Enter the matrix: Examining the psychosocial determinants of support for a technocracy of artificial intelligence

Political Psychology. 2024;00:1–27. | 1 wileyonlinelibrary.com/journal/pops DOI: 10.1111/pops.13048 SPECIAL ISSUE ARTICLE Enter the matrix: Examining the psychosocial determinants of support for a technocracy of artificial intelligence MarcosDono 1 | EvaMorenoBella 2 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2024 The Author(s). Political Psychology published by Wiley Periodicals LLC on behalf of International Society of Political Psychology. 1Institute of Psychology (IPsiUS), Universidade de Santiago de Compostela, Santiago de Compostela, Spain 2Social and Organizational Psychology Department, Universidad Nacional de Educación a Distancia, Madrid, Spain Correspondence Marcos Dono, Institute of Psychology (IPsiUS), Universidade de Santiago de Compostela, Campus Vida, Calle Xosé María Suárez Núñez, s/n, Santiago de Compostela, A Coruña (Office number: 87), Santiago de Compostela 15782, Spain. Email: [email protected] Funding information European Association of Social Psychology Abstract Artificial Intelligence (AI) stands as the most transformative technology since the Industrial Revolution. As AI is integrated into different domains of life, it is likely to also permeate politics. In fact, some voices already advocate for AI to make political decisions. This paper emphasizes the necessity to delve into the psychosocial determinants shaping support for supporting a technocracy of AI (TecAI). The research aims to ascertain whether anomie, moderated by NCC and SDO, predicts support for TecAI across three studies involving Spanish participants (Ntotal = 754). The findings confirm anomie's predictive power on support of TecAI, albeit without establishing causality consistently, as anomie's condition failed to alter support for TecAI in experiments 2 and 3. However, an exploratory analysis showed a small effect of the subdimension of “breakdown of leadership” on support for TecAI in a directional test, informing future research. Moreover, NCC and SDO did not significantly moderate the anomiesupport- for- TecAI link. Ideology also emerged as a consistent predictor, with conservatives consistently showing more support, a finding that ought to be reexamined and clarified. This study marks an initial empirical exploration into AI's political perceptions, hinting at the need for further investigation into this critical domain. KEYWORDS anomie, artificial intelligence, democracy, leadership, technocracy 2 | DONO and MORENO BELLA INTRODUCTION In the future, everything will make perfect sense. So you might as well wear a suit and live every day as though it was your last. InspiroBot Artificial Intelligence (AI) is one of the latest technological revolutions to impact human life at every level (Makridakis, 2017). The capability of AI to learn and develop complex models based on huge amounts of data allows it to efficiently execute many tasks that humans cannot tackle and outperform us in those that we can do (Enholm etal.,2022; Sjödin etal.,2021). Due to its prowess, AI is a tool that is rapidly integrating into many different activities. Today, AI is incorporated in diverse fields, from finance (Milana & Ashta,2021) to sports (Beal etal.,2019) to addressing the climate crisis (Leal Filho etal.,2022). Thus, the potential market value of AI is humongous (Fredström etal.,2022; Wamba- Taguimdje etal.,2020), with leading consulting firm PricewaterhouseCoopers forecasting the economic impact of AI and Machine Learning technologies at $15.7 trillion in the period between 2017 and 2030 alone (PricewaterhouseCoopers,2017). This has generated a fullblown “AI arms race,” with top companies competing to generate ever faster and more impactful developments of this technology (Chow & Perrigo,2023). However, this burgeoning technological business may bring consequences other than economic growth and new, exciting tools that improve our ability to work and quality of life. All technological breakthroughs seem to carry unexpected consequences for human society that will be of greater importance depending on the impact of the technology itself (Kranzberg,1986). For instance, the advent of the internet and social networks improved our capacity to stay in touch with people geographically apart from us but also originated mental health issues (Sun & Zhang,2021). AI is no exception. Among these “collateral effects” there are some that entail deleterious consequences and threats to people's lives. The side effect of the AI revolution that has received the most attention to this date is the impact that it could have on the job market, potentially rendering thousands of jobs obsolete (Acemoglu etal.,2022; Frank etal.,2019). Some other negative impacts of AI refer to “algorithmic biases,” lack of transparency in the decisionmaking process, or unsatisfactory human- AI interactions (Enholm etal.,2022; Sætra,2020). We argue that there is yet another underappreciated and potentially more severe threat, one that directly affects democracies, represented by popular support of a technocracy of AI (TecAI). A classical technocracy is usually defined as a government choice that consists of the citizenship transferring the governing powers to a select group of nonpolitician experts, under the premise that technical expertise provides the best possible political decisions (Centeno,1993). In the case of a TecAI, this power will be transferred to an AI that will be endowed with the task of making political decisions (Sætra,2020). Far from being a trivial matter, the idea of trusting all or at least part of the governance of society to AI has recently started to gain traction. In a recent study conducted on European adults, up to 51% of participants agreed to free up a portion of parliamentary seats in order to transfer that “representation” to AI systems, with agreement reaching up to 66% in Spain, the country most in favor of the measure (Jonsson & Luca de Tena,2021). Beyond popular support, some voices from Academia have also advocated for a TecAI, claiming that this would be the best option to implement good political decisions (Sætra,2020). Some models of AI- based political decisionmaking have already been built (Lee etal.,2019), with results suggesting that the decisions of AI are sounder when it comes to tax policymaking (Zheng etal.,2020) or that people are more satisfied when 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 3 SUPPORT FOR A TECHNOCRACY OF AI AI is used in small, politicaladministrative tasks (Chatterjee etal.,2022). Also, in support of this governing choice becoming a real possibility is the fact that people have been shown to be capable of attributing competence and warmth to the AI, with special emphasis on the former (McKee etal.,2021). Nevertheless, trusting governance to an AI is subject to multiple objections and potential problems. Some of them have been mentioned above and can be considered general caveats of AI, like the lack of transparency, the presence of “deep fake” information, or algorithmic biases (Kordzadeh & Ghasemaghaei,2022; Panch etal.,2019; Thomson,2022; Westerlund,2019). Others are specific to the application of AI to politics, like the intrinsic need of citizens to participate in the process of political decisionmaking, questions about the ability of AI to balance moral choices, or its lack of accountability (Sætra,2020). This being said, customary politics already show at least some of the problems that we outline a TecAI can cause. Corruption, lobbying, biased accounts of reality, or lacks of transparency are not alien issues to the traditional exercise of politics (Campos & Giovannoni,2017). Ultimately, we side with those voices that argue that a TecAI would be detrimental to our societies. Participating in politics is an essential aspect of—at least some—people's lives (Duncan & Stewart,2007), and political systems are legitimized by universal participation (Manin,1987). Moreover, the problems of deepfaking and algorithmic biases (Kordzadeh & Ghasemaghaei,2022; Westerlund,2019) cast doubt on whether AI choices can be guaranteed to be fully independent or rather they can be manipulated. This could mean that illintentioned actors could use the premise of an allegedly neutral AI government to selfinterestedly push a specific political agenda. For all these reasons, we argue that a TecAI may pose a threat to democracies and the correct functioning of our societies, and thus we aim to explain what may drive popular support for this initiative. Within this line, the main goal of this work is to examine psychosocial predictors of popular support for a TecAI. It is important to remark that, regardless of the opinion anyone may have about TecAI, this work constitutes a neutral examination of the