Governing Artificial Intelligence (AI) Through Creating Governance Norms: Identifying the Emerging Global AI Governance Norms.

Governing Artificial Intelligence (AI) Through Creating Governance Norms: Identifying the Emerging Global AI Governance Norms. | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF open-forum Governing Artificial Intelligence (AI) Through Creating Governance Norms: Identifying the Emerging Global AI Governance Norms. Muhammed Hasan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9383807/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Efforts to establish a global Artificial Intelligence (AI) regulatory regime are occurring in various locations through transnational debates and governance proposals. Yet, our understanding of the dynamics of a global AI governance regime is unclear. This study identifies emerging global norms governing the development and deployment of AI technologies. I examine (a) 345 policies, white papers, and executive orders published by fifty-nine states; (b) 873 AI governance-related debates, conferences, and discussions hosted by or on the platforms of fifteen international organizations between January 2000 and January 2022; and (c) 87 AI governance frameworks proposed by 54 non-state actors. Utilizing G. H. von Wright’s norm analysis framework, I identify 1,028 prescriptive norms that permit, obligate, and prohibit AI development and deployment practices. I then categorize them using the Ideal Type Analysis method, suggesting five constitutive norms emerging: ‘AI-fication,’ ‘datafication,’ ‘trustworthiness,’ ‘anthropocentrism,’ and ‘enviro-centrism.’ These emergent norms will influence AI technologies in two ways. First, they will regulate human behavior and create new actors through socialization processes regarding how AI technologies are developed and used. Second, since norms function as secondary rules, these five AI governance norms will define and influence the dynamics of an AI governance regime globally. Governance of Artificial Intelligence Norm Emergence AI Governance Norms Technology Governance Norm Identification. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Previous AI studies into global AI governance are mostly normatively grounded – that is, they analyze the risks and benefits of AI technology and propose governance solutions based on their expectations of what should be done, which sometimes extrapolate governance mechanisms that have evolved in similar technologies. Empirical studies, such as Bode (Bode 2023 ), however, focus on specific cases of AI in a particular location that examine the emerging themes in AI governance and thus lack a comprehensive picture of society’s collective expectations about AI governance. My argument is that global AI governance is already happening. According to the Organization for Economic Cooperation and Development’s (OECD) AI Observatory (OECD 2021 ), around sixty countries have released over nine hundred strategic documents. Besides, non-state actors have also published AI governance guidelines. These discursive practices indicate emerging patterns of behavior and stakeholders’ preferences regarding how this technology may or may not be developed and used by problematizing various issues and setting a governance agenda to address them (Rosert 2019 ). In this study, I examine these AI governance discourses to identify emerging global AI governance norms. Norms are understood as socially agreed-upon expectations consisting of practices and rules defining appropriate and inappropriate behavior that shape and regulate behavior (Ullmann-Margalit 1977 ; Wright 1963 ; Finnemore and Sikkink 1998; Jepperson et al. 1996 ; March and Olsen 1998 ). Norms influence behavior through two pathways. One way is by regulating human actions. This type of norm is known as a regulative norm, which guides an already-existing activity and is intended to have causal effects on human behavior (Ruggie 1998 ). Regulative norms are sometimes referred to as prescriptive norms (Katzenstein 1996 ), as they guide human behavior by providing dos and don’ts. Henceforth, I use the term ‘prescriptions’ to indicate these norms. Wright (Wright 1963 ) posits that prescription operates in three ways to determine social actions: it permits, obligates, or prohibits certain social practices over others. He suggests that these normative expectations are expressed in language, conveying that “something ought to, may, must not be or must be done” (Wright 1963 , 100; Opp 1982 ). The second pathway through which norms affect human behavior is called constitutive norms, which define the practices that constitute a particular class of consciously organized social activity. Constitutive norms specify what qualifies as that activity, while regulative norms dictate how a specific activity can be performed. Constitutive norms form the institutional foundation of all social life, without which no consciously organized realm of human activity can be envisioned (Ruggie 1998 ). They create new actors, interests, or categories of action (Finnemore and Sikkink 1998). Constitutive norms can be interchangeably referred to as principles (Hart 1994 ). Lawless and her colleagues (Lawless et al. 2020 ) argue that constitutive norms, or principles, typically represent fundamental global principles that emerge as aspirational goals through international agreements and guidelines, providing normative guidance for best practices. Henceforth, I use ‘constitutive norms’ to indicate these aspirational goals that contain a set of prescriptions. Thus, this study accomplishes two tasks. First, it identifies prescriptions by examining AI governance related documents published by state and non-state actors. I accomplish this task by employing Wright’s ( 1963 ) norm analysis framework. Second, I construct nonabstract generalizable conceptual typologies of these identified prescriptions into five constitutive norms. These are: (1) AI-fication, (2) Datafication, (3) Trustworthiness, (4) Anthropocentrism, and (5) Enviro-centrism. To construct the conceptual typology, I employed the ‘ideal type analysis method’ (Gerhardt 1994 ). Each of the constitutive norm holds and describes the set of prescriptions that permit, obligate, or prohibit specific AI development and deployment practices. 1.2 Significance of the study From a functional perspective, norms regulate human behavior, having prescriptive and proscriptive effects on our social actions (Wright 1963 ). This is particularly true for the international political system, which is anarchic and structurally non-hierarchical. Global politics is marked as much by power maximization as it is by morality (Carr 1946 , 89). The prescriptive and constitutive norms presented here are at the early stages of their emergence (Finnemore and Sikkink 1998). These are and will continue to be debated and negotiated in various contexts involving state and non-state actors. Therefore, I do not claim that these norms are the benchmark of a global AI regulation since they are yet to be formalized into codified regulations. However, I argue that, in the absence of an overarching central government, these norms would influence actors in articulating their preferences about a global AI governance architecture (Goldstein and Keohane 1993 ; Keohane 1989 ; Katzenstein 1996 ; Legro 1995 ; Kratochwil 1989 ). In addition, these norms will provide justifications for political actions by acting as guiding principles, drawing clear distinctions between appropriate and inappropriate behaviors, and setting the standard of behavior for future AI governance actions (Risse 2002 ; March and Olsen 1998 ; Goldstein and Keohane 1993 ; Keohane 1989 ; 1984 ). In doing so, these norms will provide the background and lay the foundations for formalized and codified AI governance mechanisms (Tieku 2019 ). 1.3 Contribution of the study This study makes several contributions. First, there is no global AI governance mechanism that exists today except the regional-level EU AI Act (The AI Act of the European Parliament and of the Council 2024 ). Therefore, our knowledge about the dynamics of global AI governance architecture is limited. This study fills this gap by systematically examining individual AI development and deployment policies, shedding light on the normative foundations of the nature and direction of such a potential governance regime. Since norms function as informal rules (Tieku 2019 ), the five constitutive norms and the prescriptions within them will play the roles of secondary rules that would shape the formulation of primary types of laws (Hart 1994 ) in the form of treaties, conventions, or institutional arrangements like those that exist in telecommunication and internet governance. Even though there are several pathways, such as the one-country-one-vote, weighted voting to advantage major powers, weighted voting to advantage vulnerable countries, and double-weighted majority voting (Gupta et al. 2024 ), I do not claim how and when these norms may be institutionalized. However, this study suggests that a consensus is emerging among state and non-state actors regarding the norms presented here. Therefore, I argue that these emergent norms may reach a tipping point as an increasing number of actors begin to act in accordance with them, thus triggering a norm cascade, which may occur through various socialization processes (Finnemore and Sikkink 1998; VanDeveer 2004 ; 2011 ; Nadelmann 1990 ; Price 1998 ), hence becoming institutionalized through their adoption in domestic and international formal legal codes. Second, Prior AI governance studies primarily focus on exploring frequently discussed themes within national AI policies, understanding how ethics are integrated into AI governance, and examining governance methods in specific countries (Radu 2021 ; Gherhes et al. 2023 ; Saheb and Saheb 2023 ; Nelson and Gorichanaz 2019; Stix 2021 ; Palladino 2023 ; Tan and Taeihagh 2021b; 2021a). The thematic analyses do not effectively capture society’s collective expectations about what actions are or are not permissible or prohibited. Contrarily, due to prescriptive and constitutive effects, norms dictate what individuals can and cannot do while also constituting new actors and identities through socialization processes. Once a norm emerges, it undergoes a life cycle that includes emergence, cascade, and, ultimately, institutionalization through its incorporation into formal national and international laws (Finnemore and Sikkink 1998). Third, though studies into norms date back at least several centuries across multiple disciplines, the question of ‘how we know a norm when we see one’ (Finnemore and Sikkink 1998) remains puzzling. Constructivists study the roles, evolution, and diffusion of norms in global politics using interpretive methodologies, such as genealogy (Kowert and Legro 1996). In the process, they ignore the issue of emergence. On the other hand, positivists study when, how, and what types of norms may emerge by utilizing methodological individualism, such as game-theoretic models (Axelrod 1986 ; Ullmann-Margalit 1977 ; Coleman 1990 ). However, they too forget that social change happens through social interaction. So does norm emergence. This study presents an interpretive quantification method for examining and identifying norm emergence through a systematic analysis of language because languages are the social reality, practices, and social actions embedded in a particular social context (Fairciough 2013 ; Habermas et al. 2005; Wodak 2014 ). Therefore, language, being the social practices themselves, is what people say, and what they do in practice, carrying the meaning-in-use of the norms (Milliken 1999 ; Wiener 2009 ). In what follows, in Section Two, I discuss the methodology, define the terms, present the types and sources of data, and finally discuss the methods of identifying the prescriptions and conceptual typology development. In the Third Section, I introduce and define the five constitutive norms. In the Fourth Section, I elaborate on each of the constitutive norms and prescriptions within them, highlighting prescriptive and proscriptive effects as well as shedding light on agreements and contentions among actors. The article concludes with a discussion of future directions for research on AI governance. 2. Methodology I employed an interpretive quantification method to identify the prescriptions. This design effectively produces intersubjective and relational knowledge (Barkin and Sjoberg 2017 ). It is interpretive in that prescriptions are promulgated exclusively through language (Wright 1963 ), and thus, a systemic interpretation of language uncovers the meaning of prescriptions and their effects. It is quantitative because I assessed the resonance of the identified prescriptions by analyzing their frequency across various types of actors. I define AI as the art, science, and engineering of creating machines that act like humans and perform tasks better than humans (Turing 2009 ). Though AI could be deployed in different domains, the fundamental building blocks and their functionalities are the same - automation. This study considers AI as a technology for automation, rather than specific use cases. I define global governance as an “intersubjectively recognized purposive order consisting of a system of formal rules embodied in institutions or informal patterned regularities observed through individuals’ habitual actions that define, constrain, and shape actor expectations and conduct in a given issue domain at the global level” (Biersteker 2015 ) I define global as being an intercultural interaction between agents of different roots (Wiener 2014 ). Understanding this way includes state and non-state actors who are interacting with one another through which patterns of behavior emerge (Zürn 2018 ; Wiener 2014 ). I adopt Karl-Dieter Opp’s definition of emergence. Opp ( 1982 ) argues that “if we explain the emergence of norms, we explain under what conditions individuals express new normative statements” (61). Finally, I adopt the norm definition offered by Wright ( 1963 ) and Opp ( 1982 ), who define norm in terms of its effects that something ought to or may or must not be or must be done. 2.1 Data I have used three types of data for analysis: 1) states’ AI policy documents, 2) AI documents put forward by non-state actors, and argumentative documents. The non-state actors involve multinational tech corporations, civil society groups, academia, AI research institutes, and International Organizations (IOs). Argumentative documents include AI governance-related seminars, debates, and inter-governmental meetings that took place on international organizations’ platforms participated in by either state or non-state actors. Several reasons for including diverse documents. First, prescriptions are promulgated exclusively through language (Wright 1963 ). Therefore, the inclusion of diverse documents is helpful for a comprehensive analysis, irrespective of the nature and characteristics of the documents themselves. Second, since norms indicate society’s collective expectations, such inclusion would suggest AI stakeholders’ collective expectations and preferences. Finally, the inclusion of three types of documents captures the dynamics of global governance, which is a system of formal rules and informal patterned regularities. I collected this data primarily from the OECD’s digital library. Its AI Policy Observatory archives states’ AI initiatives on the National AI Policies and Strategies page; conversations, meetings, and AI-related seminars on the AI Policy Events page; and it stores non-state actors’ AI-related activities on the OECD’s Stakeholders’ Initiative page within the same digital library. The video recordings of meetings and conferences on IO’s platforms span from January 2000 to January 2022. I downloaded the recordings using “4K Downloader” and “iTubeGo” software. I have used NVIVO and “Ötter.ai” software to transcribe the recordings. Additionally, I conducted an open Google search, first to locate documents listed but missing in the OECD’s archive and second to find additional AI governance related documents that may not be stored in the AI Policy Observatory archive. I used a total of 345 AI policies from 59 states for the final analysis out of 900 documents, 873 recordings collected from 15 international organizations, and 87 AI guidelines from 54 distinct non-state actors for this analysis. 2.2 Analysis The analysis proceeded in two stages. The first stage involved identifying prescriptions, and the second stage involved constructing typologies of the prescriptions into constitutive norms. 2.2.1 Identifying Prescriptions I utilized Wright’s ( 1963 ) norm analysis framework to identify the prescriptions. Wright ( 1963 ) proposes both morphic and semantic analysis of language to distinguish between various types of prescriptions. According to Wright, a prescription may be promulgated using imperative, indicative sentences, or deontic modal verbs. However, Wright ( 1963 ) contends that “neither as a morphological nor as a semantic category is the notion of the imperative mood clear and homogeneous enough to make even a provisional identification of prescriptions plausible” (99). In addition, Wright ( 1963 ) suggests that indicative sentences are typically used in the future tense, and their use as a prescription is more common in legal language. It follows trivially that every prescription can be expressed in a deontic sentence,” (Wright 1963 , 100), which is promulgated exclusively through deontic modal verbs—‘may,’ ‘ought,’ and ‘must.’ Thus, I first isolated the deontic sentences based on morphic criteria of deontic modal verbs using NVIVO data analysis software in which I created a ‘text search’ query using the words, ‘may’, ‘may not’, ‘must’, ‘mustn’t’, ‘ought’, ‘ought to’, ‘ought not to’ including their stemmed words. Following Wright ( 1963 ), I conducted a semantic analysis. To do that, I analyzed the identified deontic sentences on several semantic dimensions: (1) isolating sentences containing the deontic modal verb ‘may’ to determine whether the sentence indicates permission or possibility; (2) separating prescriptions from ideal rules based on whether the deontic modal verbs are conjoined with the auxiliary verb ‘be’. According to Wright ( 1963 ), ideal rules are conjoined with the auxiliary verb ‘be’ along with deontic modal verbs, such as the form of ‘may be’. (3) Since I am concerned with the identification of prescriptions relevant only to the actions and activities in the social domain, I separated the prescriptions belonging to the cognitive domain by using a taxonomy of cognitive verbs (Krathwohl 2002 ). The third stage involved examining whether the deontic sentences contain the six mandatory components that set apart prescriptions from other types of norms (Wright 1963 ). These are: (1) the character indicating permission, obligation, or prohibition; (2) the content to indicate the prescribed actions and activities; (3) the conditions of applications to determine whether a prescription is categorical or hypothetical; (4) the authority meaning the prescription givers; (5) the subject meaning the prescription takers; (6) and the occasion to indicate time, scope and location. Based on the analysis, I identified 1,028 prescriptions guiding the development and deployment of AI. I determined whether a prescription resonated based on its frequency among actors. Since collective agreement serves as the ontological foundation for a prescription to exist, I have used a minimum criterion that ‘for a prescription to exist, it had to be promulgated by at least two or more state and non-state actors.’ Any prescription meeting these criteria is considered to resonate among all actors and, therefore, considered to be emerging. 