There is an extensive debate on the opportunities and risks of applying artificial intelligence (AI) in increasingly diverse areas. The use of AI in fields such as finance, biotechnology and urban mobility has been undergoing development and experimentation for decades. However, empirical evidence on the development of AI in the sphere of government and public administration is scarce, albeit emergent. It is crucial to know what citizens think about the use of AI in public decision-making processes in order to ascertain to what extent the debate is real or is actually blown out of proportion or at least premature (Innerarity, 2023). To know what place the Spanish population wants AI to occupy in political and administrative matters, first it is necessary to know what citizens think about AI, both in general terms and in more specific areas in which they might have more experience or might have accrued knowledge.

The results of the survey show that two-thirds of Spanish citizens claim to know, broadly speaking, what AI is, although only one in five feels capable of explaining what it actually consists of. The respondents are more ambivalent about the impact of AI on the economy. On the one hand, they are moderately optimistic about the impact AI may have on the creation of new economic opportunities. On the other hand, however, they appear to show some concern about the possible disappearance or replacement of jobs, although it is striking that this concern is expressed more about other people’s jobs than about their own.

As regards their general overview of AI, 60% express a positive opinion. However, on comparing different fields, such as medical diagnoses, mortgage loans and high-speed trains, a less favourable image is obtained: in these areas, Spanish people would prefer an expert to make the decision following his or her own judgement rather than the recommendation of an AI system. Specifically, the censoring of abusive messages on social media is the only area in which a majority of respondents would prefer the expert to rely on the recommendations of an AI system.

According to the respondents, not all tasks will be replaced by AI at the same pace. The censoring of abusive messages on social media is the activity that AI will replace most rapidly. As regards the granting of mortgage loans, 40% consider that machines will make these decisions on their own within a period of between two and four years. Another 40% think that we will have to wait a little longer, between five and ten years, for machines to perform all medical diagnoses and drive high-speed trains completely autonomously. Those who believe that machines will never perform these tasks completely, or at least that they will take more than fifty years to do so, in no case exceed 25%.

Little support is found for the use of AI in the public sphere, particularly when it comes to political decision-making. On comparing this support with that expressed for more consolidated political decision-making models, such as representation, participation, and expertise (Font  et al., 2015), AI is the worst rated mechanism or actor.

The Spanish population shows very limited support for AI, expressing only a mildly favourable opinion when it is applied to simple routine tasks in the sphere of public administration. The greater the complexity and the democratic nature of the decision-making process, the stronger the rejection felt towards using AI. Are the Spanish especially reluctant to incorporate AI in government and public administration? The answer is “no”, as similar studies carried out in other countries show an identical pattern (König, 2022).

In Spain, 65% of the population are against politicians relying on the recommendations of an AI system when making decisions, as opposed to relying on their own judgement, while 15% are in favour and 20% are neither for nor against. On comparing the support given to AI with that given to a wide range of mechanisms and key actors in political decision-making, the former is found to be very low. Support for referendums and experts is around 7 out of 10, being the best-rated systems. The idea that political decisions should be made solely in accordance with scientific evidence receives a score of 5.5, while assemblies of citizens chosen by lot, political representatives and business leaders score between 3.5 and 4. Support for AI as a political decision maker stands at 1.6 out of 10.

An experimental design was used to analyse the degree of satisfaction with the functioning of a participatory budget in which the participants can select a set of proposals on public policies using various different decision-making mechanisms, including the use of algorithms. The results indicate that the degree of satisfaction depends mainly on whether the final outcome of the decision is favourable to the participant’s preferences, and less so (albeit with some nuances) on the way in which the decisions were reached.

According to the experiment, based on the simulation of a participatory budget in which the participants are divided into groups and then assigned different decision-making scenarios, the main factor influencing their satisfaction with the process is whether or not the outcome of the decision matches their own preferences. This conclusion is in line with previous studies that have compared the effect of various decision-making processes on the degree of democratic legitimacy perceived by citizens (Esaiasson et al., 2016). However, the novelty of this research study lies in the fact that this is the first time that algorithms have been included in the equation as a potential political actor.

The study raises new questions. Contrary to expectations, the results suggest that higher doses of transparency on the way decisions are reached does not always lead to greater satisfaction for the participants. The expected result would be that higher doses of transparency ‒ an explanation of the criteria used to make the decision – would mitigate the negative effect among those participants whose proposals were not chosen. However, the observed outcome is that greater transparency not only fails to mitigate this effect but sometimes may even increase dissatisfaction with the process.

ESAIASSON, P., M. PERSSON, M. GILLJAM y T. LINDHOLM (2016): «Reconsidering the Role of Procedures for Decision Acceptance», British Journal of Political Science, 49(1), 291-314. 

FONT, J., M. WOJCIESZAK y C. J. NAVARRO (2015): «Participation, Representation and Expertise: Citizen Preferences for Political Decision-Making Processes», Political Studies, 63, 153-172. 
 

INNERARITY, D. (2023): «The epistemic impossibility of an artificial intelligence take-over of democracy», AI & Society, 1-5. 

KÖNIG, P. D. (2023): «Citizen conceptions of democracy and support for artificial intelligence in government and politics», European Journal of Political Research, 62(4), 1280-1300.