psychosocial motives for its support, and thus it should also be of interest to those who defend tasking AI with policymaking. To this day, the fields of Social and Political Psychology have only paid minimal attention to the potential impact of AI on human politics. Hence, there is little specialized literature that can be used for theorization and modelbuilding. However, the rich existing literature on Social and Political Psychology can provide robust theoretical foundations for a predictive model of popular support for a TecAI based on literature on radicalism, extremism, and other unconventional political choices that go against the established political system (i.e., Duckitt & Fisher,2003; Hogg & Adelman,2013; Jost etal.,2017; Kruglanski etal.,2019; Pratto etal.,1994). Despite these being different phenomena that may come accompanied by different implications or even behavioral consequences, they share the feature of confronting the current political system (Kruglanski,2018; Trip etal.,2019). Therefore, we argue that this literature may represent a fitting starting point for initial theorization of the motives that can drive support for TecAI, as this will also represent totally transforming society and the political system by departing from democracy. Anomie as a predictor of popular support for a technocracy of the AI The central idea of the present work is that a perception of a certain malfunction within the current system and/or its actors may cause support for a TecAI. Overall, trust in institutions, political actors, and the system itself are all key elements that predict the stability of a political system, while a lack of trust fosters a desire for change (Devos etal.,2002; Jost etal.,2017). Aside from political trust, general social trust felt toward other citizens is also a key factor in predicting a stable status quo and a strong democratic commitment (Zmerli & Newton,2008). Lastly, we propose that a final factor that we can include within a general sense of societal 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4 | DONO and MORENO BELLA malfunction is the perception of an inadequate moral code. Since politics is about discussing and agreeing on the best way of managing society (Hague & Harrop,2004), morals understandably are a key factor in political life, as they constitute a guide for ideal conduct for different individuals and groups (Gert & Gert,2017). In fact, moral convictions and obligations are key explanations of political collective action (Sabucedo etal.,2019; van Zomeren etal.,2012), while perceived threats or transgressions of moral values seem to predict extreme political choices and behavior (Fiske & Rai,2014; Skitka,2002). Whether they are about political actors or fellow citizens, perceptions of immorality should encourage support for a TecAI. All of these sources of discontent with the way society is arranged and ruled are, we argue, blended in the broader construct of anomie (Teymoori etal.,2017) based on the original sociological definition of anomie coined by Durkheim(1897). Teymoori etal.(2017) carried out a reconceptualization of this construct from an eminently psychological approach and defined it as a perception of profound deterioration of social and political conditions in society, including moral values. Additionally, anomie is conceptualized as having two different dimensions: breakdown of the social fabric and breakdown of leadership. The first represents the perception that within society at large there is low social trust, low solidarity, and deteriorated moral standards. The latter is understood as the perception of political leadership as corrupt, unjust, and illegitimate (Teymoori etal.,2017). Due to the profound pessimistic vision that anomie entails about society, we argue that it should be a powerful predictor of public support for a TecAI. We argue that, under anomie, an AI government could be seen as a good choice to return society to proper functioning due to its exceptional efficiency and its incorruptible nature, contrasting with a corrupt and rotten society and political class. This is supported by the fact that the competence of AI is not merely a technical fact but also a human attribution, as people have been found to— primarily—judge AI as competent (McKee etal.,2021). Effectively, trusting AI with the government of society would represent an efficacious, taskcentered way of regaining order and would eliminate the problem of selfinterest, which are among the main arguments of the defendants of this type of government (Sætra,2020). Since the main strength of AI is its technical performance rather than moral deliberation, we argue that, if both subdimensions of anomie were to be contrasted, breakdown of leadership will have a stronger impact on predicting support for a TecAI than breakdown of the social fabric. Further strengthening our case for anomie as a predictor, we find that anomie combines different considerations that are related to radical and/or extreme political attitudes like a catastrophic view of society (Duckitt & Fisher,2003), distrust in the system and perception of corruption (Gáspár etal.,2021; Wintrobe,2006), and threatened moral values (Fiske & Rai,2014; Skitka,2002). Anomie has specifically been observed to predict support for strong and authoritarian leaders (Heydari etal.,2012; Sprong etal.,2019), adoption of authoritarian attitudes (Teymoori etal.,2017), and political extremism (Ionescu etal.,2021b). We argue that supporting a TecAI would be a radical political idea bringing forth a desired structural change that nevertheless avoids the high social and moral costs associated with involvement in political extremism (Lake,2002; Wintrobe,2006). Thus, as the capabilities of AI and the public become more familiar with them, a TecAI becomes an increasingly feasible, less risky option for anomic people to defend. However, we also argue that the relation between anomie and support for a TecAI should be moderated by additional factors. Hereafter, we describe some variables that we believe could condition this association. Moderating variables of the anomiesupport for AI technocracy relation In the present work, we posit two variables that may act as moderators of anomie's effect on support for a TecAI: the Need for Cognitive Closure (NCC; Webster & Kruglanski,1994) and the Social Dominance Orientation (SDO; Pratto etal.,1994). The main idea behind this addition is to examine the effects of one intrapersonal moderator and another moderator that 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 5 SUPPORT FOR A TECHNOCRACY OF AI represents an ideological attitude in an attempt to capture the whole range of the personby- context interaction (Lewin,1936). The first hypothesized moderator is NCC, defined as the tendency to feel distressed in the face of uncertainty that elicits a desire to achieve fast, definitive answers to ambiguity (Webster & Kruglanski,1994). It has been associated with people's willingness to join radical and extreme groups, as these groups tend to promote and exercise more absolute, black- and- white thinking (Hogg & Adelman,2013; Webber etal.,2018). Similarly, a TecAI will put forward clear and definitive rules based on a data analysis capability that cannot be disputed by human beings, supported not only by AI's characteristics but also by human attributions of high competence to the AI (McKee etal.,2021). The fact that AI is essentially a neutral, incorruptible machine should also help reduce the uncertainty surrounding not only the decision itself but also the motives behind those choices, as AI is incapable of acting on selfinterest. Moreover, we argue that individuals high in NCC would have further reasons to support a TecAI beyond the satisfaction of ambiguityreducing needs. A TecAI would not only imply trusting governance to an extremely efficient machine but also a political system where citizen autonomy and inference are reduced as political representatives are substituted by AI. This reduction of personal autonomy we believe would equate a ‘harsh’ power tactic, a strategy of management consisting in being much more directive about the tasks and roles of people in a team, which those high in NCC have shown to prefer, and that for these individuals has even been associated with lower levels of workrelated psychological distress (Bélanger etal.,2015, 2016). We argue that anomie can be addressed in many potential fashions, yet those high in NCC will be more prone to support TecAI as they perceive it to be more proficient in making decisions and less tentative on their solutions. Thus, NCC should enhance the relationship between anomie and support for TecAI. Once we have discussed an intrapersonal variable that may enhance support for a TecAI, we seek to complete our model by adding a variable that