2.2.2 Constructing Conceptual Typologies The second stage of the analysis involved constructing typologies of the identified prescriptions into constitutive norms. This construction is required to solve the problem to be able to say, ‘what this prescription is an example of’. A typology can be theoretical or descriptive (Doty and Glick 1994). However, I am focusing on developing a conceptual nonabstract generalizable typology by creating a hierarchical system of categories to organize objects according to their similarities and differences (Mandara 2003 ; Bailey 1994). The main task of this typology is to describe the dimensions and attributes of prescriptions and find ideal types to which those dimensions and attributes may belong. I employed the ideal type analysis method (Gerhardt 1994 ). This method, sometimes known as a qualitative clustering method, involves a systematic comparison of cases or participants within a qualitative dataset to create ideal types or groups of similar cases. Together, these ideal types form a typology (Stapley, O’Keeffe, and Midgley 2022, 2; Stapley et al. 2021). I have used the already identified 1028 prescriptions as a data basis for this analysis. First, a “criterion” needed to be determined. “These criteria define what a case must be like that fulfills in optimal fashion the characteristics postulated as indicators of the relationship under investigation” (Gerhardt 1994 ; 100). In this study, I am using ‘actions and activities’ associated with AI governance and their stated/intended purpose as my criterion since prescription is intended to achieve some goals by permitting some actions while prohibiting others. Additionally, typology could be constructed using two constructs (Bailey 1994; Doty and Glick 1994). Given the nature of the data, I used three constructs – indicators, dimensions, and the ideal type – to construct the constitutive norm type. 2.2.3 Identifying Indicators Indicators are the empirical reality (Babbie 2016 ). Since I am concerned with prescriptions that order, prohibit, or permit actions or activities, the indicators are the contents of a given prescription. I identified the contents using two linguistic clues: (1) nouns and their various forms in a sentence formation, and (2) verbs and their various forms. From a syntactical perspective, the content of a norm is the ‘object’ of a sentence or the verb in the sentence. The answer to the question ‘what is to be done, and what is prescribed?’ is the content of a prescription. 2.2.4 Identifying Dimensions The second stage involved grouping the indicators into dimensions. Since a single word can be used in different contexts, the only separating matrix is the goals that the actions are intended to achieve. Usually, the sentence structures, such as infinitives, infinitive phrases, and infinitive clauses, hold the clues to such goals. Additionally, I have examined the sentences and paragraphs that appeared before and after the prescriptions to contextualize and make an interpretation of the intended goals by asking ‘what this action/activity is aimed at’? 2.3 The Ideal Norm Types/Norm Typologies The final step involved developing the norm typologies. Gerhardt ( 1994 ) suggests that the ideal type must be constructed based on the pre-set criterion for typology development. The criterion I have set is the rational-purposive social actions, which are ‘means to the ends.’ Finding conceptual names was an iterative process. The chief question I kept asking for each of the dimensions is this: to what purposes such and such prescriptions are promulgated? Contextualization – by reading the prescriptions in the context in which they were written, and existing literature was instrumental in constructing the concepts. The benchmark I set is to determine which concepts can best capture and describe the dimensions and indicators. The tasks of coming up with the terminologies were more conceptual than technical. The dimensions and indicators within each constitutive norm type are neither exhaustive nor mutually exclusive. Such overlap occurs both within and across norm types for two main reasons. First, the AI is a multifaceted and multi-layered technological system (Gasser and Almeida 2017), and so are its applications. Consequently, the context in which a prescription is promulgated vary significantly. For example, the Datafication norm includes prescriptions such as ‘generate value’ from data to advance AI development and ‘protect privacy’ to foster data flow through a social contract between data owners and collectors. Though these prescriptions differ in contexts, they serve the same purpose –datafication. Moreover, data’s diverse and expanding utility beyond AI development and automation adds to the heterogeneity of prescriptions within the Datafication norm. This applies to other norm type as well given the multi-layered nature of AI. Similarly, the prescription ‘enact law’ appears across AI-fication, Datafication, and Trustworthiness norm, yet its purposes differ: enabling AI development at scale, building data infrastructure, and creating conditions for trustworthiness. Secondly, from a methodological standpoint, the indicators and dimensions of ideal types need not be exhaustive and mutually exclusive (Utech 1963; Gerhardt 1994 ; Stapley et al. 2022; 2022). Rather, each should capture and reflect the description of the ideal type to a greater degree (Stapley et al. 2022). 3. The Five Constitutive Norms Following the ideal type analysis method, this study categorize the 1028 identified prescriptions into five constitutive norm types: (1) AI-fication, (2) Datafication, (3) Trustworthiness, (4) Anthropocentrism, and (5) Enviro-centrism. Figure 1 illustrates the frequency of each norm and the actors promoting them, while Fig. 2 displays the frequency of normative effects for each norm. In what follows, I define the concept of each constitutive norm before turning to discuss the prescriptions contained within them. 3.1 AI-fication I use the term AI-fication to describe the dimension and indicators related to the development and deployment of AI. I define the concept in terms of the prescriptions aimed at developing and deploying AI technologies at scale. These included creating an ecosystem, investing in infrastructure, and adopting new regulations to facilitate AI development or modify existing ones. Here, AI-fication is the ends. The dimensions and indicators are means to those ends. Technologization and digitalization could have been used instead of AI-fication. However, technologization is a broader concept that encompasses technologies beyond AI, while digitalization can be seen as just one of the indicators of AI. 3.2 Datafication I use the term ‘Datafication’ to describe the prescriptions related to data collection, sharing, processing, and maintaining data quality. The term “datafication” was introduced by Mayer-Schönberger and Cukier (Mayer-Schönberger and Cukier 2013). According to them, datafication refers to assigning value to information or facts (Mayer-Schönberger and Cukier 2013). “To datafy a phenomenon is to convert it into a quantified format so it can be tabulated and analyzed” (Mayer-Schönberger and Cukier 2013, 82). However, this definition does not capture the full range of dimensions and indicators that have emerged. They emphasized one aspect of valuation that pertains to the extensive collection and processing of data due to its value to different groups for various reasons. However, data valuation is not a one-way street. It requires a relationship of trust between the data collector and data subject whereby rights and responsibilities could be ensured for each stakeholder. Therefore, data valuation should not only refer to valuation in the strict sense of commodification or quantification but also to the associated rights and responsibilities. Hence, I define datafication as attempts to assign value to data as public or private assets and categorize the prescriptions aimed at ensuring the free flow of data under constitutive norm datafication . 3.3 Trustworthiness The third emerging AI governance norm is trustworthiness. There is confusion between ‘trust’ and ‘trustworthiness’ regarding which concept could be considered the norm (Hardin 2002 ; Bicchieri et al. 2011). Trustworthiness refers to both cognitive and non-cognitive situations in which “B ( AI ) is trustworthy regarding A ( human ) in the domain of interaction D ( human-AI interactions ) if and only if she ( B, that is, AI ) is competent in that domain, and she ( B, that is, AI ) considers the fact that A ( human ) is relying on her, should A ( human ) choose to do so in this domain, to be a compelling reason for acting as expected,” (Jones 2012 , 70–73). This concept of “three-place trustworthiness” suggests two strategies for promoting it: 1) increasing the prevalence of “ motivational structures” that fundamentally enable responsiveness to dependency; 2) reducing the field of “ competing considerations” so that responsiveness to dependency will often carry the day,” (Jones 2012 , 73). The conceptual clarification follows Hardin's distinctions between trust and trustworthiness (Hardin 2002 ). Hardin ( 2002 ) argues that “trust is little more than knowledge; trustworthiness is a motivation or a set of motivations for acting” (Hardin 2002 , 31). To establish an environment of trustworthiness or the motivations for it, the creation of institutions is one of the processes through which trust is supported and fostered. In the context of AI governance, there appear to be prescriptions aimed at creating or modifying motivational structures by adopting new policies and practices of data collection to promote or foster trust among developers, deployers, and users, thereby enhancing the motivational framework for trustworthiness. I categorized these prescriptions aimed at enabling human responsiveness to dependence on AI as constitutive norm of Trustworthiness. 3.4 Anthropocentrism The fourth emerging AI governance norm is anthropocentrism. Anthropocentrism refers to the concept that puts humans at the center of everything. Therefore, all other beings are “means to humans’ ends” (Kopnina et al. 2018 ; Crist and Kopnina 2014). However, putting ‘Anthropos’ at the center of everything is contested. At the heart of the debate is the issue of assigning agency and whether humans should be put at the center of material and ethical concerns (Crist and Kopnina 2014). Anthropocentrism, understood from the ethical perspective, would mean that humans are at the center, and everything is a means to human ends, but that everything would require preservation and nurturing for the sake of humans’ survival. On the other hand, from a material perspective, anthropocentrism would mean that humans are the center of everything, and at the expense of others. Therefore, this concept of anthropocentrism regards everything else as subhuman and not worthy, and that they can be destroyed, dislocated, and need to be dominated and controlled (Crist and Kopnina 2014). In this study, I do not define anthropocentrism from either a material or an ethical perspective of putting humans at the center. Instead, I adopt the definition of anthropocentrism more broadly and let the data speak for itself. Hence, I categorized prescriptions promulgated to serve humans and place humans at the center as they relate to AI under constitutive norm Anthropocentrism. 3.5 Enviro-centrism Like anthropocentrism, some prescriptions related to AI were promulgated that focus on ecology. These could be categorized as eco-centric. However, Gellers (Gellers 2021 ) argues that Eco-Centrism as a concept prioritizes the entire natural ecosystem or individual organisms within it. From this holistic perspective, the empirical evidence suggests that the prescriptions made to define the scope of the relationship between nature and AI systems do not truly aim to extend moral and ethical rights to either AI systems or the ecology involved in the AI-environment relationship. Therefore, they cannot be accurately characterized as ecocentrism. Therefore, I categorize the prescriptions that guide AI’s development, deployment, and interaction with the natural environment under constitutive norm Enviro-centrism. 4. Understanding the Prescriptions of Five Constitutive Norms 4.1 The Prescriptions of AI-fication The prescriptions categorized under AI-fication belong to three dimensions: (1) governance and regulation, (2) development and deployment, and (3) infrastructure development. These prescriptions prescribe how AI can be developed and utilized, how to regulate AI, and how to strengthen infrastructural capabilities. Table − 1 below summarizes the dimensions and prescriptions within AI-fication. Table 1 Dimensions and the Prescriptions of AI-fication. Dimensions of AI-fication Prescribed Norms/Indicators Governance & Regulations Adaptive Enact Laws Global Regulation Proportional Regulation Private Regulation State Regulation Compliance Participatory Governance Regional Regulation Dev. & Dep. Approach Cocreation Technological Sovereignty Exploit Potential Patronization Proportional Dev. & Dep. Exploit Data Equitability Technologization Public Awareness Digitization Logistics/Infrastructure Eliminate Barriers Develop Infrastructure Investment Dev. Human Resources Interoperability Safety Net Two conclusions can be drawn from the prescriptions of AI-fication. First, the development of AI technology at scale will continue to grow. This applies to both civilian and military uses of AI, even in historically less technologically advanced countries. Second, the prescriptions of AI-fication suggest a flexible governance over rigid regulations. This preference arises because AI technology is still evolving, and its full benefits and harms are yet to be understood. However, there are considerable contentions and zones of agreement among state and non-state actors. The following network Fig. 3 represents the frequency of actors promulgating a particular prescription that highlights the contentions and agreements. 4.1.1 Permitted Actions for Developing AI ‘Co-creation,’ ‘proportionality,’ and ‘equitability’ are the three most promulgated prescriptions. Both state and non-state actors have prescribed that AI development and deployment must occur through collaborative processes between AI developers and state actors, as well as between the developers and end-users. Furthermore, they have prescribed that AI’s development and deployment must be proportional and equitable. Being proportional means maximizing AI’s benefits while mitigating its risks. Additionally, it is prescribed that state and non-state actors must ‘leverage AI’s potential’ and achieve ‘technological sovereignty’ through AI development. They have prescribed ‘mass technologization’ and ‘state patronage.’ Among the actors promulgating these norms are the United States, the United Kingdom, Sweden, Denmark, Finland, France, the European Union, the OECD, Singapore, Germany, IEEE, and Microsoft Inc. 4.1.2 Building AI Infrastructure AI’s development, however, does not occur without the appropriate infrastructure. The prescriptions for advancing AI infrastructure encompass activities such as developing physical infrastructure, making investments, and cultivating human resources. The two most promulgated prescriptions from state actors are ‘human resources’ and ‘infrastructure development.’ States have directed relevant stakeholders to create training tools, implement worker re-skilling programs, and redesign education systems. Some states, including Belgium, have prescribed amending existing immigration laws to attract students and skilled AI workers from around the globe. 4.1.3 Permitted Regulatory Approaches ‘Adaptability’ in AI regulation and ‘participatory’ governance are two prescriptions promulgated by both state and non-state actors, including the EU, OECD, IEEE, and the United States. These two prescriptions highlight the importance of including all relevant stakeholders, such as the public sector, industry, and end-users, emphasizing the collaboration of scientific experts and policymakers in technology governance. 4.1.4 Prohibited AI Development, Regulatory, and Infrastructure Approaches Almost all the prescriptions of AI-fication involve permission, command, and obligation. Only a few prohibitive prescriptions are promulgated within AI-fication. These prohibitive prescriptions restrict actions that could hinder the development and deployment of AI, including restrictive regulation or insufficient investments, clearly prohibiting any actions that may inhibit AI innovation. Nonetheless, there is evidence, particularly in the EU, which suggests that AI development and deployment might be prohibited when it poses risks to human rights, democracy, and the rule of law. Therefore, future prohibition on AI may be expected when AI presents clear threats to human rights, democracy, and the rule of law. 4.2 The Prescriptions of Datafication The prescriptions categorized under datafication belong to five dimensions: (1) approaches to data, (2) governance and regulation, (3) infrastructure/capacity, (4) data trust, and (5) data quality. These prescriptions guide how data may be perceived, regulated, infrastructure capabilities may be developed, and measures to enhance trust in and the quality of data. Table 2 below summarizes the dimensions in the left column and the prescriptions within each dimension in the right column. Table 2 The Dimensions and the Prescriptions of Datafication. Dimensions of Datafication Prescribed Norms/Indicators Approaches to Data Generate Value Commodification Generate Data Data Sovereignty Governance & Regulations Enact Laws Compliance Monitoring & Assessment Litigation Infrastructure/Capacity Dev. Data Infrastructure Management Data Resources Data Sharing Data Partnership Data Trust Transparency Data Subjects’ Sovereignty Privacy Data Justice Data Protection Proportionality Accountability Disclosure Profiling Awareness Data Quality Data Standardization Accuracy and Reliability Integrity Remove Noise Representative Source Integrity Validity The prescriptions of datafication suggest, since the development of AI relies on data, that there are normative tensions between the prescriptions of datafication and the norms of human rights, especially concerning privacy, individual liberty, and sovereignty. This holds true in both democratic and non-democratic countries. The following network Fig. 4 highlights these contentions among actors and their governance preferences. 4.2.1 Permitted Approaches to Data It is permitted to treat data as having intrinsic value. Notably, states prescribe to view data as ‘strategic assets’, ‘a commodity’ that can be produced, assigned property rights, and sold to intended buyers under state control. Some states have commanded the creation of ‘data commons’ to generate data and to be data-sovereign nations. 4.2.2 Prescribed Approaches to Building Infrastructure Maintaining a data collection ecosystem requires proper infrastructure and capacity. The prescriptions for developing data infrastructure primarily comes from states, which see this as an obligation, acting individually or collaboratively through mechanisms like a union-wide data hub in the EU. Both state and non-state actors have prescribed for data sharing within and across borders. This includes countries from EU member states to the USA, UK, UAE, and India. EU member states prescribed to share personal data collected within their territories with other EU members. Additionally, the EU GDPR prescribed member states to share personal data with countries outside the region and with international organizations through bilateral or multilateral agreements. These prescriptions suggest a consensus around state-sponsored data collection and sharing mechanisms and possibly a forthcoming international agreement regarding these practices. 4.2.3 Prescriptions of Data Trust State and non-state actors promulgated prescriptions—such as data privacy, data protection, proportionality, data justice, and the autonomy of data subjects. These prescriptions are aimed at fostering a relationship of trust between the data collectors and data subjects to facilitate data collection processes. These prescriptions are promulgated by technologically advanced, economically prosperous, and liberal democratic countries, such as the USA, the UK, and the EU member states. The prescriptions of ‘data privacy,’ ‘data protection,’ and ‘proportionality’ are promulgated so data subject may trust data collector. Two prescriptions— ‘data disclosure’ and ‘data subject’s sovereignty’—indicate a shift in how states manage the confidentiality of the information they collect. For instance, the prescription of data subjects’ sovereignty prescribes that data owners must have complete control over their personal information. This implies that any data collection, whether public or private, must obtain consent before collecting personal data, inform data subjects of any processes performed on their data, and ensure that data subjects always have access to their information. 4.2.4 Prescriptions for Data Governance State and non-state actors prescribed requiring data collectors to ‘comply with’ applicable privacy laws and prescribing to ‘enact laws.’ ‘Monitoring and assessment’ is another prescribed obligation for state-appointed data stewards and prescribing non-state actors, particularly technology corporations, to have their own data stewards. 4.2.5 Prescribed Prohibitions of Datafication There are some prescriptions categorized under datafication. The EU, India, and the UK forbid any organization, whether public or private, from discriminating based on race, color, or religion or from obstructing access to services based on the personal data they hold about an individual. The EU, the UK, India, the Trade Union Congress (TUC), and the USA prescribed that no organization, public or private, should collect personal identifiers during data collection, nor disclose any personal data without obtaining clear consent from the data subjects. Singapore prescribes that consent cannot be forced or manipulated. It also prohibits organizations from disclosing when an individual’s data is shared with law enforcement and bans entities from transferring personal data outside the country. The EU prohibits the transfer and sharing of EU citizens’ data outside the region without their permission, either through bilateral or multilateral agreements. 4.3 The Prescriptions of Trustworthiness The prescriptions of trustworthiness belong to four dimensions: (1) AI functionality, (2) outcomes, (3) development processes, and (4) governance. Table 3 below summarizes the dimensions in the left column and the promulgated prescriptions within each dimension in the right column. Table 3 Dimensions and the Prescriptions of Trustworthiness. Dimensions of Trustworthiness Prescribed Norms/Indicators AI Functionality Accuracy & Reliability Safety & Security Adaptability Transparency Traceability Robustness Explainability Reproducibility AI Outcome Fairness Non-discrimination Equitability No-Harm Sustainability Transparency Development & Deployment Diversity & Inclusion Human-centric Proportionality Public Good Empowerment Surveillance Complementarity Governance & Regulations Accountability Ethical Fiduciary Duty Monitoring & Assessment Audits & Oversight Social Safety Net Compliance Private Regulations State Regulations Proportionality Enact Laws The prescriptions suggest societal-level tensions about trusting AI technology. This mistrust could impact the widespread adoption of AI technologies. The following network Fig. 5 shows the frequency of prescriptions by actors. 4.3.1 Prescriptions for AI’s Functionality, Outcome, and Development Processes To establish trust in AI technology, state and non-state actors have prescribed that the development of AI must incorporate diverse groups of people and cultures, and remain transparent throughout the design, development, and deployment processes. Furthermore, they prescribed that AI must function safely, accurately, and reliably, that its outcomes be transparent, fair, secure, and explainable to users, and that AI’s functionalities be traceable. State and non-state actors promulgated that AI systems must be made ‘safe and secure’ from development through implementation and use cases. The prescription of ‘transparency’ involves activities, such as disclosing how AI systems are constructed, how they operate, and what, how, and why decisions are made by the AI system and the individuals involved in these processes. 4.3.2 Prescribed Regulatory Approaches to Trustworthiness The actors prescribed to ensure ‘accountability’ in the event of adverse effects. This applies to both civil and criminal liabilities. State and non-state actors prescribed creation of a ‘monitoring and assessment’ and ‘audit and oversight’ mechanisms throughout the entire lifecycle of AI systems. Evidence indicates that a consensus exists among most states for implementing such a ‘monitoring and evaluation’ system. For instance, member states of UNESCO have agreed in principle to establish such a mechanism. 4.3.3 Prohibitions to Enhance Trustworthiness ‘No-harm’ and ‘non-discrimination’ are two prohibitive prescriptions promulgated within trustworthiness. These prescriptions prohibit the outcome of any AI system from being ‘harmful’ to its developers, deployers, and users directly or indirectly, intentionally or unintentionally. Some of the prohibitive statements specified harms in a broader context, including physical and psychological harms. For example, companies that produce AI systems are prohibited from selling them to actors with a history of human rights violations or are deemed likely to violate them. The ‘non-discrimination’ prescription prohibits AI developers from deploying AI systems that might result in discriminatory outcomes, particularly in judicial systems, law enforcement agencies, or in contexts in which AI systems filter based on religion, race, ethnicity, locations, opinion, or due to membership of a group. 4.4 The Prescriptions of Anthropocentrism The prescriptions of anthropocentrism belong to three dimensions: (1) the ‘man behind the machine,’ (2) ‘status,’ and (3) the ‘development and deployment’ of AI. The prescriptions within each dimension delineate roles and status in human-machine interactions and their primacy in relation to one another. Table − 4 below summarizes the dimensions in the left column and the prescriptions within each dimension in the right column. Table 4 Dimensions and the Prescriptions of Anthropocentrism. Dimensions of Anthropocentrism Prescribed Norms/Indicators Man Behind Machine Accountability Human-Control Statusing Human Dignification Human Autonomy Development & Deployment Beneficial Safety & Security The prescriptions suggest that ‘human control’ is likely to be a contentious governance feature, with differentiated application in civilian and military uses. They also suggest a strong predominance of material rather than ethical anthropocentrism discussed earlier. The following network Fig. 6 highlights the frequency of prescriptions and actors promulgating them. 4.4.1 Prescriptions of Human Agency ‘Human autonomy’ and ‘human dignification’ are two main prescriptions. State and non-state actors have prescribed that humans should be valued above machines and that machines must retain their status as tools, obeying commands from humans rather than intelligent entities that could be granted rights. Machines are prescribed to be subordinate to human commands and to adhere to the ethical and normative values of human society, thereby upholding human sovereignty and autonomy in human-machine interactions. Actors who mainly prescribed these are the U.N., EU, ITU, IEEE, and UNESCO. 4.4.2 Prescriptions of Man Behind the Machine States and non-state actors have prescribed that, in developing AI systems, ‘human control’ must be established and that AI systems must serve to ‘benefit humans.’ This prescription is promulgated in response to the potential for AI systems to be given full autonomy, allowing them to operate independently and adapt to various environments. It is also prescribed that, in the event of an adverse outcome, ‘accountability’ must rest with humans, not the machine. Actors like the U.N., EU, G-20, IGF, IEEE, Internet Society, ITU, ITI, OECD, and UNESCO are non-state actors that promulgated the prescriptions. States such as France, Germany, the Netherlands, and the U.S. have also promulgated the prescriptions. 4.4.3 Prohibitions of Anthropocentrism The prohibitive prescriptions address actions and activities related to violations of human sovereignty and dignity, ensuring that humans remain in control of AI systems and that no harmful AI systems are developed. The prescriptions prohibiting these actions and activities can be interpreted as obligatory, as both obligations and prohibitions are inter-definable in this case. 4.5 The Prescriptions of Enviro-centrism The prescriptions of Enviro-centrism belong to one dimension: (1) development and deployment. The prescriptions include emission reduction, energy optimization, greening the value chain, and proportionality. Table 5 below presents the dimension in the left column and the promulgated prescriptions in the right column. Table 5 The Dimension and the Prescriptions of Enviro-centrism. Dimensions of Enviro-centrism Prescribed Norms/Indicators Development & Deployment Emission Reduction Energy Optimization Greening Value Chain Proportionality Sustainability Among the five constitutive norms, Enviro-centrism is the least prescribed. This may be due to the impacts of AI on the environment, and the relationship between the two is still largely unknown. The following network Fig. 7 highlights the frequency of prescriptions and actors promulgating them. All prescriptions are obligatory in their effects except for one, which is a prohibitive promulgation. The EU Commission, France, Hungary, Switzerland, and the ITU promulgate these prescriptions. They prescribe for the development and deployment of AI systems to reduce emissions, enhance the value chain, and, most importantly, align with environmental costs and benefits. Similarly, they promulgate that AI must be used to enhance the value chain, reduce emissions, and mitigate environmental degradation, thereby establishing a relationship where technologies serve the environment. The only prohibitive promulgation comes from Switzerland, which prescribes that the energy and materials consumption by AI must not increase in line with the growing use of these technologies. 5. Conclusion This study demonstrates that, even though there is no formal global agreement on AI regulation as yet, AI governance is already happening through the creation of norms. This study identifies these norms by examining publicly available AI governance discourses. The findings suggest five main types of emergent constitutive AI governance norms, each containing a set of prescriptions regulating AI development and deployment. It is unknown when and how a global AI governance architecture will formally take shape. But as theory suggests, these five emergent [constitutive] norms, acting as the rules of the rulemaking, are likely to guide any future global, transnational, and local AI governance. The prescriptions of each of the main norm types would determine how AI is developed and used. These norms are still in the early stages of their emergence. As this study suggests, these are still being contested nationally and internationally, involving various state and non-state actors for local interpretation and adaptation. Aside from states, these actors include norm entrepreneurs such as international organizations, civil society, human rights organizations, networks of NGOs, special interest groups such as technology corporations, and intergovernmental organizations, who have interests and stakes in the ways AI governance takes shape. This contestation among the various AI stakeholders and their participation in different venues indicates the norm emergence mechanism of what Finnemore and Sikkink (1998) suggest, the strategic construction of norms by actors to solve the AI governance problem that affects them all (Ullman-Margalit 1977; Coleman 1990 ; Finnemore and Sikkink 1998). Future studies may examine AI governance to explain why AI governance norms are emerging in their current form and the influence these actors may have on one another in setting the AI governance agenda globally. In addition, norm emergence theories suggest a co-constitutive mechanism, in which the emergence of a norm is shaped by actors’ interests, opportunity structures, and fitness of new norms with the existing normative foundation (O’Neill et al. 2004 ; VanDeveer 2004 ; 2011 ; Klotz 2002 ; Nadelmann 1990 ; Sandholtz 2008 ; Risse-Kappen 1994 ; Risse 2000 ). As this study suggests, numerous non-state actors are involved in setting the global AI governance agenda and attempting to influence the direction, nature, and shape of global AI governance architecture. These actors carry significant meaning about their preferences and power dynamics in global politics. Studies suggest that non-technical actors and scientific experts are both involved in the design, implementation, and adoption of science and technologies in society, as well as in technological decision-making at both global and local levels in different issues such as environmental governance, biotechnology, and nanotechnology governance (Bucchi and Neresini 2008; VanDeveer 2004 ; 2011 ). It is not clear how and to what extent these actors exert influence on one another in the case of AI governance. Future studies may examine the dynamics of actors’ preferences and the extent and mechanisms of their influence on one another, as well as the material and ideational opportunity structures that facilitate such actors’ participation in norm-generating platforms in shaping the global AI governance regime. Furthermore, norms do not only emerge, they also disappear and die (Panke and Petersohn 2016; 2012). It will be interesting to observe whether these constitutive norms continue to gain momentum or might as well disappear with the evolution of AI technology and the governance landscape. Declarations Author Contribution Agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Acknowledgement Acknowledgment I express my gratitude to Professor Stacy D. VanDeveer, Professor J. Samuel Barkin, and Professor Karen Ross at the University of Massachusetts Boston for their important contribution in conducting this research and feedback on an earlier version of this manuscript. I also thank Professor Joshua C. Gellers of the University of North Florida for his important contribution to the development of typologies in this study. Data Availability All data supporting the findings of this study are available within the paper and its Supplementary Information. Microsatellite primer sequences are provided in Supplementary Table 2, along with original reference describing the microsatellites used in this study. References Axelrod R (1986) An Evolutionary Approach to Norms. Am Polit Sci Rev 80(4):1095–1111. https://doi.org/10.2307/1960858 Babbie ER (2016) The Practice of Social Research . Fourteenth edition. Cengage Learning Bailey, Kenneth D (1994) Typologies and Taxonomies An Introduction to Classification Techniques. Sage Barkin JS, Sjoberg L (eds) (2017) Interpretive Quantification: Methodological Explorations for Critical and Constructivist IR. University of Michigan Press Bicchieri C, Erte Xiao, and Ryan Muldoon (2011) Trustworthiness Is a Social Norm, but Trusting Is Not. Politics Philos Econ 10(2):170–187. https://doi.org/10.1177/1470594X10387260 Biersteker T (2015) Global Governance. On Governance: What It Is, What It Means and Its Policy Uses. McGill-Queen’s University. http://www.jstor.org/stable/j.ctt1jktrbx.12 Bode I (2023) Practice-Based and Public-Deliberative Normativity: Retaining Human Control over the Use of Force. Eur J Int Relat 29(4):990–1016. https://doi.org/10.1177/13540661231163392 Bucchi M, and Federico Neresini (2008) Science and Public Participation. The Handbook of Science and Technology Studies. The MIT Press Carr EH (1946) The Twenty Years’ Crisis, 1919–1939: Reissued with a New Preface from Michael Cox . Edited by Michael Cox. Palgrave Macmillan Coleman JS (1990) Foundations of Social Theory. Harvard University Press Crist E, and Helen Kopnina (2014) Unsettling Anthropocentrism. Dialect Anthropol 38(4):387–396. https://doi.org/10.1007/s10624-014-9362-1 Doty D, Harold, Glick WH (1994) Typologies as a Unique Form of Theory Building: Toward Improved Understanding and Modeling. Acad Manage Rev 19(2):230. https://doi.org/10.2307/258704 Fairciough N (2013) Language and Power, 2nd edn. Taylor and Francis Finnemore M, and Kathryn Sikkink (1998) International Norm Dynamics and Political Change. Int Org 52(4):887–917. https://doi.org/10.1162/002081898550789 Gasser U, Virgilio AF, Almeida (2017) A Layered Model for AI Governance. IEEE Internet Comput 21(6):58–62. https://doi.org/10.1109/MIC.2017.4180835 Gellers J (2021) Rights for Robots: Artificial Intelligence, Animal and Environmental Law. Taylor & Francis Gerhardt U (1994) The Use of Weberian Ideal-Type Methodology in Qualitative Data Interpretation: An Outline for Ideal-Type Analysis. Bull Sociol Methodology/Bulletin de Méthodologie Sociologique 45(1):74–126. https://doi.org/10.1177/075910639404500105 Gherhes C, Yu Z, Vorley T, and Lan Xue (2023) Technological Trajectories as an Outcome of the Structure-Agency Interplay at the National Level: Insights from Emerging Varieties of AI. World Dev 168(August):106252. https://doi.org/10.1016/j.worlddev.2023.106252 Goldstein J, Keohane RO (eds) (1993) Ideas and Foreign Policy: Beliefs, Institutions, and Political Change . Cornell Studies in Political Economy. Cornell University Press Gupta A, Biermann F, Van Driel E et al (2024) Towards a Non-Use Regime on Solar Geoengineering: Lessons from International Law and Governance. Transnatl Environ Law 13(2):368–399. https://doi.org/10.1017/S2047102524000050 Habermas Jürgen, McCarthy TA (2005) and Jürgen Habermas. Lifeworld and System: A Critique of Functionalist Reason . 1. digital-Print ed. The Theory of Communicative Action / Jürgen Habermas. Transl. by Thomas MacCarthy, Vol. 2. Beacon Hardin R (2002) Trust and Trustworthiness. The Russell Sage Foundation Series on Trust, vol 4. Russell Sage Foundation Hart HLA (1994) The Concept of Law, 2nd edn. Clarendon Law Series. Clarendon Jepperson RL, Wendt A, Peter JK (1996) Norms, Identity and Culture in National Security. The Culture of National Security: Norms and Identity in World Politics. Columbia University, New York Jones K (2012) Trustworthiness Ethics 123(1):61–85. https://doi.org/10.1086/667838 Katzenstein PJ (ed) (1996) The Culture of National Security: Norms and Identity in World Politics. New Directions in World Politics. Columbia Univ. Keohane RO (1984) After Hegemony: Cooperation and Discord in the World Political Economy . 