. “One concern is that this is a lot to ask from people. Asking staff to keep up with everything surrounding AI and the implications for their jobs is a big responsibility. The AI landscape is constantly changing, it’s near impossible to keep up. Asking employees to have a general understanding of AI, how it’s being used, the role of the organisation and awareness about legal and ethical aspects is a lot. You would almost need a new bachelor’s or master’s degree in AI.”

The other concern Cecília points out is that one could be sceptical about whether AI literacy is the answer. “It’s one of the possible solutions to reduce risks and negative impact of AI. But if you look at the broader AI context, it makes you wonder if AI literacy is enough to prepare civil servants for AI’s impact on public values and on their jobs. Can we convincingly tell when efficiency becomes more important than impersonality for instance? And if we offer an AI solution that’s impersonal, can we still strive for equity, where different needs are met with different solutions? People need to be considered in their differences. Can AI do that, and can we be AI literate enough to tell if an AI indeed can?”

Losing the ability to be critical

Cecília’s argument is just one example of how AI literacy may fall short in determining whether AI’s impact on values is acceptable. Part of what can make our judgement fall short is the way AI operates. “AI can provide answers within milliseconds, taking away the time we would usually spend on critical thinking. Subconsciously, we lean towards accepting the answer AI gives us, because we lose the time to ‘feel the pain’ or feel empathy. We’re placing more and more value on efficiency, so that value overrides other important factors in determining what’s right and wrong.”

Because this technology mimics human behaviour, accepting what comes out of an AI chatbot for instance, becomes even more prevalent, almost as if showing empathy for the chatbot. “’Well, it searched millions of data sources quicker than I ever could and came up with an answer so fast, it must be true’, one might think subconsciously. This is another aspect that could lead to conflicts and misalignments.”

The current AI race is impossible to win

Thus, Cecília is critical when it comes to AI, without being pessimistic. “AI isn’t inherently good or bad. It’s a tool. But we’re treating it as a promise to solve everything. And governments fear ending up behind in the AI race, just as companies are, priming them to override important values and systems to be on top. But when we feel rushed, we tend to very easily fast by other people’s priorities.”

Prioritising is a key aspect of mitigating risk and negative impact, as far as Cecília is concerned. “That means taking the time to actually ask ourselves what it is we’re trying to solve. What’s the reason we’re doing what we’re doing? Questions that are of paramount importance for civil servants. We shouldn’t lose sight of the bigger picture. Being able to take our time and think about the problems we want to solve is a privilege. And AI is just a tool to help us address some of these problems. It’s not the holy grail.”

Cecília rounds off by saying: “I don’t think anyone will be a winner in this AI race if we don’t take enough time to properly think about what we want. We’re too focused on trying to solve problems with AI. But are these the real problems we should focus on? Should AI implementation be a priority in all processes and sectors? We’re asking too few questions. But as governments offering public services, we’re required to think about our priorities. We don’t have infinite resources, and in the end, all that matters is that we create a society, a system, that respects the values of all the people in it.”