expresses a more structural belief. Since a TecAI would be a hierarchical organization of society, we argue that such a moderating role would be played by SDO, defined as a preference for a hierarchical organization of society in which each social stratum assumes its place in that hierarchy (Pratto etal.,1994). The first argument in favor of this relation is that people high in SDO tend to prefer autocratic and directive systems (Duriez & Van Hiel,2002), similar to what a TecAI would represent. Furthermore, AI's “beyond human” competence may lead to a perception of AI as an unquestioned authority, reinforcing its rightful place as a governing device in the eyes of those high in SDO, as attributions of competence tend to legitimize hierarchical social systems (Oldmeadow & Fiske,2007). We argue that the superiority in performance of the AI would represent a dominant trait, which tends to be preferred in leaders by those high in SDO (Laustsen etal.,2015). Moreover, preferred leadership styles by people high in SDO focus on production while ignoring the particular desires of people (Nicol,2009), much in line with what AI would offer as a governing tool. In this same breath, SDO has also been found to be associated with a greater tendency to engage in mechanistic dehumanization (Kuljian & Hohman,2023), which translated to the social realm could also imply a preference for the “productiveonly” approach of AI. For these reasons, we argue that those people high in anomie that are also high in SDO would be more willing to support a TecAI. This model of governance implies trusting a single agent to make the relevant decisions and thus is much more hierarchic than a democratic political system, while the technical prowess of AI would also make it much more performanceoriented. Therefore, we posit SDO would positively affect the relation between anomie and supporting a TecAI. THE PRESENT RESEARCH Across three studies, we examined whether anomie, NCC, and SDO could explain a preference for a TecAI. We conducted a first, crosssectional study that aimed to examine the 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6 | DONO and MORENO BELLA hypothesized associations among anomie, NCC, SDO, and support for TecAI. In Study 2, we conducted a singlefactor, twolevel (anomie vs. control), betweensubjects experimental design to examine the causality of the anomie support for TecAI link and to test the effects of NCC as an intrapersonal moderator and the effects of SDO as an ideological attitude. Finally, in another experimental design in Study 3, we will seek to replicate the results of Study 2 while also examining anomie's subdimensions. We expected that the breakdown of society subdimensions would have a stronger impact on support for the TecAI than the subdimension of breakdown of the social fabric. Thus, this time we will resort to a betweensubjects, singlefactor, fourlevel design (composite anomie vs. breakdown of society vs. breakdown of leadership vs. control). In all studies, we test the hypothesis that anomie leads to support for the TecAI (H1). Specifically, in the first study, we test this association from a correlational standpoint (H1a). We also expect to find a positive relationship between NCC anomie (H2a) and between NCC and support for TecAI (H2b). Additionally, we anticipate a positive relationship between SDO and anomie (H3a) and between SDO and support for TecAI (H3b). In studies 2 and 3, we examine the hypothesis that anomie causes support for TecAI to increase (H1b).1 Additionally, in the second study we test the moderating role of an intrapersonal variable, that is NCC; anticipating that support for the TecAI is stronger for those participants who score high in both anomie and NCC (H4). In our second study, we also test an ideological attitude as moderator: SDO. We expect the effect of anomie on support for a TecAI to be stronger for those who score higher on SDO (H5). Furthermore, in Study 3, we test the following secondary hypotheses: First, the hypothesis that when comparing the effects of the breakdown of social fabric and the breakdown of leadership subdimensions of anomie, the latter will cause a rate of support for a TecAI that is statistically significantly higher than that caused by the breakdown of social fabric (H6). Moreover, we expect the composite anomie condition to cause significantly higher levels of support for the TecAI than any of anomie's subdimensions (H7). And finally, we hypothesize that all anomie conditions (composite, breakdown of social fabric, and breakdown of leadership) cause higher support for the TecAI than the control condition (H8). All hypotheses have been preregistered, and materials hosted in the Open Science Framework (OSF), available at the link https:// osf. io/ nyvtf/ ? view_ only= f6e88 07bfa ff4c7 9a640 8d77f 0a4b3c0. All studies included in this work have been approved by the Ethics Committee of the University of Santiago de Compostela. STUDY 1 In the first study, we used crosssectional data to assess the hypothesized associations between anomie, NCC, SDO, and support for a TecAI. Method Sample size We determined our sample size via an a priori power analysis for linear multiple regression (Fixed model, R2 deviation from zero) using G*Power (Faul etal.,2009). The power was set to the standard of .90, and alpha was also set at the typical value of .05. As a reference for the expected effect size, we employed the mean effect size from research on attitudes in Social 1We numerated the hypotheses H1a and H1b because they express testing the same relationship from—first—a correlational nature and—then—establishing a causal relationship, respectively. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 7 SUPPORT FOR A TECHNOCRACY OF AI Psychology, which is d = .53 (Richard etal., 2003), a medium- to- big effect size (Cohen,1988). Accordingly, we have considered it appropriate to fix the expected effect size at a medium effect size, which for this particular analysis would be f2 = .15 (Faul etal.,2009). The analytical strategy chosen was an a priori analysis for a fixed model multiple regression testing deviation from zero, that allows us to test our full model of 6 predictors (1 covariable: political orientation, 5 focal predictors: Anomie, NCC, SDO, Anomie × NCC, Anomie × SDO). The resulting ideal sample size was a minimum N = 123. However, we decided to collect at least 200 observations recommended by Schönbrodt and Perugini(2013). The resulting sample would thus be able to provide more power and/or detect a smaller effect size than that originally calculated via power analysis, with a minimum detectable effect size of f2 = .09, according to a sensitivity analysis also performed on G*Power. Participants and procedure All participants were recruited through the Prolific Academic crowdsourcing platform— which research has shown to provide data of sound quality (http:// www. proli fic. co; Palan & Schitter,2018; Peer etal.,2017)—and they completed an online questionnaire. Upon finishing the survey, participants were debriefed and thanked for their participation. They also received a monetary compensation proportional to a $10.20 per hour rate. The requirements for participation were being over 18 years old, having Spanish citizenship, and having Spanish as their native language. We set a criterion of proportional distribution of gender (≈50% women). All these criteria were established in Prolific's demographic prescreening, and we obtained 210 responses. In addition, we excluded those participants who fail an attention check item (“This is an attention check question, please select option number ‘2’”). Therefore, our final sample will be composed of 201 participants (MAge = 33.06, SDAge = 10.49, %women = 46.8). Measures The full scales of all the presented measures can be consulted in Section1 of the Supplementary Materials (https:// osf. io/ nyvtf/ ? view_ only= f6e88 07bfa ff4c7 9a640 8d77f 0a4b3c0). Political orientation This variable was measured using one item (“In politics, we sometimes talk about ‘left’ and ‘right’. Where would you place yourself on this scale? 