1st Princeton classic ed. A Princeton Classic Edition. Princeton University Press Keohane RO (1989) International Institutions and State Power: Essays in International Relations Theory, 1st edn. Westview. https://doi.org/10.4324/9780429032967 Klotz A (2002) Transnational Activism and Global Transformations: The Anti-Apartheid and Abolitionist Experiences. Eur J Int Relat 8(1):49–76. https://doi.org/10.1177/1354066102008001002 Kopnina H, Washington H, Taylor B, John JP (2018) Anthropocentrism: More than Just a Misunderstood Problem. J Agric Environ Ethics 31(1):109–127. https://doi.org/10.1007/s10806-018-9711-1 Kowert P, and Jeffrey Legro (1996) Norms, Identity, and Their Limits: A Theoretical Reprise. The Culture of National Security: Norms and Identity in World Politics. Columbia Univ. Krathwohl DR (2002) A Revision of Bloom’s Taxonomy: An Overview. Theory Into Pract 41(4):212–218. https://doi.org/10.1207/s15430421tip4104_2 Kratochwil FV (1989) Rules, Norms, and Decisions: On the Conditions of Practical and Legal Reasoning in International Relations and Domestic Affairs. Reprint. Cambridge Studies in International Relations, vol 2. Cambridge Univ. Press Lawless S, Song AM, Cohen PJ, and Tiffany H. Morrison (2020) Rights, Equity and Justice: A Diagnostic for Social Meta-Norm Diffusion in Environmental Governance. Earth Syst Gov 6(December):100052. https://doi.org/10.1016/j.esg.2020.100052 Legro J (1995) Cooperation under Fire: Anglo-German Restraint during World War II. Cornell University Press Mandara J (2003) The Typological Approach in Child and Family Psychology: A Review of Theory, Methods, and Research. Clin Child Fam Psychol Rev 6(2):129–146. https://doi.org/10.1023/A:1023734627624 March JG, Olsen JP (1998) The Institutional Dynamics of International Political Orders. Int Org 52(4):943–969. https://doi.org/10.1162/002081898550699 Mayer-Schönberger V, and Kenneth Cukier (2013) Big Data: A Revolution That Will Transform How We Live, Work, and Think. Houghton Mifflin Harcourt Milliken J (1999) The Study of Discourse in International Relations: A Critique of Research and Methods. Eur J Int Relat 5(2):225–254. https://doi.org/10.1177/1354066199005002003 Nadelmann EA (1990) Global Prohibition Regimes: The Evolution of Norms in International Society. Int Org 44(4):479–526. https://doi.org/10.1017/S0020818300035384 Nelson J, and Tim Gorichanaz (2019) Trust as an Ethical Value in Emerging Technology Governance: The Case of Drone Regulation. Technol Soc 59:101131. https://doi.org/10.1016/j.techsoc.2019.04.007 OECD (2021) Database of National AI Policies. https://oecd.ai O’Neill K, Balsiger Jörg, VanDeveer SD (2004) ACTORS, NORMS, AND IMPACT: Recent International Cooperation Theory and the Influence of the Agent-Structure Debate. Annu Rev Polit Sci 7(1):149–175. https://doi.org/10.1146/annurev.polisci.7.090803.161821 Opp K-D (1982) The Evolutionary Emergence of Norms. Br J Soc Psychol 21(2):139–149. https://doi.org/10.1111/j.2044-8309.1982.tb00522.x Palladino N (2023) A ‘Biased’ Emerging Governance Regime for Artificial Intelligence? How AI Ethics Get Skewed Moving from Principles to Practices. Telecomm Policy 47(5):102479. https://doi.org/10.1016/j.telpol.2022.102479 Panke D, and Ulrich Petersohn (2012) Why International Norms Disappear Sometimes. Eur J Int Relat 18(4):719–742. https://doi.org/10.1177/1354066111407690 Panke D, and Ulrich Petersohn (2016) Norm Challenges and Norm Death: The Inexplicable? Cooperation Confl 51(1):3–19. https://doi.org/10.1177/0010836715597948 Price R (1998) Reversing the Gun Sights: Transnational Civil Society Targets Land Mines. Int Org 52(3):613–644. https://doi.org/10.1162/002081898550671 Radu R (2021) Steering the Governance of Artificial Intelligence: National Strategies in Perspective. Policy Soc 40(2):178–193. https://doi.org/10.1080/14494035.2021.1929728 Risse T (2000) Let’s Argue!’: Communicative Action in World Politics. Int Org 54(1):1–39. https://doi.org/10.1162/002081800551109 Risse T (2002) Nationalism and Collective Identities. Europe versus the Nation-State? Developments West Eur Politics no. 2:77–93 Risse-Kappen T (1994) Ideas Do Not Float Freely: Transnational Coalitions, Domestic Structures, and the End of the Cold War. Int Org 48(2):185–214. https://doi.org/10.1017/S0020818300028162 Rosert E (2019) Norm Emergence as Agenda Diffusion: Failure and Success in the Regulation of Cluster Munitions. Eur J Int Relat 25(4):1103–1131. https://doi.org/10.1177/1354066119842644 Ruggie JG (1998) What Makes the World Hang Together? Neo-Utilitarianism and the Social Constructivist Challenge. Int Org 52(4):855–885. https://doi.org/10.1162/002081898550770 Saheb T, and Tayebeh Saheb (2023) Topical Review of Artificial Intelligence National Policies: A Mixed Method Analysis. Technol Soc 74:102316. https://doi.org/10.1016/j.techsoc.2023.102316 Sandholtz W (2008) Dynamics of International Norm Change: Rules against Wartime Plunder. Eur J Int Relat 14(1):101–131. https://doi.org/10.1177/1354066107087766 Stapley E, Sally O’Keeffe, and Nick Midgley (2022) Developing Typologies in Qualitative Research: The Use of Ideal-Type Analysis. Int J Qualitative Methods 21(April):16094069221100633. https://doi.org/10.1177/16094069221100633 Stix C (2021) The Ghost of AI Governance Past, Present and Future: AI Governance in the European Union. Version 1. https://doi.org/10.48550/ARXIV.2107.14099 . Preprint, arXiv Tan SY, and Araz Taeihagh (2021a) Adaptive Governance of Autonomous Vehicles: Accelerating the Adoption of Disruptive Technologies in Singapore. Government Inform Q 38(2):101546. https://doi.org/10.1016/j.giq.2020.101546 Tan SY, and Araz Taeihagh (2021b) Governing the Adoption of Robotics and Autonomous Systems in Long-Term Care in Singapore. Policy Soc 40(2):211–231. https://doi.org/10.1080/14494035.2020.1782627 The AI Act of the European Parliament and of the, Council (2024) /1689 Pub. L. No. (2024). https://eur-lex.europa.eu/eli/reg/2024/1689 Tieku TK (2019) Ruling from the Shadows: The Nature and Functions of Informal International Rules in World Politics. Int Stud Rev 21(2):225–243. https://doi.org/10.1093/isr/viz012 Turing AM (2009) Computing Machinery and Intelligence. In Parsing the Turing Test , edited by Robert Epstein, Gary Roberts, and Grace Beber. Springer Netherlands. https://doi.org/10.1007/978-1-4020-6710-5_3 Ullmann-Margalit E (1977) The Emergence of Norms . 1. published in paperback. Clarendon Library of Logic and Philosophy. Clarendon Press Utech, Myron R (1963) The Social Action Approach of Max Weber. Thesis, Bowling Green State University VanDeveer S (2004) Ordering Environments: Regions in European International Environmental Cooperation. In Earthly Politics: Local and Global in Environmental Governance VanDeveer S (2011) Networked Baltic Environmental Cooperation. J Baltic Stud 42(1):37–55. https://doi.org/10.1080/01629778.2011.538516 Wiener A (2009) Enacting Meaning-in-Use: Qualitative Research on Norms and International Relations. Rev Int Membr 35(1):175–193. https://doi.org/10.1017/S0260210509008377 Wiener A (2014) A Theory of Contestation. SpringerBriefs in Political Science. Springer Wodak R (ed) (2014) Methods of Critical Discourse Analysis . 2. Aufl., Repr. Introducing Qualitative Methods. SAGE von Wright GH (1963) Norm and Action: A Logical Enquiry . Reprinted. International Library of Philosophy and Scientific Method. Routledge & Kegan Paul [u.a.] Zürn M (2018) A Theory of Global Governance: Authority, Legitimacy, and Contestation . First edition. Oxford University Press Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9383807","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"open-forum","associatedPublications":[],"authors":[{"id":634553116,"identity":"0b795b1d-d613-4b14-81a1-d5b0c8dffa7c","order_by":0,"name":"Muhammed Hasan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+0lEQVRIiWNgGAWjYFACHjDJ2CABpiV42NgbGJjBIkRqsZHj4zlAmpY0YzmJBPxadGfkHvx0M8dGtkG6+fGHnzsOJ7ZJvjHdXMBgI7vhAHYtZjfykqVzt6UZN8gcM5PsPQPUIp1jdnsG0DrcWnIMgFoOJzZIJJgx8LZBtfAwHE7Eo8X4d+62/0At6Z8//gVpkTwD0vIfnxYzoC0HgFqA1vG2pRmzSfCAtBzAreXMGzPr3G3Jxm0SOWXSsm02cmw8aWW3ZxgkG8/EpeV4jvHt3G12sv0S6Zs/vm2T4JFvP7ztdkGFnWwfDi1wwIbKNSCgfBSMglEwCkYBXgAAofphdanqOi0AAAAASUVORK5CYII=","orcid":"","institution":"University of Massachusetts Boston","correspondingAuthor":true,"prefix":"","firstName":"Muhammed","middleName":"","lastName":"Hasan","suffix":""}],"badges":[],"createdAt":"2026-04-11 02:08:40","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9383807/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9383807/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108696484,"identity":"d7f73a9d-60e5-4b34-8f5d-c71a0c227383","added_by":"auto","created_at":"2026-05-07 11:57:04","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":33447,"visible":true,"origin":"","legend":"\u003cp\u003eThe Frequency Distribution of Five Constitutive Norms \u0026amp; Actors Promulgated Them. The ‘blue’ represents the number of prescriptions identified in documents from State actors, while ‘green’ represents non-state actors, ‘pink’ represents arguments/debates, and ‘purple’ represents the documents produced by both state and non-state actors. Of the total 1028 prescriptions, 347 are categorized under the constitutive norm type ‘AI-fication’ (221 in states’, 55 in argumentative documents, 69 in non-state actors, and 2 identified in documents produced by state and non-state actors); 313 are categorized under the constitutive norm type ‘Trustworthiness’ (154 in states, 50 argumentative documents, 107 in non-state actors, and 2 identified in documents produced by state and non-state actors); 271 are categorized under the constitutive norm type ‘Datafication’ (189 in states, 22 documents, 59 in non-state actors, and 1 is found in documents produced by state and non-state actors); 89 are categorized under the constitutive norm type ‘Anthropocentrism’ (23 in states, 14 in argumentative documents, 51 in non-state actors, and 1 is found in documents produced by state and non-state actors), finally, 8 are categorized under the constitutive norm type ‘Enviro-centrism’ (4 in states, 1 in argumentative documents, and 3 promulgated by non-state actors).\u003c/p\u003e","description":"","filename":"FigureOne.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/1dedaff0d6d25ccc8e2172fd.png"},{"id":108696486,"identity":"a3ba110b-c952-4c22-a621-ffbea089c782","added_by":"auto","created_at":"2026-05-07 11:57:04","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":31720,"visible":true,"origin":"","legend":"\u003cp\u003eThe Frequency Distribution of Constitutive Norms \u0026amp; Their Effects. The color ‘green’ indicates the number of prescriptions with obligatory effects, ‘blue’ with permissive effects, ‘pink’ with command effects, and ‘purple’ with prohibitive effects. Within AI-fication, 314 prescriptions are to the effects of obligation, 10 command, 9 permissions, and 14 are prohibition. Within Trustworthiness, 265 prescriptions are to the effects of obligation, 16 are permission, 2 are command, and 30 are prohibition. Within Datafication, 202 prescriptions are to the effects of obligation, 54 are permissions, 1 is a command, and 14 are prohibition. Within Anthropocentrism, 72 prescriptions are to the effects of obligation, 1 is permission, and 16 are prohibition. Within Enviro-centrism, 7 prescriptions are to the effects of obligation, and one prescription is prohibition.\u003c/p\u003e","description":"","filename":"FigureTwo.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/ff8f4dcb9590c352b301c249.png"},{"id":108696502,"identity":"db43e89d-fd77-4e09-9947-8ffe3becffee","added_by":"auto","created_at":"2026-05-07 11:57:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":8097558,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 3 presents the prescriptions of AI-fication. It shows the in-degree frequency of each prescription and the linkages of actors promulgating them. Actors are represented with an in-degree frequency of zero and are red color-coded. The prescriptions are represented with in-degree frequency in ascending order, starting from 1, indicating the number of actors promulgating the prescriptions. A lower frequency means a smaller number of actors and is red color-coded. A higher frequency indicates that more actors promulgated the prescriptions, which are blue color-coded.\u003c/p\u003e","description":"","filename":"FigureThree.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/ed08abbc8384204b4521dd56.png"},{"id":108696503,"identity":"795c5bd1-b497-44a8-ae06-0233455920fd","added_by":"auto","created_at":"2026-05-07 11:57:15","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":7773090,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 4 presents the prescriptions of Datafication. It shows the in-degree frequency of each prescription and the linkages of actors promulgating them. Actors are represented with an in-degree frequency of zero and are red color-coded. The prescriptions are represented with in-degree frequency in ascending order, starting from 1, indicating the number of actors promulgating the prescriptions. A lower frequency means a smaller number of actors and is red color-coded. A higher frequency indicates that more actors promulgated the prescriptions, which are blue color-coded.\u003c/p\u003e","description":"","filename":"FigureFour.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/87b91cb2cd03383f5bddd4bd.png"},{"id":108696508,"identity":"f146c407-7523-4b23-844b-2ddee5740ef7","added_by":"auto","created_at":"2026-05-07 11:57:15","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":11648758,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 5 presents the prescriptions of Trustworthiness. It shows the in-degree frequency of each prescription and the linkages of actors promulgating them. Actors are represented with an in-degree frequency of zero and are red color-coded. The prescriptions are represented with in-degree frequency in ascending order, starting from 1, indicating the number of actors promulgating the prescriptions. A lower frequency means a smaller number of actors and is red color-coded. A higher frequency indicates that more actors promulgated the prescriptions, which are blue color-coded.\u003c/p\u003e","description":"","filename":"FigureFive.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/5df7620358abe6320c1d4d77.png"},{"id":108696491,"identity":"99941df3-862f-45ab-be5b-b1981b1c594a","added_by":"auto","created_at":"2026-05-07 11:57:11","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2068805,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 6 presents the prescriptions of Anthropocentrism. It shows the in-degree frequency of each prescription and the linkages of actors promulgating them. Actors are represented with an in-degree frequency of zero and are red color-coded. The prescriptions are represented with in-degree frequency in ascending order, starting from 1, indicating the number of actors promulgating them. A lower frequency means a smaller number of actors and is red color-coded. A higher frequency indicates that more actors promulgated the prescriptions, which are blue color-coded.\u003c/p\u003e","description":"","filename":"FigureSix.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/339cd851f7b9fc1b7882623c.png"},{"id":108696485,"identity":"3c2228dc-4739-423c-965c-717b7719aa5f","added_by":"auto","created_at":"2026-05-07 11:57:04","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":494703,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 7 presents the prescriptions of Enviro-centrism. It shows the in-degree frequency of each prescription and the linkages of actors promulgating them. Actors are represented with an in-degree frequency of zero and are red color-coded. The prescriptions are represented with in-degree frequency in ascending order, starting from 1, indicating the number of actors promulgating them. A lower frequency means a smaller number of actors and is red color-coded. A higher frequency indicates that more actors promulgated the prescriptions, which are blue color-coded.\u003c/p\u003e","description":"","filename":"FigureSeven.png","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/2692bd4125aaab4207c737c9.png"},{"id":108696470,"identity":"110d5024-90ce-48a9-9533-de8667384c4b","added_by":"auto","created_at":"2026-05-07 11:56:58","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":415685,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9383807/v1/ceb9606c-259e-4bde-a5f3-86570f5a8283.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Governing Artificial Intelligence (AI) Through Creating Governance Norms: Identifying the Emerging Global AI Governance Norms.","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003ePrevious AI studies into global AI governance are mostly normatively grounded \u0026ndash; that is, they analyze the risks and benefits of AI technology and propose governance solutions based on their expectations of what should be done, which sometimes extrapolate governance mechanisms that have evolved in similar technologies. Empirical studies, such as Bode (Bode \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), however, focus on specific cases of AI in a particular location that examine the emerging themes in AI governance and thus lack a comprehensive picture of society\u0026rsquo;s collective expectations about AI governance. My argument is that global AI governance is already happening. According to the Organization for Economic Cooperation and Development\u0026rsquo;s (OECD) AI Observatory (OECD \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), around sixty countries have released over nine hundred strategic documents. Besides, non-state actors have also published AI governance guidelines. These discursive practices indicate emerging patterns of behavior and stakeholders\u0026rsquo; preferences regarding how this technology may or may not be developed and used by problematizing various issues and setting a governance agenda to address them (Rosert \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn this study, I examine these AI governance discourses to identify emerging global AI governance norms. Norms are understood as socially agreed-upon expectations consisting of practices and rules defining appropriate and inappropriate behavior that shape and regulate behavior (Ullmann-Margalit \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e1977\u003c/span\u003e; Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e; Finnemore and Sikkink 1998; Jepperson et al. \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1996\u003c/span\u003e; March and Olsen \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). Norms influence behavior through two pathways. One way is by regulating human actions. This type of norm is known as a regulative norm, which guides an already-existing activity and is intended to have causal effects on human behavior (Ruggie \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). Regulative norms are sometimes referred to as prescriptive norms (Katzenstein \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e1996\u003c/span\u003e), as they guide human behavior by providing dos and don\u0026rsquo;ts. Henceforth, I use the term \u0026lsquo;prescriptions\u0026rsquo; to indicate these norms. Wright (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) posits that prescription operates in three ways to determine social actions: it permits, obligates, or prohibits certain social practices over others. He suggests that these normative expectations are expressed in language, conveying that \u0026ldquo;something ought to, may, must not be or must be done\u0026rdquo; (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e, 100; Opp \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1982\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe second pathway through which norms affect human behavior is called constitutive norms, which define the practices that constitute a particular class of consciously organized social activity. Constitutive norms specify what qualifies as that activity, while regulative norms dictate how a specific activity can be performed. Constitutive norms form the institutional foundation of all social life, without which no consciously organized realm of human activity can be envisioned (Ruggie \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). They create new actors, interests, or categories of action (Finnemore and Sikkink 1998). Constitutive norms can be interchangeably referred to as principles (Hart \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). Lawless and her colleagues (Lawless et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) argue that constitutive norms, or principles, typically represent fundamental global principles that emerge as aspirational goals through international agreements and guidelines, providing normative guidance for best practices. Henceforth, I use \u0026lsquo;constitutive norms\u0026rsquo; to indicate these aspirational goals that contain a set of prescriptions.\u003c/p\u003e \u003cp\u003eThus, this study accomplishes two tasks. First, it identifies prescriptions by examining AI governance related documents published by state and non-state actors. I accomplish this task by employing Wright\u0026rsquo;s (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) norm analysis framework. Second, I construct nonabstract generalizable conceptual typologies of these identified prescriptions into five constitutive norms. These are: (1) AI-fication, (2) Datafication, (3) Trustworthiness, (4) Anthropocentrism, and (5) Enviro-centrism. To construct the conceptual typology, I employed the \u0026lsquo;ideal type analysis method\u0026rsquo; (Gerhardt \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). Each of the constitutive norm holds and describes the set of prescriptions that permit, obligate, or prohibit specific AI development and deployment practices.