Artificial intelligence is generating extraordinary expectations, particularly in those areas where we have been stuck as a society for years. One of these areas is the clear obsolescence of administration: AI will cut the Gordian knot of inefficiency and ineffectiveness and, finally, put an end to the mess of bureaucracy and the headlong rush of administrative procedures. These hopes are not unfounded. At an unprecedented speed for the administration, we are already seeing proofs of concept and pilot tests that are far more than promising and with real impacts on both satisfaction and the time required to achieve that satisfaction. Without intending to overwhelm the wine, the question that must be asked is whether these improvements are palliative—and welcome—or are truly transformative—and consequently will have a multiplier effect on the entire administration and society.
There are three things we should expect from artificial intelligence when it's promoted by public institutions: that it works, that it does good things, and that it doesn't harm anyone.
For it to work, extraordinary developments in our country are still needed in terms of data governance, interoperability, and the digitalization of public services. The good news is that these developments are being made, based on the positive experience of implementing the General Data Protection Regulation, state interoperability and security schemes, and, since the summer, the European AI Law, among many other things. However, this is a very instrumental vision of AI: oriented toward improving services (which is good) and day-to-day life, but without a strategic, systemic, or impactful vision.
This higher vision is what we want when we ask AI to "do good things," that is, to look after the general interest. This requires thinking about its purpose: AI in government, for what? And, above all, government, with AI, for what? The answer places us at a higher level than the previous, strictly utilitarian point. And, in fact, there are two answers. On the one hand, it is necessary for the government, the governments, to keep in mind what part of the world they want to change and how, what their long-term strategic planning is, what their short-term operations are, what their objectives and short-term goals are, and with what resources and with what teams. Unfortunately, strategic planning is more the exception than the rule in government. On the other hand, it requires approaching AI not as a solution, but as a public infrastructure. In the same way that the road network enables communication and exchange, reconfigures territories, changes economies, and marks generations, digital public infrastructure can also have this profoundly transformative effect—positively and negatively. However, this approach has nothing to do with improving public services, but rather requires an administration that participates in a national debate and wants to have a structural impact: it requires its own profiles, resources, and time.
Finally, we want the impact of AI to be positive, so that it doesn't harm anyone. For both, but especially for the latter, it is necessary that everything—the purpose, the design, the implementation, the monitoring, the evaluation, the assessment of the impact—be collective and agreed upon at the level of all Society. Due to both the complexity of the technology itself and the scope of what we're asking of it, it's impossible for the administration to succeed alone. There are too many factors to consider to be able to be everywhere without slowing down technological progress or overlooking all the blind spots that may arise. The only solution is for the governance of artificial intelligence in the administration to be a collective mission. A mission in which the administration opens the door and sets tables and chairs so that all the actors involved can sit down to debate, design, and implement, each from their area of expertise and interest.
Planning for the general interest and a systemic vision with the participation of social stakeholders require a profound transformation of the administration as we know it today. A transformation that must come before that the transformation that the incorporation of AI into government promises us. We therefore have two complementary but different paths ahead of us. The first is to leverage AI to unclog pipes and improve inefficient and ineffective processes. The second is to commit to amends, rethinking functions, structure, and organizational chart to be capable of transformation in order to be able to transform.