0 means ‘markedly leftwing’ and 10 means ‘markedly rightwing’. Choose the number that best reflects your position.”) on a 11- point Likert scale, where 0 indicated “leftwing” and 10 indicated “rightwing” (Kroh,2007; M = 4.50, SD = 2.55). Sociodemographic and socioeconomic data We also asked participants about their age, gender (1 = Man, 2 = Woman, 3 = Other), as well as educational attainment (1 = Less than high school degree, 8 = Doctoral degree), and their monthly family income level (1 = Below €500, 7 = Above €5000). Subjective socioeconomic status (SES) was measured with the MacArthur Scale of Subjective Social Status (Adler etal.,2000; M = 5.57, SD = 1.37). Participants indicated the rung they thought they stood on within a ladder of 10 rungs that represents the different (SES) in society. High scores indicated higher subjective SES. We also asked participants to what extent they knew about AI and its functionality before taking part in our study (1 = Nothing at all, 7 = Very much). We also asked participants about their subjective knowledge of AI before participating in our study (1 = Nothing at all, 7 = Very much). We calculated a score for the two items. It is a 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8 | DONO and MORENO BELLA subjective measure of knowledge of AI—not the actual knowledge—since it relies on selfreporting (M = 5.08, SD = 1.03). Anomie It was measured using the 12- item scale developed by Teymoori etal.(2016), assessing the perceived moral decay of society. Six items assessed the breakdown of the social fabric (e.g., “In [country] today, everyone thinks of him/herself and does not help others in need”), and six items measured the breakdown of the leadership dimension (e.g., “In [country], politicians don't care about the problems of the average person”). All items were measured on a 7- point scale (1 = Strongly disagree, 7 = Strongly agree). Five items were reversed coded (see Section1 of Supplementary Materials). The two scores were moderately correlated with each other (r = .28, p < .001). Hence, we computed a variable including the mean scores of both dimensions. The items showed adequate internal consistency (α = .82).2 Higher scores indicate higher perceived anomie (M = 4.56, SD = .85). Need for cognitive closure (NCC) This variable was assessed with the short version of the Need for Closure Scale (NFCS; Roets & Van Hiel,2011; Webster & Kruglanski,1994). This scale is composed of 15 items (e.g., “I feel uncomfortable when I don't understand the reason why an event occurred in my life.”). Participants indicated their answers on a 6- point Likert scale (1 = Strongly disagree, 6 = Strongly agree, M = 3.83, SD = .72). The items' internal consistency was good (α = .87). Social dominance orientation (SDO) We used the 4- item scale (e.g., “We should not push for group equality”) of Social Dominance Orientation (Pratto etal.,2013). The answer format was a 10- point Likert scale (1 = Extremely oppose, 10 = Extremely in favor; M = 2.98, SD = 1.60). Two items were reversecoded to calculate the mean score in SDO. The items showed good internal consistency (α = .76). Support for technocracy of AI (TecAI) We asked participants about their support for allowing AI to take charge of different governance decisions. For this purpose, we used a 9- item scale (e.g., “I think Parliament should be replaced by AI and AI should create the laws of my country”), and participants answered on a 7- point Likert scale (1 = Strongly disagree, 7 = Strongly agree; M = 3.57, SD = 1.36). The internal consistency of the items was adequate (α = .92). Data analyses First, we explored missing data in our sample. We decided to not replace missing values as long as they did not exceed 10% of the data (Dong & Peng, 2013). Preliminary analyses for this study consisted of an observation of descriptive statistics as well as an assessment of the internal consistency of the scales. Since no scale for public support for a TecAI was available at this time, we used the results from the internal consistency analysis and Exploratory Factor Analysis of the scale to modify it going forward. For instance, if reliability analysis shows that an item should be dropped to bring this statistic to an acceptable level (i.e., α > .60), the scale would have been then reduced, and that version would have been used in Studies 2 and 3. Correlational bivariate analysis was conducted to examine the relation among all the variables 2Reliability scores were also calculated for the breakdown of social fabric (α = .82) and breakdown of leadership (α = .83) subscales. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 9 SUPPORT FOR A TECHNOCRACY OF AI of interest. We included in this analysis the mean scores of composite anomie, breakdown of social fabric, breakdown of leadership, NCC, SDO, support for TecAI, and political orientation. Provided we found a positive relation between anomie, NCC, and SDO with support for TecAI, we ran a hierarchical regression analysis. We included political orientation as a covariate in the first step of the hierarchical regression analysis to control for their possible influence, hinted by previous research (Ionescu etal.,2021a). In the second step, anomie, NCC, and SDO were included as predictors, whereas in the third step, the interaction terms of the latter two with the former were included as predictor variables. Support for TecAI was included as the criterion variable. Results Considering that the support for TecAI was an adhoc scale, we ran a pilot study to observe key psychometric indicators (see Supplementary Materials, Section3). This analysis indeed supported its use, and no alterations to the scale were made based on this data. According to our analytical plan, we then moved on to the analysis of bivariate correlations among our main variables. Correlation analysis showed that general anomie—and breakdown of leadership—were positively correlated with support for TecAI, supporting H1a. Regarding the moderating variables, results show that NCC was positively associated with the breakdown of social fabric but not with general anomie, nor with the breakdown of leadership (partial support to H2a) nor with support for TecAI (H2b not supported). We also found that SDO was significantly and positively correlated with the breakdown of leadership and general anomie, supporting H3a. Moreover, results revealed that SDO was significantly and positively correlated with support for TecAI (confirming H3b). Finally, political orientation was also significantly associated with supporting a TecAI, with conservative participants favoring the idea more (see Table1). We performed our hierarchical regression in three steps. Step 1 included political orientation, Step 2 the predictors in anomie, NCC, and SDO, and Step 3 included the interaction terms of anomie by NCC and anomie by SDO. Model 1 was statistically significant (F = 15.45, p < .001) showing that political orientation predicted support for the TecAI. Model 2 was also statistically significant (F = 6.79, p < .001), as was the increase in variance (ΔR2 = .05, p = .013). In this second step, anomie emerged as the only significant predictor of support for TecAI (B = .29, p = .012). Finally, Model 3 was also significant (F = 4.59, p < .001), yet the increase in explained variance was not (ΔR2 = .003, p = .75). This suggests that adding the interaction terms had no impact on the predictive power of the model. Results on the impact of coefficients in TABLE 1 Bivariate correlations between the variables included in Study 1. 1 2 3 4 5 6 1. General anomie – 2. Breakdown of society .75*** – 3. Breakdown of leadership .84*** .26*** – 4. NCC .13 .28*** −.05 – 5. SDO .29*** .05 .38*** −.11 – 6. Support for TecAI .25*** .05 .32*** −.08 .27*** – 7. Political orientation .32*** .10 .37*** −.04 .60*** .27*** Abbreviations: NCC, need for cognitive closure; SDO, social dominance orientation; TecAI, technocracy of artificial intelligence. ***p < .001; 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 16 | DONO and MORENO BELLA was substituted with more details on the deterioration of moral values, including a particular fictitious episode where an elderly lady is mugged and no bystanders come to her aid. This condition was accompanied by a picture of a group of people struggling, allegedly entitled “Cooperation today.” Conversely, the breakdown of leadership conditions contained no information on society's values of trust and cooperation and discussed at length the dysfunctions of the political system. In it, Spain was described as a country built upon social movements, collective efforts that are now completely ignored by politicians, regardless of the ideology of both the movements and the politicians in charge. It was accompanied by an illustration of a male politician that projects a shade of himself with a “Pinocchio nose,” a symbol associated with lying. Lastly, the control condition consisted of a similar news piece that discussed societal values and political leadership using the same structure as the composite anomie piece. However, it concluded that, while values and political leadership had not improved, they had not gotten worse either, remaining “stagnant at previous acceptable levels.” The main idea was to put across the impression that values and quality of leadership remain at a placid stalemate. This last condition was complemented by a picture of a graph that replicated that of the composite anomie condition but that conveyed the information that cooperation, trust, and