\u003c/p\u003e \u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Significance of the study\u003c/h2\u003e \u003cp\u003eFrom a functional perspective, norms regulate human behavior, having prescriptive and proscriptive effects on our social actions (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e). This is particularly true for the international political system, which is anarchic and structurally non-hierarchical. Global politics is marked as much by power maximization as it is by morality (Carr \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1946\u003c/span\u003e, 89). The prescriptive and constitutive norms presented here are at the early stages of their emergence (Finnemore and Sikkink 1998). These are and will continue to be debated and negotiated in various contexts involving state and non-state actors. Therefore, I do not claim that these norms are the benchmark of a global AI regulation since they are yet to be formalized into codified regulations. However, I argue that, in the absence of an overarching central government, these norms would influence actors in articulating their preferences about a global AI governance architecture (Goldstein and Keohane \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Keohane \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; Katzenstein \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e1996\u003c/span\u003e; Legro \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e1995\u003c/span\u003e; Kratochwil \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1989\u003c/span\u003e). In addition, these norms will provide justifications for political actions by acting as guiding principles, drawing clear distinctions between appropriate and inappropriate behaviors, and setting the standard of behavior for future AI governance actions (Risse \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; March and Olsen \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e1998\u003c/span\u003e; Goldstein and Keohane \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Keohane \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1984\u003c/span\u003e). In doing so, these norms will provide the background and lay the foundations for formalized and codified AI governance mechanisms (Tieku \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.3 Contribution of the study\u003c/h2\u003e \u003cp\u003eThis study makes several contributions. First, there is no global AI governance mechanism that exists today except the regional-level EU AI Act (The AI Act of the European Parliament and of the Council \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Therefore, our knowledge about the dynamics of global AI governance architecture is limited. This study fills this gap by systematically examining individual AI development and deployment policies, shedding light on the normative foundations of the nature and direction of such a potential governance regime. Since norms function as informal rules (Tieku \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), the five constitutive norms and the prescriptions within them will play the roles of secondary rules that would shape the formulation of primary types of laws (Hart \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1994\u003c/span\u003e) in the form of treaties, conventions, or institutional arrangements like those that exist in telecommunication and internet governance. Even though there are several pathways, such as the one-country-one-vote, weighted voting to advantage major powers, weighted voting to advantage vulnerable countries, and double-weighted majority voting (Gupta et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), I do not claim how and when these norms may be institutionalized. However, this study suggests that a consensus is emerging among state and non-state actors regarding the norms presented here. Therefore, I argue that these emergent norms may reach a tipping point as an increasing number of actors begin to act in accordance with them, thus triggering a norm cascade, which may occur through various socialization processes (Finnemore and Sikkink 1998; VanDeveer \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Nadelmann \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e1990\u003c/span\u003e; Price \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e1998\u003c/span\u003e), hence becoming institutionalized through their adoption in domestic and international formal legal codes.\u003c/p\u003e \u003cp\u003eSecond, Prior AI governance studies primarily focus on exploring frequently discussed themes within national AI policies, understanding how ethics are integrated into AI governance, and examining governance methods in specific countries (Radu \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Gherhes et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Saheb and Saheb \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Nelson and Gorichanaz 2019; Stix \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Palladino \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Tan and Taeihagh 2021b; 2021a). The thematic analyses do not effectively capture society\u0026rsquo;s collective expectations about what actions are or are not permissible or prohibited. Contrarily, due to prescriptive and constitutive effects, norms dictate what individuals can and cannot do while also constituting new actors and identities through socialization processes. Once a norm emerges, it undergoes a life cycle that includes emergence, cascade, and, ultimately, institutionalization through its incorporation into formal national and international laws (Finnemore and Sikkink 1998).\u003c/p\u003e \u003cp\u003eThird, though studies into norms date back at least several centuries across multiple disciplines, the question of \u0026lsquo;how we know a norm when we see one\u0026rsquo; (Finnemore and Sikkink 1998) remains puzzling. Constructivists study the roles, evolution, and diffusion of norms in global politics using interpretive methodologies, such as genealogy (Kowert and Legro 1996). In the process, they ignore the issue of emergence. On the other hand, positivists study when, how, and what types of norms may emerge by utilizing methodological individualism, such as game-theoretic models (Axelrod \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1986\u003c/span\u003e; Ullmann-Margalit \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e1977\u003c/span\u003e; Coleman \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e1990\u003c/span\u003e). However, they too forget that social change happens through social interaction. So does norm emergence. This study presents an interpretive quantification method for examining and identifying norm emergence through a systematic analysis of language because languages are the social reality, practices, and social actions embedded in a particular social context (Fairciough \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Habermas et al. 2005; Wodak \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Therefore, language, being the social practices themselves, is what people say, and what they do in practice, carrying the meaning-in-use of the norms (Milliken \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e1999\u003c/span\u003e; Wiener \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn what follows, in Section Two, I discuss the methodology, define the terms, present the types and sources of data, and finally discuss the methods of identifying the prescriptions and conceptual typology development. In the Third Section, I introduce and define the five constitutive norms. In the Fourth Section, I elaborate on each of the constitutive norms and prescriptions within them, highlighting prescriptive and proscriptive effects as well as shedding light on agreements and contentions among actors. The article concludes with a discussion of future directions for research on AI governance.\u003c/p\u003e \u003c/div\u003e"},{"header":"2. Methodology","content":"\u003cp\u003eI employed an interpretive quantification method to identify the prescriptions. This design effectively produces intersubjective and relational knowledge (Barkin and Sjoberg \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). It is interpretive in that prescriptions are promulgated exclusively through language (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e), and thus, a systemic interpretation of language uncovers the meaning of prescriptions and their effects. It is quantitative because I assessed the resonance of the identified prescriptions by analyzing their frequency across various types of actors.\u003c/p\u003e \u003cp\u003eI define AI as the art, science, and engineering of creating machines that act like humans and perform tasks better than humans (Turing \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Though AI could be deployed in different domains, the fundamental building blocks and their functionalities are the same - automation. This study considers AI as a technology for automation, rather than specific use cases. I define global governance as an \u0026ldquo;intersubjectively recognized purposive order consisting of a system of formal rules embodied in institutions or informal patterned regularities observed through individuals\u0026rsquo; habitual actions that define, constrain, and shape actor expectations and conduct in a given issue domain at the global level\u0026rdquo; (Biersteker \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) I define global as being an intercultural interaction between agents of different roots (Wiener \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Understanding this way includes state and non-state actors who are interacting with one another through which patterns of behavior emerge (Z\u0026uuml;rn \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Wiener \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). I adopt Karl-Dieter Opp\u0026rsquo;s definition of emergence. Opp (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1982\u003c/span\u003e) argues that \u0026ldquo;if we explain the emergence of norms, we explain under what conditions individuals express new normative statements\u0026rdquo; (61). Finally, I adopt the norm definition offered by Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) and Opp (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1982\u003c/span\u003e), who define norm in terms of its effects that something ought to or may or must not be or must be done.\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Data\u003c/h2\u003e \u003cp\u003eI have used three types of data for analysis: 1) states\u0026rsquo; AI policy documents, 2) AI documents put forward by non-state actors, and argumentative documents. The non-state actors involve multinational tech corporations, civil society groups, academia, AI research institutes, and International Organizations (IOs). Argumentative documents include AI governance-related seminars, debates, and inter-governmental meetings that took place on international organizations\u0026rsquo; platforms participated in by either state or non-state actors. Several reasons for including diverse documents. First, prescriptions are promulgated exclusively through language (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e). Therefore, the inclusion of diverse documents is helpful for a comprehensive analysis, irrespective of the nature and characteristics of the documents themselves. Second, since norms indicate society\u0026rsquo;s collective expectations, such inclusion would suggest AI stakeholders\u0026rsquo; collective expectations and preferences. Finally, the inclusion of three types of documents captures the dynamics of global governance, which is a system of formal rules and informal patterned regularities.\u003c/p\u003e \u003cp\u003eI collected this data primarily from the OECD\u0026rsquo;s digital library. Its AI \u003cem\u003ePolicy Observatory\u003c/em\u003e archives states\u0026rsquo; AI initiatives on the \u003cem\u003eNational AI Policies and Strategies\u003c/em\u003e page; conversations, meetings, and AI-related seminars on the \u003cem\u003eAI Policy Events\u003c/em\u003e page; and it stores non-state actors\u0026rsquo; AI-related activities on the \u003cem\u003eOECD\u0026rsquo;s Stakeholders\u0026rsquo; Initiative\u003c/em\u003e page within the same digital library. The video recordings of meetings and conferences on IO\u0026rsquo;s platforms span from January 2000 to January 2022. I downloaded the recordings using \u0026ldquo;4K Downloader\u0026rdquo; and \u0026ldquo;iTubeGo\u0026rdquo; software. I have used NVIVO and \u0026ldquo;\u0026Ouml;tter.ai\u0026rdquo; software to transcribe the recordings. Additionally, I conducted an open Google search, first to locate documents listed but missing in the OECD\u0026rsquo;s archive and second to find additional AI governance related documents that may not be stored in the AI Policy Observatory archive. I used a total of 345 AI policies from 59 states for the final analysis out of 900 documents, 873 recordings collected from 15 international organizations, and 87 AI guidelines from 54 distinct non-state actors for this analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Analysis\u003c/h2\u003e \u003cp\u003eThe analysis proceeded in two stages. The first stage involved identifying prescriptions, and the second stage involved constructing typologies of the prescriptions into constitutive norms.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 Identifying Prescriptions\u003c/h2\u003e \u003cp\u003eI utilized Wright\u0026rsquo;s (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) norm analysis framework to identify the prescriptions. Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) proposes both morphic and semantic analysis of language to distinguish between various types of prescriptions. According to Wright, a prescription may be promulgated using imperative, indicative sentences, or deontic modal verbs. However, Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) contends that \u0026ldquo;neither as a morphological nor as a semantic category is the notion of the imperative mood clear and homogeneous enough to make even a provisional identification of prescriptions plausible\u0026rdquo; (99). In addition, Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) suggests that indicative sentences are typically used in the future tense, and their use as a prescription is more common in legal language. It follows trivially that every prescription can be expressed in a deontic sentence,\u0026rdquo; (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e, 100), which is promulgated exclusively through deontic modal verbs\u0026mdash;\u0026lsquo;may,\u0026rsquo; \u0026lsquo;ought,\u0026rsquo; and \u0026lsquo;must.\u0026rsquo; Thus, I first isolated the deontic sentences based on morphic criteria of deontic modal verbs using NVIVO data analysis software in which I created a \u0026lsquo;text search\u0026rsquo; query using the words, \u0026lsquo;may\u0026rsquo;, \u0026lsquo;may not\u0026rsquo;, \u0026lsquo;must\u0026rsquo;, \u0026lsquo;mustn\u0026rsquo;t\u0026rsquo;, \u0026lsquo;ought\u0026rsquo;, \u0026lsquo;ought to\u0026rsquo;, \u0026lsquo;ought not to\u0026rsquo; including their stemmed words.\u003c/p\u003e \u003cp\u003eFollowing Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e), I conducted a semantic analysis. To do that, I analyzed the identified deontic sentences on several semantic dimensions: (1) isolating sentences containing the deontic modal verb \u0026lsquo;may\u0026rsquo; to determine whether the sentence indicates permission or possibility; (2) separating prescriptions from ideal rules based on whether the deontic modal verbs are conjoined with the auxiliary verb \u0026lsquo;be\u0026rsquo;. According to Wright (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e), ideal rules are conjoined with the auxiliary verb \u0026lsquo;be\u0026rsquo; along with deontic modal verbs, such as the form of \u0026lsquo;may be\u0026rsquo;. (3) Since I am concerned with the identification of prescriptions relevant only to the actions and activities in the social domain, I separated the prescriptions belonging to the cognitive domain by using a taxonomy of cognitive verbs (Krathwohl \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2002\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe third stage involved examining whether the deontic sentences contain the six mandatory components that set apart prescriptions from other types of norms (Wright \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e1963\u003c/span\u003e). These are: (1) the character indicating permission, obligation, or prohibition; (2) the content to indicate the prescribed actions and activities; (3) the conditions of applications to determine whether a prescription is categorical or hypothetical; (4) the authority meaning the prescription givers; (5) the subject meaning the prescription takers; (6) and the occasion to indicate time, scope and location.\u003c/p\u003e \u003cp\u003eBased on the analysis, I identified 1,028 prescriptions guiding the development and deployment of AI. I determined whether a prescription resonated based on its frequency among actors. Since collective agreement serves as the ontological foundation for a prescription to exist, I have used a minimum criterion that \u0026lsquo;for a prescription to exist, it had to be promulgated by at least two or more state and non-state actors.\u0026rsquo; Any prescription meeting these criteria is considered to resonate among all actors and, therefore, considered to be emerging.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Constructing Conceptual Typologies\u003c/h2\u003e \u003cp\u003eThe second stage of the analysis involved constructing typologies of the identified prescriptions into constitutive norms. This construction is required to solve the problem to be able to say, \u0026lsquo;what this prescription is an example of\u0026rsquo;. A typology can be theoretical or descriptive (Doty and Glick 1994). However, I am focusing on developing a conceptual nonabstract generalizable typology by creating a hierarchical system of categories to organize objects according to their similarities and differences (Mandara \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Bailey 1994). The main task of this typology is to describe the dimensions and attributes of prescriptions and find ideal types to which those dimensions and attributes may belong.\u003c/p\u003e \u003cp\u003eI employed the ideal type analysis method (Gerhardt \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). This method, sometimes known as a qualitative clustering method, involves a systematic comparison of cases or participants within a qualitative dataset to create ideal types or groups of similar cases. Together, these ideal types form a typology (Stapley, O\u0026rsquo;Keeffe, and Midgley 2022, 2; Stapley et al. 2021). I have used the already identified 1028 prescriptions as a data basis for this analysis.\u003c/p\u003e \u003cp\u003eFirst, a \u0026ldquo;criterion\u0026rdquo; needed to be determined. \u0026ldquo;These criteria define what a case must be like that fulfills in optimal fashion the characteristics postulated as indicators of the relationship under investigation\u0026rdquo; (Gerhardt \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1994\u003c/span\u003e; 100). In this study, I am using \u0026lsquo;actions and activities\u0026rsquo; associated with AI governance and their stated/intended purpose as my criterion since prescription is intended to achieve some goals by permitting some actions while prohibiting others. Additionally, typology could be constructed using two constructs (Bailey 1994; Doty and Glick 1994). Given the nature of the data, I used three constructs \u0026ndash; indicators, dimensions, and the ideal type \u0026ndash; to construct the constitutive norm type.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3 Identifying Indicators\u003c/h2\u003e \u003cp\u003eIndicators are the empirical reality (Babbie \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Since I am concerned with prescriptions that order, prohibit, or permit actions or activities, the indicators are the contents of a given prescription. I identified the contents using two linguistic clues: (1) nouns and their various forms in a sentence formation, and (2) verbs and their various forms. From a syntactical perspective, the content of a norm is the \u0026lsquo;object\u0026rsquo; of a sentence or the verb in the sentence. The answer to the question \u0026lsquo;what is to be done, and what is prescribed?\u0026rsquo; is the content of a prescription.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.2.4 Identifying Dimensions\u003c/h2\u003e \u003cp\u003eThe second stage involved grouping the indicators into dimensions. Since a single word can be used in different contexts, the only separating matrix is the goals that the actions are intended to achieve. Usually, the sentence structures, such as infinitives, infinitive phrases, and infinitive clauses, hold the clues to such goals. Additionally, I have examined the sentences and paragraphs that appeared before and after the prescriptions to contextualize and make an interpretation of the intended goals by asking \u0026lsquo;what this action/activity is aimed at\u0026rsquo;?\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.3 The Ideal Norm Types/Norm Typologies\u003c/h2\u003e \u003cp\u003eThe final step involved developing the norm typologies. Gerhardt (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1994\u003c/span\u003e) suggests that the ideal type must be constructed based on the pre-set criterion for typology development. The criterion I have set is the rational-purposive social actions, which are \u0026lsquo;means to the ends.