Autor: José Luis Fernández-Martínez

Universidad de Málaga

 

Autor: Sara Pasadas del Amo

Universidad de Córdoba

 

Modalidad: Presencial

 

Abstract: 

This paper investigates how the use of AI algorithms in political decision-making affects satisfaction with the performance and outcomes of the process. We also analyze if outcome favourability, which according to the literature, is ""the strongest determinant of individuals’ willingness to accept authoritative decisions' (Esaiasson et al. 2016) mediates this effect. Using data from a July 2023 web survey of 3,077 Spanish respondents aged 18 to 64, we analyze the results of a population-based experiment simulating online Participatory Budgeting (PB). First, participants were randomly assigned to four different scenarios of decision-making: i) direct democracy, ii) argument visualization, iii) opaque AI algorithm and iii) transparent AI algorithm. Second, they chose their preferred policy proposals. They were then presented the process outcomes produced using different decision-making approaches. Half of the participants in each scenario were shown an outcome that coincided with the proposals chosen by them, whereas the other half were shown non-coincident outcomes. Lastly, all participants were asked to assess the process using different measures that capture their level of satisfaction.

 

Consistent with prior research, our findings highlight that outcome favourability stands out as the primary driver of satisfaction with the decision-making process, both in terms of its performance and outcomes. Its impact is so robust that it overrides the effects associated with the decision-making procedures. The majority of procedural effects manifest in experimental groups where outcomes deviate from participants' choices. Within these groups (unfavourable outcomes groups), the procedure yielding the highest satisfaction involves the selection of the optimal choice by an opaque AI algorithm. In all alternative scenarios, even when employing the transparent AI algorithm that considers the votes of all participants to select the optimal outcome, satisfaction with both the process's performance and its outcomes decreases significantly.