political corruption maintain a steady and unchanged trend within acceptable levels. Same as in Study 2, after being exposed to the manipulation, participants answered two comprehension checks and then responded to the independent and dependent variables. Measures Political orientation We measured it using the single item of political orientation we used in Studies 1 and 2 on a 11- point Likert scale, where 0 indicates “leftwing” and 10 indicates “rightwing” (M = 4.62, SD = 2.49). Sociodemographic and socioeconomic data We asked participants about their age, gender (1 = Man, 2 = Woman, 3 = Other), educational attainment (1 = Less than high school degree, 8 = Doctoral degree), and their monthly family income level (1 = Below €500, 7 = Above €5000). Moreover, we used the MacArthur Scale of Subjective Social Status (Adler etal.,2000; M = 5.57, SD = 1.48) to measure subjective SES. We also asked participants about their subjective knowledge of AI (1 = Nothing at all, 7 = Very much; M = 4.92, SD = 1.21). Anomie Once again, we used the same 12 items that will be used in Study 2 to check whether the experimental manipulation work. Participants indicated their answer on a 7- point Likert scale (1 = Strongly disagree, 7 = Strongly agree; M = 4.54, SD = 1.00; α = .87). Need for cognitive closure (NCC) We evaluated it like in Studies 1 and 2 with the NFCS (Roets & Van Hiel,2011; Webster & Kruglanski, 1994). The answer format was a 6- point Likert scale (1 = Strongly disagree, 6 = Strongly agree; M = 3.75, SD = .74; α = .86). Social dominance orientation (SDO) We used the 4- item scale (e.g., “We should not push for group equality”) of Social Dominance Orientation (Pratto etal.,2013). The answer format was a 10- point Likert scale (1 = Extremely 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 17 SUPPORT FOR A TECHNOCRACY OF AI oppose, 10 = Extremely favor; M = 2.95 SD = 1.64). Two items were reversecoded to calculate the mean score in SDO. The items show good internal consistency (α = .78). Support for technocracy of AI (TecAI) It was assessed using our own 9- item scale of support for TecAI. The answer format was a 7- point Likert scale (1 = Strongly disagree, 7 = Strongly agree; M = 3.53, SD = 1.37; α = .91) as in the prior studies. Data analyses Again, no missing data had to be replaced according to the same criteria as in previous studies. Preliminary analyses for this study consisted of an observation of descriptive statistics, an assessment of the internal consistency of the scales, and a correlational bivariate analysis to observe which variables are associated. An ANCOVA conceptually replicating that conducted in Study 2 (i.e., including at least political orientation as a covariate) but using the two additional factors that try to isolate the effects of each of anomie's subdimensions was run. This allowed us to test whether our H1b is replicated in this study and to assess the particular effects of each of anomie's subdimensions. To assess the whole model, we examined whether NCC (H4) and SDO (H5) moderate the effects of anomie. To do that, we ran a moderation model (Model 2) using the PROCESS macro for SPPS (Hayes,2022) with 10,000 bootstraps using the experimental condition (composite anomie vs. breakdown of social fabric vs. breakdown of leadership vs. control) as the predictor of support for TecAI, and NCC and SDO as moderators. The categorical predictor was dummycoded using PROCESS' indicator method, which individually compares each experimental condition with the control. Finally, to examine the potential differential impact of anomie's subdimensions on support for TecAI, we conducted a post hoc test on the aforementioned ANOVA while applying Tukey's correction, where we contrasted the effects of anomie's subdimensions against each other (H6), the effects of anomie's subdimensions against the composite anomie condition (H7), and the effect of all anomie conditions against the control condition (H8). Results As in the previous study, we first conducted bivariate correlations. General anomie, breakdown of society, and breakdown of leadership were positively correlated with support for TecAI, confirming H1a. Concerning the moderating variables, NCC was significantly associated with the breakdown of society, breakdown of leadership, and general anomie, thus supporting H2a. Moreover, lending support to H2b, we also found a positive and statistically significant correlation between NCC and support for TecAI. SDO was significantly and positively associated to anomie and its subdimensions, in line with H3a. Moreover, SDO was positively related to support for TecAI, dovetailing results of Studies 1 and 2 in supporting H3b. Also replicating Study 2, conservatives showed a stronger preference for a TecAI (see Table4). We checked the experimental manipulation through a oneway ANOVA (betweensubjects) with four levels (composite anomie vs. breakdown of society vs. breakdown of leadership vs. control condition) and composite anomie as the dependent variable; and through a oneway MANOVA with the breakdown of leadership, and breakdown of the social fabric as dependent variables. This time, results supported the success of the experimental manipulation, observing that it affected general anomie [F (3, 316) = 5.16, p = .001, 𝜂2 p = .05], breakdown of leadership and breakdown of society dimensions [F (3, 316) = 4.03, p = .008, 𝜂2 p = .04; F (3, 316) = 4.13, p = .007, 𝜂2 p = .04]. When looking at post hoc effects, we 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 18 | DONO and MORENO BELLA found that both the composite anomie and the breakdown of leadership conditions generated significantly higher scores for composite anomie and its two subdimensions, while the breakdown of society did not generate statistically significant differences, all compared to the control condition (see Table5). To examine the causal effect of anomie on support for a TecAI, we performed the planned ANCOVA. Again, contradicting our expectations, we did not find a significant effect of the experimental manipulation on the support for TecAI [F (3, 315) = 1.15, p = .327, 𝜂2 p = .01]. Moreover, we did not find significant statistical differences between the composite anomie (M = 3.41, SD = 1.38), breakdown of the society (M = 3.66, SD = 1.36), breakdown of leadership (M = 3.72, SD = 1.41), and control conditions (M = 3.34, SD = 1.31). The analysis once again revealed political orientation as a significant predictor of support for TecAI, F (1, 315) = 48.43, p < .001, 𝜂2 p = .13. Results about the anticipated moderators NCC and SDO also dovetailed those of Study 2, again failing to support H4 and H5, with no significant interaction between both NCC and SDO with either of the combinations of the experimental manipulation (see Table6). Exploratory analysis Despite the null results of the manipulation on support for TecAI and considering the significant effects anomie consistently showed throughout the studies, we wanted to further explore TABLE 4 Bivariate correlations between the variables included in Study 3. 1 2 3 4 5 6 1. General anomie – 2. Breakdown of society .88*** – 3. Breakdown of leadership .88*** .53*** – 4. NCC .23** .26*.13*– 5. SDO .18** .14*.18*** .13*– 6. Support for TecAI .28** .20** .29** .12*.22*** – 7. Political orientation .34*** .23*** .37*** .09** .62*** .37*** Abbreviations: NCC, need for cognitive closure; SDO, social dominance orientation; TecAI, technocracy of artificial intelligence. ***p < .001; **p < .010; *p < .050. TABLE 5 Descriptive statistics derived from the post hoc tests of the MANOVA (Study 3). Anomie experimental conditions and Control condition General anomie Breakdown of society Breakdown of leadership Control condition M (SD)MDp M (SD)MDp M (SD)MDp M (SD) 1. General anomie 4.71 (.91) .54 .004 4.54 (.84) .36 .120 4.73 (1.02) .55 .003 4.18 (1.12) 2. Breakdown of society 4.44 (1.06) .52 .020 4.28 (1.02) .36 .269 4.47 (1.17) .55 .014 3.92 (1.23) 3. Breakdown of leadership 4.97 (1.05) .54 .015 4.80 (1.10) .37 .245 4.98 (1.12) .55 .014 4.43 (1.22) Note: MD indicates the mean difference between the experimental manipulation and the control condition. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 19 SUPPORT FOR A TECHNOCRACY OF AI the impact of our manipulations. We attempted a directional ttest comparing the control condition and the breakdown of the leadership condition, as we had a directional hypothesis regarding its effect, and these were the most diverse groups when looking at descriptive statistics. This analysis did indeed reveal a small yet significant result, with participants in the breakdown of leadership condition scoring significantly higher than those in the control condition in the support for TecAI measure t(156) = 1.73, p = .042, d = .27 (see Figure1). Robustness test To better evaluate the impact that anomie may have on supporting TecAI, we decided to perform a minimeta- analysis to assess the average effect size of the association. Conducting such analysis on the average correlation between composite anomie and support for TecAI yielded TABLE 6 Results of the moderation analysis (PROCESS Model 2) of NCC and SDO on the relation between the anomie experimental conditions and support for TecAI (Study 3). BSE 95% Confidence interval pLower Upper (General anomie vs. Control) × NCC .130 .286 −.434 .694 .649 (General anomie vs. Control) × SDO .200 .137 −.070 .471 .145 (Breakdown of social fabric vs. Control) × NCC −.377 .302 −.973 .218 .213 (Breakdown of social fabric vs. Control) × SDO .242 .133 −.020 .504 .070 (Breakdown of leadership vs. Control) × NCC −.146 .289 −.715 .423 .613 (Breakdown of leadership vs. Control) × SDO .008 .125 −.238 .255 .945 Abbreviations: NCC, need for cognitive closure; SDO, social dominance orientation. FIGURE 1 Directional ttest between control and breakdown of leadership conditions (St.3). *p < .05. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 20 | DONO and MORENO BELLA an average effect size of μ = .27, p < .001 (see Figure2 for forest plot), indicating a medium effect size according to Cohen's conventions (Cohen,1988). Discussion In general terms, our third study replicated the results of the second. Anomie and both its subdimensions seem to be significantly associated with increased support for TecAI, as well as SDO and a conservative political ideology. Additionally, NCC also significantly and positively associated with support for TecAI, offering mixed results and calling for further inquiry on the subject. Regarding our experimental manipulation, we were not able to confirm the causal role of anomie on support for TecAI. However, contrary to the previous study, the conditions of composite anomie and breakdown of leadership did impact the scores on anomie and its subdimensions. This fact led us to further explore the role of the experimental condition that, in principle, seemed most likely to be effective: breakdown of leadership. In a directional ttest, we found support for a small effect of the condition on support for TecAI when compared to the control group. This may suggest that support for technocracy is uniquely associated with the perception that the political system is disintegrating but unaffected by wider anomicsocietal considerations. GENERAL DISCUSSION With this research on the psychosocial drives of support toward a TecAI, we believe we are contributing to fill an important lacuna currently present in Political Psychology. To the best of our knowledge, this is the first time that the psychosocial motives leading to support for a TecAI have been examined. Moreover, it should be among the first works—if not first—to focus on the interaction between AI and politics from a psychosocial perspective altogether. The reason for this gap in the literature is the novelty of AI itself (at least “generative AI”; Makridakis,2017). However, as stated in the introduction, AI is a technology with immense potential that is developing hastily. It is already being integrated into many human practices and jobs. Pretending that similar integration will not happen in politics seems naïve, especially FIGURE 2 Forest plot for the minimeta- analysis of composite anomie on support for TecAI (Analysis was carried out using Fisher- to- z transformed correlations). 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 21 SUPPORT FOR A TECHNOCRACY OF AI when AI decisionmaking has characteristics that could be—at least in principle—interesting for citizens. At the very least, we as social scientists need to understand what may drive people to want to task AI with political decisionmaking. Whether we think that is a good option, how could it be implemented, or even if it will ever happen, the question is worthy of examining nonetheless. Moreover, independently of whether its application becomes a reality, assessing willingness to introduce AI in policymaking could be a measure of systemic political disconformity, perhaps more subtle and less risky than other behaviors such as political extremism. Ultimately, we hope that with this research we can spark future inquiry about the AI- Politics interaction. Regarding our specific findings, we sought to uncover the role of anomie in promoting this support for the TecAI as our main goal. In three different studies, we found that anomie, and especially its subdimension of breakdown of leadership, are meaningfully associated with such support. We also tried to assess the causality of such a relationship in two different experiments, yet we only found support for a small effect of the breakdown of the leadership subdimension in an exploratory directional analysis. Thus, it seems that the role of anomie as a cause for increased public support for the TecAI is not supported by the data, but political dysregulation may have some causal effects. This suggests that the perception of the disintegration of the political system is more important than that of the decomposition of society in explaining this support for a TecAI. This finding contributes to existing literature, highlighting the worth of thoroughly examining anomie and its subdimensions, as they may potentially influence in different ways on political preferences (e.g., Crimston etal.,2022; Ionescu etal.,2021a). However, further studies should aim to replicate the causal link we found between the breakdown of leadership and support for TecAI, as it was only explored in the current research. Perhaps the causal role of anomie and its subdimensions on TecAI can be clarified by strengthening its manipulation. An intriguing option would be to ask people to provide their own reasons as to why society is deteriorating after having read the text pertaining to the assigned condition this could enhance the effects of the manipulation and hopefully produce clearer results. Another possibility we wish to elaborate is that, since our evidence of anomie as a predictor of TecAI is mostly correlation, it is the fact that a third variable is driving the effect. In this sense, we argue that political cynicism seems a good candidate for such a role. Political cynicism goes beyond anomie as, on top of potential beliefs about corruption or lack of competence, represents a deeply negative conception of politicians as inherently bad people (Pattyn etal.,2012). As such, cynics think not only that there is a crisis of leadership but also attribute such a crisis to the malicious nature and intent of politicians, potentially engulfing and driving the effects of anomie. Thus, we believe research on this possibility would be an interesting alley to explore for this novel field of inquiry. As for the role of our hypothesized moderators, we found no support for their importance as conditioners in the relationship between anomie and support for the TecAI. Nevertheless, we found valuable correlational insights that show that SDO is consistently associated with a preference for TecAI. Results regarding NCC, on the other hand, are mixed, since it only was significantly—and positively—associated with support for TecAI in our last study. The fact that our moderators did not yield statistically significant results is another issue that may have obscured the causal impact of anomie on TecAI. It is possible that the relationship between anomie and TecAI—for which we found robust correlational evidence—may be moderated by other variables, which constitutes an interesting avenue for future research. Research informs that people can attribute competence and warmth to AI (McKee etal.,2021), and perhaps some of these attributions may condition the causal impact of anomie. Among the rest of the variables that we measured in an exploratory fashion, the role of political orientation is especially notable. This variable was also significantly associated with support, with more conservative people showing greater support for the idea of a TecAI. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 22 | DONO and MORENO BELLA It is also key to highlight a piece of descriptive data. Despite the fact that this kind of data does not usually receive much attention, we think it is important in this case due to the novelty of the inquiry. Specifically, we want to stress that, in all three studies, support for the TecAI is slightly above the midpoint of the scale. We interpret this data as confirmation of the fact that for citizens, AI has potential to be used in political decisionmaking. Again, we would like to highlight that what we attempt to assess is whether citizens come to regard AI as a solution for the political problems of their societies, an even silver bullet against them. We have tried to do so from a neutral perspective, albeit acknowledging that it may pose a threat to democracy as currently constructed. The debate of whether AI can improve political decisionmaking needs from further research from several disciplines, and while some authors have argued against AI being used in politics (König & Wenzelburger,2022), others believe in its potential benefits (Cavaliere & Romeo,2022; Sætra,2020). Nevertheless, our research shows that perceiving a disintegrating society, especially in what pertains to political leadership, seems to promote a favorable idea among citizens about AI being entrusted with at least some governing responsibilities. Whether AI can do so in a way that is positive for our societies remains open to debate, but such uncertainty is even more reason to pay attention to whether the public would wish to go down this path, as it is a choice that would have a deep impact that could, at least potentially, be noxious. Whatever the case may be, we consider support for AI to be a signal of generalized discontent and/or disconnection with politics. Be that as it may, support for TecAI also implies a preference for political change, as prior research has also shown to be derived from anomie (e.g., Teymoori etal.,2016). We argue that, jointly, all these results constitute a good starting point to advance the research on the psychological determinants of public opinion on the use of AI in politics. Limitations and future directions As with any piece of research, our study is not exempt from limitations that have to be cautioned for proper interpretation of the data. First, it is important to notice that, despite a preliminary test of its psychometric qualities (see Supplementary Materials), the support for the TecAI measure has not been validated and published as such. It also measures what we define as a strong or full version of TecAI, as it does not consider a partial integration of AI within politics but rather assesses opinions on the fact of substituting politicians for an AI. As this is a new field of inquiry, having reliable and—ideally crossculturally—validated measurements would be a great step toward a better understanding of the topic. The fact that we measured this strong TecAI and got some null results should also encourage the research community to investigate whether our predictors could have a greater impact on a more subtle measure of support for AI integration in politics. Another limitation resides in the fact that, for Study 2, our manipulation of anomie seemed to not work properly. This could partially explain the absence of statically significant causal effects, as the manipulation check itself did not produce the expected results. Since its redefinition from a more psychological approach (Teymoori etal.,2016), relatively few experimental research has been conducted that manipulates anomie. This could mean that, despite the fact that we base our manipulation on a successful previous effort (Ionescu etal.,2021a), some nuances about how to best manipulate the construct may remain unknown. In our case, it could be that the control about the decay of art consumption could somehow also be interpreted by participants as a disintegration of society. Moreover, in cases where political discontent is high, perhaps it would be more appropriate to prime a moderately adequate (as we did in Study 3) functioning of society as a comparison. This would allow to generate meaningfully different groups, while in the absence of such a moderate condition, people may retain an already highly negative view on society. Also, 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 23 SUPPORT FOR A TECHNOCRACY OF AI and informed by the fact that its two subdimensions seem to have different effects, perhaps isolated manipulations of each should be encouraged. Finally, we think it is important to mention that all the studies have been performed in Spain while the country was under a leftwing government. This could, at least partially, account for the strong association observed between ideology and support for TecAI. We have found that being conservative is associated with this preference, yet with a leftwing government in office, this could indeed be biased by partisan interests. In the end, supporting a TecAI should require at least some degree of political discontent. Although anomie represents a more systemic disgruntlement, we cannot at the moment rule out a partisan hypothesis. Again, to disentangle whether this bias exists or if rather conservatives are more supportive of this hypothetical government of AI is, in our opinion, an exciting research topic. It is also important to note that results concerning ideological selfplacement should be interpreted with caution due to the use of a singleitem measure. Research conducted in countries where a rightwing party is in power could help shed light on this. In a similar vein, future research that examines crosscultural differences based on different sets of cultural values (Kingsbury & Avonius,2008) or that compares countries with different political regimes (i.e., authoritarian or postauthoritarian vs. democratic) seems in our eyes another interesting avenue for future research. CONCLUSION In the film The Matrix, humankind has been enslaved by “machines” that simulate their lives to make them pleasant while exploiting their resources. Although we do not augur for such a bleak future, AI poses certain challenges to our political systems. This lifechanging technology is being integrated into most human activities and thus could eventually be introduced in the exercise of politics. If we want to preserve a correct democratic functioning, we need to understand how people perceive and prefer to shape the AI- politics relationship. With this, we could be avoiding general discontent, negligence, and even foul use of AI as a tool for undemocratic groups to gain support for their cause. ACKNOWLEDGMENTS This article has been funded via a Seedcorn Grant by the European Association of Social Psychology. Additionally, we would like to thank Marco Brambilla for his feedback in the initial stages of the project and Miguel Moya for providing essential support in the pretests necessary for this publication. ORCID Marcos Dono https://orcid.org/0000-0002-6895-2433 Eva Moreno Bella https://orcid.org/0000-0002-1299-3148 REFERENCES Acemoglu, D., Autor, D., Hazell, J., & Restrepo, P. (2022). Artificial intelligence and jobs: Evidence from online vacancies. Journal of Labor Economics, 40(S1), S293–S340. Adler, N. E., Epel, E. S., Castellazzo, G., & Ickovics, J. R. (2000). Relationship of subjective and objective social status with psychological and physiological functioning: Preliminary data in healthy white women. Health Psychology, 19(6), 586–592. https:// doi. org/ 10. 1037// 0278- 6133. 19.6. 586 Beal, R., Norman, T. J., & Ramchurn, S. D. (2019). Artificial intelligence for team sports: A survey. The Knowledge Engineering Review, 34, e28. Bélanger, J. J., Pierro, A., Barbieri, B., De Carlo, N. A., Falco, A., & Kruglanski, A. W. (2016). One size doesn't fit all: The influence of supervisors' power tactics and subordinates' need for cognitive closure on burnout and 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 24 | DONO and MORENO BELLA stress. European Journal of Work and Organizational Psychology, 25(2), 287–300. https:// doi. org/ 10. 1080/ 13594 32X. 2015. 1061999 Bélanger, J. J., Pierro, A., & Kruglanski, A. W. (2015). Social power tactics and subordinates' compliance at work: The role of need for cognitive closure. European Review of Applied Psychology, 65(4), 163–169. https:// doi. org/ 10. 1016/j. erap. 2015. 05. 001 Campos, N., & Giovannoni, F. (2017). Political institutions, lobbying and corruption. Journal of Institutional Economics, 13(4), 917–939. Cavaliere, P., & Romeo, G. (2022). From poisons to antidotes: Algorithms as democracy boosters. European Journal of Risk Regulation, 13(3), 421–442. https:// doi. org/ 10. 1017/ err. 2021. 57 Centeno, M. A. (1993). The new leviathan: The dynamics and limits of technocracy. Theory and Society, 22(3), 307–335. Chatterjee, S., Khorana, S., & Kizgin, H. (2022). Harnessing the potential of artificial intelligence to foster citizens' satisfaction: An empirical study on India. Government Information Quarterly, 39(4), 101621. https:// doi. org/ 10. 1016/j. giq. 2021. 101621 Chow, A., & Perrigo, B. (2023). The AI arms race is on. Start worrying. Time. https:// time. com/ 62559 52/ aiimpac tchatg ptmicro softgoogle/ Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Academic Press. Crimston, C. R., Selvanathan, H. P., & Jetten, J. (2022). Moral polarization predicts support for authoritarian and progressive strong leaders via the perceived breakdown of society. Political Psychology, 44(4), 671–691. https:// doi. org/ 10. 