\u0026rsquo; Finding conceptual names was an iterative process. The chief question I kept asking for each of the dimensions is this: to what purposes such and such prescriptions are promulgated? Contextualization \u0026ndash; by reading the prescriptions in the context in which they were written, and existing literature was instrumental in constructing the concepts. The benchmark I set is to determine which concepts can best capture and describe the dimensions and indicators. The tasks of coming up with the terminologies were more conceptual than technical.\u003c/p\u003e \u003cp\u003eThe dimensions and indicators within each constitutive norm type are neither exhaustive nor mutually exclusive. Such overlap occurs both within and across norm types for two main reasons. First, the AI is a multifaceted and multi-layered technological system (Gasser and Almeida 2017), and so are its applications. Consequently, the context in which a prescription is promulgated vary significantly. For example, the Datafication norm includes prescriptions such as \u0026lsquo;generate value\u0026rsquo; from data to advance AI development and \u0026lsquo;protect privacy\u0026rsquo; to foster data flow through a social contract between data owners and collectors. Though these prescriptions differ in contexts, they serve the same purpose \u0026ndash;datafication. Moreover, data\u0026rsquo;s diverse and expanding utility beyond AI development and automation adds to the heterogeneity of prescriptions within the Datafication norm. This applies to other norm type as well given the multi-layered nature of AI. Similarly, the prescription \u0026lsquo;enact law\u0026rsquo; appears across AI-fication, Datafication, and Trustworthiness norm, yet its purposes differ: enabling AI development at scale, building data infrastructure, and creating conditions for trustworthiness.\u003c/p\u003e \u003cp\u003eSecondly, from a methodological standpoint, the indicators and dimensions of ideal types need not be exhaustive and mutually exclusive (Utech 1963; Gerhardt \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1994\u003c/span\u003e; Stapley et al. 2022; 2022). Rather, each should capture and reflect the description of the ideal type to a greater degree (Stapley et al. 2022).\u003c/p\u003e \u003c/div\u003e"},{"header":"3. The Five Constitutive Norms","content":"\u003cp\u003eFollowing the ideal type analysis method, this study categorize the 1028 identified prescriptions into five constitutive norm types: (1) AI-fication, (2) Datafication, (3) Trustworthiness, (4) Anthropocentrism, and (5) Enviro-centrism. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the frequency of each norm and the actors promoting them, while Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e displays the frequency of normative effects for each norm. In what follows, I define the concept of each constitutive norm before turning to discuss the prescriptions contained within them.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.1 AI-fication\u003c/h2\u003e \u003cp\u003eI use the term AI-fication to describe the dimension and indicators related to the development and deployment of AI. I define the concept in terms of the prescriptions aimed at developing and deploying AI technologies at scale. These included creating an ecosystem, investing in infrastructure, and adopting new regulations to facilitate AI development or modify existing ones. Here, AI-fication is the ends. The dimensions and indicators are means to those ends. Technologization and digitalization could have been used instead of AI-fication. However, technologization is a broader concept that encompasses technologies beyond AI, while digitalization can be seen as just one of the indicators of AI.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Datafication\u003c/h2\u003e \u003cp\u003eI use the term \u0026lsquo;Datafication\u0026rsquo; to describe the prescriptions related to data collection, sharing, processing, and maintaining data quality. The term \u0026ldquo;datafication\u0026rdquo; was introduced by Mayer-Sch\u0026ouml;nberger and Cukier (Mayer-Sch\u0026ouml;nberger and Cukier 2013). According to them, \u003cem\u003edatafication\u003c/em\u003e refers to assigning value to information or facts (Mayer-Sch\u0026ouml;nberger and Cukier 2013). \u0026ldquo;To datafy a phenomenon is to convert it into a quantified format so it can be tabulated and analyzed\u0026rdquo; (Mayer-Sch\u0026ouml;nberger and Cukier 2013, 82). However, this definition does not capture the full range of dimensions and indicators that have emerged. They emphasized one aspect of valuation that pertains to the extensive collection and processing of data due to its value to different groups for various reasons.\u003c/p\u003e \u003cp\u003eHowever, data valuation is not a one-way street. It requires a relationship of trust between the data collector and data subject whereby rights and responsibilities could be ensured for each stakeholder. Therefore, data valuation should not only refer to valuation in the strict sense of commodification or quantification but also to the associated rights and responsibilities. Hence, I define datafication as attempts to assign value to data as public or private assets and categorize the prescriptions aimed at ensuring the free flow of data under constitutive norm \u003cem\u003edatafication\u003c/em\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Trustworthiness\u003c/h2\u003e \u003cp\u003eThe third emerging AI governance norm is trustworthiness. There is confusion between \u0026lsquo;trust\u0026rsquo; and \u0026lsquo;trustworthiness\u0026rsquo; regarding which concept could be considered the norm (Hardin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Bicchieri et al. 2011). \u003cem\u003eTrustworthiness\u003c/em\u003e refers to both cognitive and non-cognitive situations in which \u0026ldquo;B (\u003cem\u003eAI\u003c/em\u003e) is trustworthy regarding A (\u003cem\u003ehuman\u003c/em\u003e) in the domain of interaction D (\u003cem\u003ehuman-AI interactions\u003c/em\u003e) if and only if she (\u003cem\u003eB, that is, AI\u003c/em\u003e) is competent in that domain, and she (\u003cem\u003eB, that is, AI\u003c/em\u003e) considers the fact that A (\u003cem\u003ehuman\u003c/em\u003e) is relying on her, should A (\u003cem\u003ehuman\u003c/em\u003e) choose to do so in this domain, to be a compelling reason for acting as expected,\u0026rdquo; (Jones \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2012\u003c/span\u003e, 70\u0026ndash;73). This concept of \u0026ldquo;three-place trustworthiness\u0026rdquo; suggests two strategies for promoting it: 1) increasing the prevalence of \u0026ldquo;\u003cem\u003emotivational structures\u0026rdquo;\u003c/em\u003e that fundamentally enable responsiveness to dependency; 2) reducing the field of \u0026ldquo;\u003cem\u003ecompeting considerations\u0026rdquo;\u003c/em\u003e so that responsiveness to dependency will often carry the day,\u0026rdquo; (Jones \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2012\u003c/span\u003e, 73).\u003c/p\u003e \u003cp\u003eThe conceptual clarification follows Hardin's distinctions between trust and trustworthiness (Hardin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2002\u003c/span\u003e). Hardin (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2002\u003c/span\u003e) argues that \u0026ldquo;trust is little more than knowledge; trustworthiness is a motivation or a set of motivations for acting\u0026rdquo; (Hardin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2002\u003c/span\u003e, 31). To establish an environment of trustworthiness or the motivations for it, the creation of institutions is one of the processes through which trust is supported and fostered. In the context of AI governance, there appear to be prescriptions aimed at creating or modifying motivational structures by adopting new policies and practices of data collection to promote or foster trust among developers, deployers, and users, thereby enhancing the motivational framework for trustworthiness. I categorized these prescriptions aimed at enabling human responsiveness to dependence on AI as constitutive norm of Trustworthiness.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Anthropocentrism\u003c/h2\u003e \u003cp\u003eThe fourth emerging AI governance norm is anthropocentrism. \u003cem\u003eAnthropocentrism\u003c/em\u003e refers to the concept that puts humans at the center of everything. Therefore, all other beings are \u0026ldquo;means to humans\u0026rsquo; ends\u0026rdquo; (Kopnina et al. \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Crist and Kopnina 2014). However, putting \u0026lsquo;Anthropos\u0026rsquo; at the center of everything is contested. At the heart of the debate is the issue of assigning agency and whether humans should be put at the center of material and ethical concerns (Crist and Kopnina 2014). Anthropocentrism, understood from the ethical perspective, would mean that humans are at the center, and everything is a means to human ends, but that everything would require preservation and nurturing for the sake of humans\u0026rsquo; survival.\u003c/p\u003e \u003cp\u003eOn the other hand, from a material perspective, anthropocentrism would mean that humans are the center of everything, and at the expense of others. Therefore, this concept of anthropocentrism regards everything else as subhuman and not worthy, and that they can be destroyed, dislocated, and need to be dominated and controlled (Crist and Kopnina 2014). In this study, I do not define anthropocentrism from either a material or an ethical perspective of putting humans at the center. Instead, I adopt the definition of anthropocentrism more broadly and let the data speak for itself. Hence, I categorized prescriptions promulgated to serve humans and place humans at the center as they relate to AI under constitutive norm Anthropocentrism.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Enviro-centrism\u003c/h2\u003e \u003cp\u003eLike anthropocentrism, some prescriptions related to AI were promulgated that focus on ecology. These could be categorized as eco-centric. However, Gellers (Gellers \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) argues that Eco-Centrism as a concept prioritizes the entire natural ecosystem or individual organisms within it. From this holistic perspective, the empirical evidence suggests that the prescriptions made to define the scope of the relationship between nature and AI systems do not truly aim to extend moral and ethical rights to either AI systems or the ecology involved in the AI-environment relationship. Therefore, they cannot be accurately characterized as ecocentrism. Therefore, I categorize the \u003cem\u003eprescriptions\u003c/em\u003e that guide AI\u0026rsquo;s development, deployment, and interaction with the natural environment under constitutive norm Enviro-centrism.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Understanding the Prescriptions of Five Constitutive Norms","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.1 The Prescriptions of AI-fication\u003c/h2\u003e \u003cp\u003eThe prescriptions categorized under AI-fication belong to three dimensions: (1) governance and regulation, (2) development and deployment, and (3) infrastructure development. These prescriptions prescribe how AI can be developed and utilized, how to regulate AI, and how to strengthen infrastructural capabilities. Table\u0026thinsp;\u0026minus;\u0026thinsp;1 below summarizes the dimensions and prescriptions within AI-fication.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDimensions and the Prescriptions of AI-fication.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimensions of AI-fication\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrescribed Norms/Indicators\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"8\" rowspan=\"9\"\u003e \u003cp\u003eGovernance \u0026amp; Regulations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdaptive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnact Laws\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGlobal Regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportional Regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrivate Regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eState Regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCompliance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eParticipatory Governance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRegional Regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"9\" rowspan=\"10\"\u003e \u003cp\u003eDev. \u0026amp; Dep. Approach\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCocreation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTechnological Sovereignty\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExploit Potential\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePatronization\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportional Dev. \u0026amp; Dep.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExploit Data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEquitability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTechnologization\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePublic Awareness\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDigitization\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e \u003cp\u003eLogistics/Infrastructure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEliminate Barriers\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDevelop Infrastructure\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInvestment\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDev. Human Resources\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInteroperability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSafety Net\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTwo conclusions can be drawn from the prescriptions of AI-fication. First, the development of AI technology at scale will continue to grow. This applies to both civilian and military uses of AI, even in historically less technologically advanced countries. Second, the prescriptions of AI-fication suggest a flexible governance over rigid regulations. This preference arises because AI technology is still evolving, and its full benefits and harms are yet to be understood. However, there are considerable contentions and zones of agreement among state and non-state actors. The following network Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e represents the frequency of actors promulgating a particular prescription that highlights the contentions and agreements.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec20\" class=\"Section3\"\u003e \u003ch2\u003e4.1.1 Permitted Actions for Developing AI\u003c/h2\u003e \u003cp\u003e\u0026lsquo;Co-creation,\u0026rsquo; \u0026lsquo;proportionality,\u0026rsquo; and \u0026lsquo;equitability\u0026rsquo; are the three most promulgated prescriptions. Both state and non-state actors have prescribed that AI development and deployment must occur through collaborative processes between AI developers and state actors, as well as between the developers and end-users. Furthermore, they have prescribed that AI\u0026rsquo;s development and deployment must be proportional and equitable. Being proportional means maximizing AI\u0026rsquo;s benefits while mitigating its risks. Additionally, it is prescribed that state and non-state actors must \u0026lsquo;leverage AI\u0026rsquo;s potential\u0026rsquo; and achieve \u0026lsquo;technological sovereignty\u0026rsquo; through AI development. They have prescribed \u0026lsquo;mass technologization\u0026rsquo; and \u0026lsquo;state patronage.\u0026rsquo; Among the actors promulgating these norms are the United States, the United Kingdom, Sweden, Denmark, Finland, France, the European Union, the OECD, Singapore, Germany, IEEE, and Microsoft Inc.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e \u003ch2\u003e4.1.2 Building AI Infrastructure\u003c/h2\u003e \u003cp\u003eAI\u0026rsquo;s development, however, does not occur without the appropriate infrastructure. The prescriptions for advancing AI infrastructure encompass activities such as developing physical infrastructure, making investments, and cultivating human resources. The two most promulgated prescriptions from state actors are \u0026lsquo;human resources\u0026rsquo; and \u0026lsquo;infrastructure development.\u0026rsquo; States have directed relevant stakeholders to create training tools, implement worker re-skilling programs, and redesign education systems. Some states, including Belgium, have prescribed amending existing immigration laws to attract students and skilled AI workers from around the globe.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e4.1.3 Permitted Regulatory Approaches\u003c/h2\u003e \u003cp\u003e\u0026lsquo;Adaptability\u0026rsquo; in AI regulation and \u0026lsquo;participatory\u0026rsquo; governance are two prescriptions promulgated by both state and non-state actors, including the EU, OECD, IEEE, and the United States. These two prescriptions highlight the importance of including all relevant stakeholders, such as the public sector, industry, and end-users, emphasizing the collaboration of scientific experts and policymakers in technology governance.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e4.1.4 Prohibited AI Development, Regulatory, and Infrastructure Approaches\u003c/h2\u003e \u003cp\u003eAlmost all the prescriptions of AI-fication involve permission, command, and obligation. Only a few prohibitive prescriptions are promulgated within AI-fication. These prohibitive prescriptions restrict actions that could hinder the development and deployment of AI, including restrictive regulation or insufficient investments, clearly prohibiting any actions that may inhibit AI innovation. Nonetheless, there is evidence, particularly in the EU, which suggests that AI development and deployment might be prohibited when it poses risks to human rights, democracy, and the rule of law. Therefore, future prohibition on AI may be expected when AI presents clear threats to human rights, democracy, and the rule of law.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e4.2 The Prescriptions of Datafication\u003c/h2\u003e \u003cp\u003eThe prescriptions categorized under datafication belong to five dimensions: (1) approaches to data, (2) governance and regulation, (3) infrastructure/capacity, (4) data trust, and (5) data quality. These prescriptions guide how data may be perceived, regulated, infrastructure capabilities may be developed, and measures to enhance trust in and the quality of data. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e below summarizes the dimensions in the left column and the prescriptions within each dimension in the right column.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe Dimensions and the Prescriptions of Datafication.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimensions of Datafication\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrescribed Norms/Indicators\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eApproaches to Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGenerate Value\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCommodification\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGenerate Data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Sovereignty\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eGovernance \u0026amp; Regulations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnact Laws\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCompliance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMonitoring \u0026amp; Assessment\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLitigation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eInfrastructure/Capacity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDev. Data Infrastructure\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eManagement\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Resources\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Sharing\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Partnership\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"9\" rowspan=\"10\"\u003e \u003cp\u003eData Trust\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTransparency\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Subjects\u0026rsquo; Sovereignty\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrivacy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Justice\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Protection\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportionality\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccountability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDisclosure\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProfiling\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAwareness\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"6\" rowspan=\"7\"\u003e \u003cp\u003eData Quality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Standardization\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy and Reliability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntegrity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRemove Noise\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRepresentative\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSource Integrity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe prescriptions of datafication suggest, since the development of AI relies on data, that there are normative tensions between the prescriptions of datafication and the norms of human rights, especially concerning privacy, individual liberty, and sovereignty. This holds true in both democratic and non-democratic countries. The following network Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e highlights these contentions among actors and their governance preferences.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003e4.2.1 Permitted Approaches to Data\u003c/h2\u003e \u003cp\u003eIt is permitted to treat data as having intrinsic value. Notably, states prescribe to view data as \u0026lsquo;strategic assets\u0026rsquo;, \u0026lsquo;a commodity\u0026rsquo; that can be produced, assigned property rights, and sold to intended buyers under state control. Some states have commanded the creation of \u0026lsquo;data commons\u0026rsquo; to generate data and to be data-sovereign nations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003e4.2.2 Prescribed Approaches to Building Infrastructure\u003c/h2\u003e \u003cp\u003eMaintaining a data collection ecosystem requires proper infrastructure and capacity. The prescriptions for developing data infrastructure primarily comes from states, which see this as an obligation, acting individually or collaboratively through mechanisms like a union-wide data hub in the EU. Both state and non-state actors have prescribed for data sharing within and across borders. This includes countries from EU member states to the USA, UK, UAE, and India. EU member states prescribed to share personal data collected within their territories with other EU members. Additionally, the EU GDPR prescribed member states to share personal data with countries outside the region and with international organizations through bilateral or multilateral agreements. These prescriptions suggest a consensus around state-sponsored data collection and sharing mechanisms and possibly a forthcoming international agreement regarding these practices.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003e4.2.3 Prescriptions of Data Trust\u003c/h2\u003e \u003cp\u003eState and non-state actors promulgated prescriptions\u0026mdash;such as data privacy, data protection, proportionality, data justice, and the autonomy of data subjects. These prescriptions are aimed at fostering a relationship of trust between the data collectors and data subjects to facilitate data collection processes. These prescriptions are promulgated by technologically advanced, economically prosperous, and liberal democratic countries, such as the USA, the UK, and the EU member states. The prescriptions of \u0026lsquo;data privacy,\u0026rsquo; \u0026lsquo;data protection,\u0026rsquo; and \u0026lsquo;proportionality\u0026rsquo; are promulgated so data subject may trust data collector. Two prescriptions\u0026mdash; \u0026lsquo;data disclosure\u0026rsquo; and \u0026lsquo;data subject\u0026rsquo;s sovereignty\u0026rsquo;\u0026mdash;indicate a shift in how states manage the confidentiality of the information they collect. For instance, the prescription of data subjects\u0026rsquo; sovereignty prescribes that data owners must have complete control over their personal information. This implies that any data collection, whether public or private, must obtain consent before collecting personal data, inform data subjects of any processes performed on their data, and ensure that data subjects always have access to their information.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section3\"\u003e \u003ch2\u003e4.2.4 Prescriptions for Data Governance\u003c/h2\u003e \u003cp\u003eState and non-state actors prescribed requiring data collectors to \u0026lsquo;comply with\u0026rsquo; applicable privacy laws and prescribing to \u0026lsquo;enact laws.\u0026rsquo; \u0026lsquo;Monitoring and assessment\u0026rsquo; is another prescribed obligation for state-appointed data stewards and prescribing non-state actors, particularly technology corporations, to have their own data stewards.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section3\"\u003e \u003ch2\u003e4.2.5 Prescribed Prohibitions of Datafication\u003c/h2\u003e \u003cp\u003eThere are some prescriptions categorized under datafication. The EU, India, and the UK forbid any organization, whether public or private, from discriminating based on race, color, or religion or from obstructing access to services based on the personal data they hold about an individual. The EU, the UK, India, the Trade Union Congress (TUC), and the USA prescribed that no organization, public or private, should collect personal identifiers during data collection, nor disclose any personal data without obtaining clear consent from the data subjects. Singapore prescribes that consent cannot be forced or manipulated. It also prohibits organizations from disclosing when an individual\u0026rsquo;s data is shared with law enforcement and bans entities from transferring personal data outside the country. The EU prohibits the transfer and sharing of EU citizens\u0026rsquo; data outside the region without their permission, either through bilateral or multilateral agreements.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e4.3 The Prescriptions of Trustworthiness\u003c/h2\u003e \u003cp\u003eThe prescriptions of trustworthiness belong to four dimensions: (1) AI functionality, (2) outcomes, (3) development processes, and (4) governance. Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e below summarizes the dimensions in the left column and the promulgated prescriptions within each dimension in the right column.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDimensions and the Prescriptions of Trustworthiness.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimensions of Trustworthiness\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrescribed Norms/Indicators\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"7\" rowspan=\"8\"\u003e \u003cp\u003eAI Functionality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy \u0026amp; Reliability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSafety \u0026amp; Security\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdaptability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTransparency\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraceability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRobustness\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExplainability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReproducibility\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e \u003cp\u003eAI Outcome\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFairness\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNon-discrimination\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEquitability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo-Harm\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSustainability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTransparency\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"6\" rowspan=\"7\"\u003e \u003cp\u003eDevelopment \u0026amp; Deployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDiversity \u0026amp; Inclusion\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHuman-centric\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportionality\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePublic Good\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEmpowerment\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurveillance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eComplementarity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"10\" rowspan=\"11\"\u003e \u003cp\u003eGovernance \u0026amp; Regulations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccountability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEthical\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFiduciary Duty\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMonitoring \u0026amp; Assessment\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAudits \u0026amp; Oversight\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSocial Safety Net\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCompliance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrivate Regulations\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eState Regulations\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportionality\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnact Laws\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe prescriptions suggest societal-level tensions about trusting AI technology. This mistrust could impact the widespread adoption of AI technologies. The following network Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows the frequency of prescriptions by actors.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec31\" class=\"Section3\"\u003e \u003ch2\u003e4.3.1 Prescriptions for AI\u0026rsquo;s Functionality, Outcome, and Development Processes\u003c/h2\u003e \u003cp\u003eTo establish trust in AI technology, state and non-state actors have prescribed that the development of AI must incorporate diverse groups of people and cultures, and remain transparent throughout the design, development, and deployment processes. Furthermore, they prescribed that AI must function safely, accurately, and reliably, that its outcomes be transparent, fair, secure, and explainable to users, and that AI\u0026rsquo;s functionalities be traceable. State and non-state actors promulgated that AI systems must be made \u0026lsquo;safe and secure\u0026rsquo; from development through implementation and use cases. The prescription of \u0026lsquo;transparency\u0026rsquo; involves activities, such as disclosing how AI systems are constructed, how they operate, and what, how, and why decisions are made by the AI system and the individuals involved in these processes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec32\" class=\"Section3\"\u003e \u003ch2\u003e4.3.2 Prescribed Regulatory Approaches to Trustworthiness\u003c/h2\u003e \u003cp\u003eThe actors prescribed to ensure \u0026lsquo;accountability\u0026rsquo; in the event of adverse effects. This applies to both civil and criminal liabilities. State and non-state actors prescribed creation of a \u0026lsquo;monitoring and assessment\u0026rsquo; and \u0026lsquo;audit and oversight\u0026rsquo; mechanisms throughout the entire lifecycle of AI systems. Evidence indicates that a consensus exists among most states for implementing such a \u0026lsquo;monitoring and evaluation\u0026rsquo; system. For instance, member states of UNESCO have agreed in principle to establish such a mechanism.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec33\" class=\"Section3\"\u003e \u003ch2\u003e4.3.3 Prohibitions to Enhance Trustworthiness\u003c/h2\u003e \u003cp\u003e\u0026lsquo;No-harm\u0026rsquo; and \u0026lsquo;non-discrimination\u0026rsquo; are two prohibitive prescriptions promulgated within trustworthiness. These prescriptions prohibit the outcome of any AI system from being \u0026lsquo;harmful\u0026rsquo; to its developers, deployers, and users directly or indirectly, intentionally or unintentionally. Some of the prohibitive statements specified harms in a broader context, including physical and psychological harms. For example, companies that produce AI systems are prohibited from selling them to actors with a history of human rights violations or are deemed likely to violate them. The \u0026lsquo;non-discrimination\u0026rsquo; prescription prohibits AI developers from deploying AI systems that might result in discriminatory outcomes, particularly in judicial systems, law enforcement agencies, or in contexts in which AI systems filter based on religion, race, ethnicity, locations, opinion, or due to membership of a group.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec34\" class=\"Section2\"\u003e \u003ch2\u003e4.4 The Prescriptions of Anthropocentrism\u003c/h2\u003e \u003cp\u003eThe prescriptions of anthropocentrism belong to three dimensions: (1) the \u0026lsquo;man behind the machine,\u0026rsquo; (2) \u0026lsquo;status,\u0026rsquo; and (3) the \u0026lsquo;development and deployment\u0026rsquo; of AI. The prescriptions within each dimension delineate roles and status in human-machine interactions and their primacy in relation to one another. Table\u0026thinsp;\u0026minus;\u0026thinsp;4 below summarizes the dimensions in the left column and the prescriptions within each dimension in the right column.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDimensions and the Prescriptions of Anthropocentrism.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimensions of Anthropocentrism\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrescribed Norms/Indicators\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eMan Behind Machine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccountability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHuman-Control\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStatusing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHuman Dignification\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHuman Autonomy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eDevelopment \u0026amp; Deployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBeneficial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSafety \u0026amp; Security\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe prescriptions suggest that \u0026lsquo;human control\u0026rsquo; is likely to be a contentious governance feature, with differentiated application in civilian and military uses. They also suggest a strong predominance of material rather than ethical anthropocentrism discussed earlier. The following network Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e highlights the frequency of prescriptions and actors promulgating them.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec35\" class=\"Section3\"\u003e \u003ch2\u003e4.4.1 Prescriptions of Human Agency\u003c/h2\u003e \u003cp\u003e\u0026lsquo;Human autonomy\u0026rsquo; and \u0026lsquo;human dignification\u0026rsquo; are two main prescriptions. State and non-state actors have prescribed that humans should be valued above machines and that machines must retain their status as tools, obeying commands from humans rather than intelligent entities that could be granted rights. Machines are prescribed to be subordinate to human commands and to adhere to the ethical and normative values of human society, thereby upholding human sovereignty and autonomy in human-machine interactions. Actors who mainly prescribed these are the U.N., EU, ITU, IEEE, and UNESCO.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec36\" class=\"Section3\"\u003e \u003ch2\u003e4.4.2 Prescriptions of Man Behind the Machine\u003c/h2\u003e \u003cp\u003eStates and non-state actors have prescribed that, in developing AI systems, \u0026lsquo;human control\u0026rsquo; must be established and that AI systems must serve to \u0026lsquo;benefit humans.\u0026rsquo; This prescription is promulgated in response to the potential for AI systems to be given full autonomy, allowing them to operate independently and adapt to various environments. It is also prescribed that, in the event of an adverse outcome, \u0026lsquo;accountability\u0026rsquo; must rest with humans, not the machine. Actors like the U.N., EU, G-20, IGF, IEEE, Internet Society, ITU, ITI, OECD, and UNESCO are non-state actors that promulgated the prescriptions. States such as France, Germany, the Netherlands, and the U.S. have also promulgated the prescriptions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec37\" class=\"Section3\"\u003e \u003ch2\u003e4.4.3 Prohibitions of Anthropocentrism\u003c/h2\u003e \u003cp\u003eThe prohibitive prescriptions address actions and activities related to violations of human sovereignty and dignity, ensuring that humans remain in control of AI systems and that no harmful AI systems are developed. The prescriptions prohibiting these actions and activities can be interpreted as obligatory, as both obligations and prohibitions are inter-definable in this case.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec38\" class=\"Section2\"\u003e \u003ch2\u003e4.5 The Prescriptions of Enviro-centrism\u003c/h2\u003e \u003cp\u003eThe prescriptions of Enviro-centrism belong to one dimension: (1) development and deployment. The prescriptions include emission reduction, energy optimization, greening the value chain, and proportionality. Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e below presents the dimension in the left column and the promulgated prescriptions in the right column.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe Dimension and the Prescriptions of Enviro-centrism.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimensions of Enviro-centrism\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrescribed Norms/Indicators\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eDevelopment \u0026amp; Deployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEmission Reduction\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnergy Optimization\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGreening Value Chain\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProportionality\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSustainability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAmong the five constitutive norms, Enviro-centrism is the least prescribed. This may be due to the impacts of AI on the environment, and the relationship between the two is still largely unknown. The following network Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e highlights the frequency of prescriptions and actors promulgating them.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAll prescriptions are obligatory in their effects except for one, which is a prohibitive promulgation. The EU Commission, France, Hungary, Switzerland, and the ITU promulgate these prescriptions. They prescribe for the development and deployment of AI systems to reduce emissions, enhance the value chain, and, most importantly, align with environmental costs and benefits. Similarly, they promulgate that AI must be used to enhance the value chain, reduce emissions, and mitigate environmental degradation, thereby establishing a relationship where technologies serve the environment. The only prohibitive promulgation comes from Switzerland, which prescribes that the energy and materials consumption by AI must not increase in line with the growing use of these technologies.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study demonstrates that, even though there is no formal global agreement on AI regulation as yet, AI governance is already happening through the creation of norms. This study identifies these norms by examining publicly available AI governance discourses. The findings suggest five main types of emergent constitutive AI governance norms, each containing a set of prescriptions regulating AI development and deployment. It is unknown when and how a global AI governance architecture will formally take shape. But as theory suggests, these five emergent [constitutive] norms, acting as the rules of the rulemaking, are likely to guide any future global, transnational, and local AI governance. The prescriptions of each of the main norm types would determine how AI is developed and used.