1111/ pops. 12787 Devos, T., Spini, D., & Schwartz, S. H. (2002). Conflicts among human values and trust in institutions. British Journal of Social Psychology, 41(4), 481–494. https:// doi. org/ 10. 1348/ 01446 66023 21149849 Dong, Y., & Peng, C. Y. J. (2013). Principled missing data methods for researchers. SpringerPlus, 2, 1–17. https:// doi. org/ 10. 1186/ 2193- 1801-2- 222. Duckitt, J., & Fisher, K. (2003). The impact of social threat on worldview and ideological attitudes. Political Psychology, 24(1), 199–222. Duncan, L. E., & Stewart, A. J. (2007). Personal political salience: The role of personality in collective identity and action. Political Psychology, 28(2), 143–164. https:// doi. org/ 10. 1111/j. 1467- 9221. 2007. 00560. x Duriez, B., & Van Hiel, A. (2002). The march of modern fascism. A comparison of social dominance orientation and authoritarianism. Personality and Individual Differences, 32(7), 1199–1213. https:// doi. org/ 10. 1016/ S0191 - 8869(01) 00086 - 1 Durkheim, E. (1897). El Suicidio. CreateSpace Independent Publishing Platform. Enholm, I. M., Papagiannidis, E., Mikalef, P., & Krogstie, J. (2022). Artificial intelligence and business value: A literature review. Information Systems Frontiers, 24(5), 1709–1734. https:// doi. org/ 10. 1007/ s1079 6- 021- 10186 - w Faul, F., Erdfelder, E., Buchner, A., & Lang, A.- G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160. https:// doi. org/ 10. 3758/ BRM. 41.4. 1149 Fiske, A. P., & Rai, T. S. (2014). Virtuous Violence. Cambridge University Press. Frank, M. R., Autor, D., Bessen, J. E., Brynjolfsson, E., Cebrian, M., Deming, D. J., Feldman, M., Groh, M., Lobo, J., Moro, E., Wang, D., Youn, H., & Rahwan, I. (2019). Toward understanding the impact of artificial intelligence on labor. Proceedings of the National Academy of Sciences, 116(14), 6531–6539. https:// doi. org/ 10. 1073/ pnas. 19009 49116 Fredström, A., Parida, V., Wincent, J., Sjödin, D., & Oghazi, P. (2022). What is the market value of artificial intelligence and machine learning? The role of innovativeness and collaboration for performance. Technological Forecasting and Social Change, 180, 121716. https:// doi. org/ 10. 1016/j. techf ore. 2022. 121716 Gáspár, A., Giommoni, T., Morelli, M., & Nicolò, A. (2021). Corruption and Extremism (SSRN Scholarly Paper 3852839). https:// doi. org/ 10. 2139/ ssrn. 3852839 Gert, B., & Gert, J. (2017). The definition of morality. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy. Metaphysics Research Lab, Stanford University. https:// plato. stanf ord. edu/ archi ves/ fall2 017/ entri es/ moral itydefin ition/ Hague, R., & Harrop, M. (2004). Comparative government and politics. Springer. Hauser, D. J., & Schwarz, N. (2015). It's a trap! Instructional manipulation checks prompt systematic thinking on “tricky” tasks. SAGE Open, 5(2), 2158244015584617. https:// doi. org/ 10. 1177/ 21582 44015 584617 Hayes, A. F. (2022). Introduction to mediation, moderation, and conditional process analysis, third edition: A regressionbased approach. Guilford. Heydari, A., Teymoori, A., Nasiri, H., & Fardzadeh, H. E. (2012). Relationship between socioeconomic status, anomie, and authoritarianism. e- BANGI, 7(1), 177. Hogg, M. A., & Adelman, J. (2013). Uncertainty–identity theory: Extreme groups, radical behavior, and authoritarian leadership. Journal of Social Issues, 69(3), 436–454. 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 25 SUPPORT FOR A TECHNOCRACY OF AI Ionescu, O., Tavani, J. L., & Collange, J. (2021a). Perceived societal anomie, collective memory, and support for collective action: Perceiving that current French society is anomic influences present support for collective action through the reconstructed national past. Asian Journal of Social Psychology, 24(3), 405–420. https:// doi. org/ 10. 1111/ ajsp. 12438 Ionescu, O., Tavani, J. L., & Collange, J. (2021b). Political extremism and perceived anomie: New evidence of political extremes' symmetries and asymmetries within French samples. International Review of Social Psychology, 34(1), 21. https:// doi. org/ 10. 5334/ irsp. 573 Jonsson, O., & Luca de Tena, C. (2021). European tech insights 2021. Part II: Embracing and governing technological disruption. Center for the Governance of Change. https:// docs. ie. edu/ cgc/ IECGC- Europ ean- Tech- Insig hts- 2021- % 28Par t- II% 29. pdf Jost, J. T., Becker, J., Osborne, D., & Badaan, V. (2017). Missing in (collective) action: Ideology, system justification, and the motivational antecedents of two types of protest behavior. Current Directions in Psychological Science, 26(2), 99–108. https:// doi. org/ 10. 1177/ 09637 21417 690633 Kingsbury, D., & Avonius, L. (2008). Human rights in Asia: A reassessment of the Asian values debate. Pallgrave McMillan. König, P. D., & Wenzelburger, G. (2022). Between technochauvinism and humancentrism: Can algorithms improve decisionmaking in democratic politics? European Political Science, 21(1), 132–149. https:// doi. org/ 10. 1057/ s4130 4- 020- 00298 - 3 Kordzadeh, N., & Ghasemaghaei, M. (2022). Algorithmic bias: Review, synthesis, and future research directions. European Journal of Information Systems, 31(3), 388–409. https:// doi. org/ 10. 1080/ 09600 85X. 2021. 1927212 Kranzberg, M. (1986). Technology and history: “Kranzberg's Laws”. Technology and Culture, 27(3), 544–560. https:// doi. org/ 10. 2307/ 3105385 Kroh, M. (2007). Measuring leftright political orientation: The choice of response format. The Public Opinion Quarterly, 71(2), 204–220. Kruglanski, A. W. (2018). Violent radicalism and the psychology of prepossession. Social Psychological Bulletin, 13(4), e27449. https:// doi. org/ 10. 32872/ spb. v13i4. 27449 Kruglanski, A. W., Fernandez, J. R., Factor, A. R., & Szumowska, E. (2019). Cognitive mechanisms in violent extremism. Cognition, 188, 116–123. Kuljian, O. R., & Hohman, Z. P. (2023). Warmth, competence, and subtle dehumanization: Comparing clustering patterns of warmth and competence with animalistic and mechanistic dehumanization. British Journal of Social Psychology, 62(1), 181–196. https:// doi. org/ 10. 1111/ bjso. 12565 Lake, D. A. (2002). Rational extremism: Understanding terrorism in the twentyfirst century. Dialogue IO, 1(1), 15–28. https:// doi. org/ 10. 1017/ S7777 77770 200002X Laustsen, L., Petersen, M. B., & Klofstad, C. A. (2015). Vote choice, ideology, and social dominance orientation influence preferences for lower pitched voices in political candidates. Evolutionary Psychology, 13(3), 1474704915600576. https:// doi. org/ 10. 1177/ 14747 04915 600576 Leal Filho, W., Wall, T., Mucova, S. A. R., Nagy, G. J., Balogun, A.- L., Luetz, J. M., Ng, A. W., Kovaleva, M., Azam, F. M. S., & Alves, F. (2022). Deploying artificial intelligence for climate change adaptation. Technological Forecasting and Social Change, 180, 121662. Lee, M. K., Kusbit, D., Kahng, A., Kim, J. T., Yuan, X., Chan, A., See, D., Noothigattu, R., Lee, S., Psomas, A., & Procaccia, A. D. (2019). WeBuildAI: Participatory framework for algorithmic governance. Proceedings of the ACM on Human- Computer Interaction, 3(CSCW), 181:1–181:35. https:// doi. org/ 10. 1145/ 3359283 Lewin, K. (1936). Principles of topological psychology (F. Heider, Trad.). Martino Publishing. Makridakis, S. (2017). The forthcoming artificial intelligence (AI) revolution: Its impact on society and firms. Futures, 90, 46–60. https:// doi. org/ 10. 1016/j. futur es. 2017. 03. 006 Manin, B. (1987). On legitimacy and political deliberation. Political Theory, 15(3), 338–368. McKee, K., Bai, X., & Fiske, S. (2021). Humans perceive warmth and competence in artificial intelligence. PsyArXiv. https:// doi. org/ 10. 31234/ osf. io/ 5ursp Milana, C., & Ashta, A. (2021). Artificial intelligence techniques in finance and financial markets: A survey of the literature. Strategic Change, 30(3), 189–209. Nicol, A. A. M. (2009). Social dominance orientation, rightwing authoritarianism, and their relation with leadership styles. Personality and Individual Differences, 47(6), 657–661. https:// doi. org/ 10. 1016/j. paid. 2009. 06. 004 Oldmeadow, J., & Fiske, S. T. (2007). Systemjustifying ideologies moderate status = competence stereotypes: Roles for belief in a just world and social dominance orientation. European Journal of Social Psychology, 37(6), 1135– 1148. https:// doi. org/ 10. 1002/ ejsp. 428 Palan, S., & Schitter, C. (2018). Prolific.ac—A subject pool for online experiments. Journal of Behavioral and Experimental Finance, 17, 22–27. https:// doi. org/ 10. 1016/j. jbef. 2017. 12. 004 Panch, T., Mattie, H., & Atun, R. (2019). Artificial intelligence and algorithmic bias: Implications for health systems. Journal of Global Health, 9(2), 020318. https:// doi. org/ 10. 7189/ jogh. 09. 020318 14679221, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.13048 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License