\u003c/p\u003e \u003cp\u003eThese norms are still in the early stages of their emergence. As this study suggests, these are still being contested nationally and internationally, involving various state and non-state actors for local interpretation and adaptation. Aside from states, these actors include norm entrepreneurs such as international organizations, civil society, human rights organizations, networks of NGOs, special interest groups such as technology corporations, and intergovernmental organizations, who have interests and stakes in the ways AI governance takes shape. This contestation among the various AI stakeholders and their participation in different venues indicates the norm emergence mechanism of what Finnemore and Sikkink (1998) suggest, the strategic construction of norms by actors to solve the AI governance problem that affects them all (Ullman-Margalit 1977; Coleman \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e1990\u003c/span\u003e; Finnemore and Sikkink 1998). Future studies may examine AI governance to explain why AI governance norms are emerging in their current form and the influence these actors may have on one another in setting the AI governance agenda globally.\u003c/p\u003e \u003cp\u003eIn addition, norm emergence theories suggest a co-constitutive mechanism, in which the emergence of a norm is shaped by actors\u0026rsquo; interests, opportunity structures, and fitness of new norms with the existing normative foundation (O\u0026rsquo;Neill et al. \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; VanDeveer \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Klotz \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Nadelmann \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e1990\u003c/span\u003e; Sandholtz \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Risse-Kappen \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e1994\u003c/span\u003e; Risse \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2000\u003c/span\u003e). As this study suggests, numerous non-state actors are involved in setting the global AI governance agenda and attempting to influence the direction, nature, and shape of global AI governance architecture. These actors carry significant meaning about their preferences and power dynamics in global politics. Studies suggest that non-technical actors and scientific experts are both involved in the design, implementation, and adoption of science and technologies in society, as well as in technological decision-making at both global and local levels in different issues such as environmental governance, biotechnology, and nanotechnology governance (Bucchi and Neresini 2008; VanDeveer \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). It is not clear how and to what extent these actors exert influence on one another in the case of AI governance. Future studies may examine the dynamics of actors\u0026rsquo; preferences and the extent and mechanisms of their influence on one another, as well as the material and ideational opportunity structures that facilitate such actors\u0026rsquo; participation in norm-generating platforms in shaping the global AI governance regime.\u003c/p\u003e \u003cp\u003eFurthermore, norms do not only emerge, they also disappear and die (Panke and Petersohn 2016; 2012). It will be interesting to observe whether these constitutive norms continue to gain momentum or might as well disappear with the evolution of AI technology and the governance landscape.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAgree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eAcknowledgment I express my gratitude to Professor Stacy D. VanDeveer, Professor J. Samuel Barkin, and Professor Karen Ross at the University of Massachusetts Boston for their important contribution in conducting this research and feedback on an earlier version of this manuscript. I also thank Professor Joshua C. Gellers of the University of North Florida for his important contribution to the development of typologies in this study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll data supporting the findings of this study are available within the paper and its Supplementary Information. Microsatellite primer sequences are provided in Supplementary Table 2, along with original reference describing the microsatellites used in this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAxelrod R (1986) An Evolutionary Approach to Norms. Am Polit Sci Rev 80(4):1095\u0026ndash;1111. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2307/1960858\u003c/span\u003e\u003cspan address=\"10.2307/1960858\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBabbie ER (2016) \u003cem\u003eThe Practice of Social Research\u003c/em\u003e. Fourteenth edition. Cengage Learning\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBailey, Kenneth D (1994) Typologies and Taxonomies An Introduction to Classification Techniques. Sage\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarkin JS, Sjoberg L (eds) (2017) Interpretive Quantification: Methodological Explorations for Critical and Constructivist IR. University of Michigan Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBicchieri C, Erte Xiao, and Ryan Muldoon (2011) Trustworthiness Is a Social Norm, but Trusting Is Not. Politics Philos Econ 10(2):170\u0026ndash;187. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1470594X10387260\u003c/span\u003e\u003cspan address=\"10.1177/1470594X10387260\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBiersteker T (2015) Global Governance. On Governance: What It Is, What It Means and Its Policy Uses. McGill-Queen\u0026rsquo;s University. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.jstor.org/stable/j.ctt1jktrbx.12\u003c/span\u003e\u003cspan address=\"http://www.jstor.org/stable/j.ctt1jktrbx.12\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBode I (2023) Practice-Based and Public-Deliberative Normativity: Retaining Human Control over the Use of Force. Eur J Int Relat 29(4):990\u0026ndash;1016. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/13540661231163392\u003c/span\u003e\u003cspan address=\"10.1177/13540661231163392\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBucchi M, and Federico Neresini (2008) Science and Public Participation. The Handbook of Science and Technology Studies. The MIT Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarr EH (1946) \u003cem\u003eThe Twenty Years\u0026rsquo; Crisis, 1919\u0026ndash;1939: Reissued with a New Preface from Michael Cox\u003c/em\u003e. Edited by Michael Cox. Palgrave Macmillan\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColeman JS (1990) Foundations of Social Theory. Harvard University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrist E, and Helen Kopnina (2014) Unsettling Anthropocentrism. Dialect Anthropol 38(4):387\u0026ndash;396. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10624-014-9362-1\u003c/span\u003e\u003cspan address=\"10.1007/s10624-014-9362-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoty D, Harold, Glick WH (1994) Typologies as a Unique Form of Theory Building: Toward Improved Understanding and Modeling. Acad Manage Rev 19(2):230. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2307/258704\u003c/span\u003e\u003cspan address=\"10.2307/258704\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFairciough N (2013) Language and Power, 2nd edn. Taylor and Francis\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFinnemore M, and Kathryn Sikkink (1998) International Norm Dynamics and Political Change. Int Org 52(4):887\u0026ndash;917. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/002081898550789\u003c/span\u003e\u003cspan address=\"10.1162/002081898550789\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGasser U, Virgilio AF, Almeida (2017) A Layered Model for AI Governance. IEEE Internet Comput 21(6):58\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/MIC.2017.4180835\u003c/span\u003e\u003cspan address=\"10.1109/MIC.2017.4180835\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGellers J (2021) Rights for Robots: Artificial Intelligence, Animal and Environmental Law. Taylor \u0026amp; Francis\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerhardt U (1994) The Use of Weberian Ideal-Type Methodology in Qualitative Data Interpretation: An Outline for Ideal-Type Analysis. Bull Sociol Methodology/Bulletin de M\u0026eacute;thodologie Sociologique 45(1):74\u0026ndash;126. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/075910639404500105\u003c/span\u003e\u003cspan address=\"10.1177/075910639404500105\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGherhes C, Yu Z, Vorley T, and Lan Xue (2023) Technological Trajectories as an Outcome of the Structure-Agency Interplay at the National Level: Insights from Emerging Varieties of AI. World Dev 168(August):106252. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.worlddev.2023.106252\u003c/span\u003e\u003cspan address=\"10.1016/j.worlddev.2023.106252\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoldstein J, Keohane RO (eds) (1993) \u003cem\u003eIdeas and Foreign Policy: Beliefs, Institutions, and Political Change\u003c/em\u003e. Cornell Studies in Political Economy. Cornell University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGupta A, Biermann F, Van Driel E et al (2024) Towards a Non-Use Regime on Solar Geoengineering: Lessons from International Law and Governance. Transnatl Environ Law 13(2):368\u0026ndash;399. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S2047102524000050\u003c/span\u003e\u003cspan address=\"10.1017/S2047102524000050\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHabermas J\u0026uuml;rgen, McCarthy TA (2005) and J\u0026uuml;rgen Habermas. \u003cem\u003eLifeworld and System: A Critique of Functionalist Reason\u003c/em\u003e. 1. digital-Print ed. The Theory of Communicative Action / J\u0026uuml;rgen Habermas. Transl. by Thomas MacCarthy, Vol. 2. Beacon\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHardin R (2002) Trust and Trustworthiness. The Russell Sage Foundation Series on Trust, vol 4. Russell Sage Foundation\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHart HLA (1994) The Concept of Law, 2nd edn. Clarendon Law Series. Clarendon\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJepperson RL, Wendt A, Peter JK (1996) Norms, Identity and Culture in National Security. The Culture of National Security: Norms and Identity in World Politics. Columbia University, New York\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones K (2012) Trustworthiness Ethics 123(1):61\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1086/667838\u003c/span\u003e\u003cspan address=\"10.1086/667838\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKatzenstein PJ (ed) (1996) The Culture of National Security: Norms and Identity in World Politics. New Directions in World Politics. Columbia Univ.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeohane RO (1984) \u003cem\u003eAfter Hegemony: Cooperation and Discord in the World Political Economy\u003c/em\u003e. 1st Princeton classic ed. A Princeton Classic Edition. Princeton University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeohane RO (1989) International Institutions and State Power: Essays in International Relations Theory, 1st edn. Westview. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4324/9780429032967\u003c/span\u003e\u003cspan address=\"10.4324/9780429032967\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlotz A (2002) Transnational Activism and Global Transformations: The Anti-Apartheid and Abolitionist Experiences. Eur J Int Relat 8(1):49\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1354066102008001002\u003c/span\u003e\u003cspan address=\"10.1177/1354066102008001002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKopnina H, Washington H, Taylor B, John JP (2018) Anthropocentrism: More than Just a Misunderstood Problem. J Agric Environ Ethics 31(1):109\u0026ndash;127. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10806-018-9711-1\u003c/span\u003e\u003cspan address=\"10.1007/s10806-018-9711-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKowert P, and Jeffrey Legro (1996) Norms, Identity, and Their Limits: A Theoretical Reprise. The Culture of National Security: Norms and Identity in World Politics. Columbia Univ.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrathwohl DR (2002) A Revision of Bloom\u0026rsquo;s Taxonomy: An Overview. Theory Into Pract 41(4):212\u0026ndash;218. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/s15430421tip4104_2\u003c/span\u003e\u003cspan address=\"10.1207/s15430421tip4104_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKratochwil FV (1989) Rules, Norms, and Decisions: On the Conditions of Practical and Legal Reasoning in International Relations and Domestic Affairs. Reprint. Cambridge Studies in International Relations, vol 2. Cambridge Univ. Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLawless S, Song AM, Cohen PJ, and Tiffany H. Morrison (2020) Rights, Equity and Justice: A Diagnostic for Social Meta-Norm Diffusion in Environmental Governance. Earth Syst Gov 6(December):100052. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.esg.2020.100052\u003c/span\u003e\u003cspan address=\"10.1016/j.esg.2020.100052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLegro J (1995) Cooperation under Fire: Anglo-German Restraint during World War II. Cornell University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMandara J (2003) The Typological Approach in Child and Family Psychology: A Review of Theory, Methods, and Research. Clin Child Fam Psychol Rev 6(2):129\u0026ndash;146. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1023734627624\u003c/span\u003e\u003cspan address=\"10.1023/A:1023734627624\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarch JG, Olsen JP (1998) The Institutional Dynamics of International Political Orders. Int Org 52(4):943\u0026ndash;969. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/002081898550699\u003c/span\u003e\u003cspan address=\"10.1162/002081898550699\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer-Sch\u0026ouml;nberger V, and Kenneth Cukier (2013) Big Data: A Revolution That Will Transform How We Live, Work, and Think. Houghton Mifflin Harcourt\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMilliken J (1999) The Study of Discourse in International Relations: A Critique of Research and Methods. Eur J Int Relat 5(2):225\u0026ndash;254. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1354066199005002003\u003c/span\u003e\u003cspan address=\"10.1177/1354066199005002003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNadelmann EA (1990) Global Prohibition Regimes: The Evolution of Norms in International Society. Int Org 44(4):479\u0026ndash;526. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0020818300035384\u003c/span\u003e\u003cspan address=\"10.1017/S0020818300035384\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNelson J, and Tim Gorichanaz (2019) Trust as an Ethical Value in Emerging Technology Governance: The Case of Drone Regulation. Technol Soc 59:101131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2019.04.007\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2019.04.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD (2021) Database of National AI Policies. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://oecd.ai\u003c/span\u003e\u003cspan address=\"https://oecd.ai\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO\u0026rsquo;Neill K, Balsiger J\u0026ouml;rg, VanDeveer SD (2004) ACTORS, NORMS, AND IMPACT: Recent International Cooperation Theory and the Influence of the Agent-Structure Debate. Annu Rev Polit Sci 7(1):149\u0026ndash;175. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1146/annurev.polisci.7.090803.161821\u003c/span\u003e\u003cspan address=\"10.1146/annurev.polisci.7.090803.161821\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOpp K-D (1982) The Evolutionary Emergence of Norms. Br J Soc Psychol 21(2):139\u0026ndash;149. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.2044-8309.1982.tb00522.x\u003c/span\u003e\u003cspan address=\"10.1111/j.2044-8309.1982.tb00522.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalladino N (2023) A \u0026lsquo;Biased\u0026rsquo; Emerging Governance Regime for Artificial Intelligence? How AI Ethics Get Skewed Moving from Principles to Practices. Telecomm Policy 47(5):102479. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.telpol.2022.102479\u003c/span\u003e\u003cspan address=\"10.1016/j.telpol.2022.102479\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanke D, and Ulrich Petersohn (2012) Why International Norms Disappear Sometimes. Eur J Int Relat 18(4):719\u0026ndash;742. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1354066111407690\u003c/span\u003e\u003cspan address=\"10.1177/1354066111407690\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanke D, and Ulrich Petersohn (2016) Norm Challenges and Norm Death: The Inexplicable? Cooperation Confl 51(1):3\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0010836715597948\u003c/span\u003e\u003cspan address=\"10.1177/0010836715597948\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrice R (1998) Reversing the Gun Sights: Transnational Civil Society Targets Land Mines. Int Org 52(3):613\u0026ndash;644. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/002081898550671\u003c/span\u003e\u003cspan address=\"10.1162/002081898550671\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadu R (2021) Steering the Governance of Artificial Intelligence: National Strategies in Perspective. Policy Soc 40(2):178\u0026ndash;193. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14494035.2021.1929728\u003c/span\u003e\u003cspan address=\"10.1080/14494035.2021.1929728\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRisse T (2000) Let\u0026rsquo;s Argue!\u0026rsquo;: Communicative Action in World Politics. Int Org 54(1):1\u0026ndash;39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/002081800551109\u003c/span\u003e\u003cspan address=\"10.1162/002081800551109\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRisse T (2002) Nationalism and Collective Identities. Europe versus the Nation-State? Developments West Eur Politics no. 2:77\u0026ndash;93\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRisse-Kappen T (1994) Ideas Do Not Float Freely: Transnational Coalitions, Domestic Structures, and the End of the Cold War. Int Org 48(2):185\u0026ndash;214. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0020818300028162\u003c/span\u003e\u003cspan address=\"10.1017/S0020818300028162\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosert E (2019) Norm Emergence as Agenda Diffusion: Failure and Success in the Regulation of Cluster Munitions. Eur J Int Relat 25(4):1103\u0026ndash;1131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1354066119842644\u003c/span\u003e\u003cspan address=\"10.1177/1354066119842644\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuggie JG (1998) What Makes the World Hang Together? Neo-Utilitarianism and the Social Constructivist Challenge. Int Org 52(4):855\u0026ndash;885. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/002081898550770\u003c/span\u003e\u003cspan address=\"10.1162/002081898550770\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaheb T, and Tayebeh Saheb (2023) Topical Review of Artificial Intelligence National Policies: A Mixed Method Analysis. Technol Soc 74:102316. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2023.102316\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2023.102316\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSandholtz W (2008) Dynamics of International Norm Change: Rules against Wartime Plunder. Eur J Int Relat 14(1):101\u0026ndash;131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1354066107087766\u003c/span\u003e\u003cspan address=\"10.1177/1354066107087766\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStapley E, Sally O\u0026rsquo;Keeffe, and Nick Midgley (2022) Developing Typologies in Qualitative Research: The Use of Ideal-Type Analysis. Int J Qualitative Methods 21(April):16094069221100633. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/16094069221100633\u003c/span\u003e\u003cspan address=\"10.1177/16094069221100633\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStix C (2021) The Ghost of AI Governance Past, Present and Future: AI Governance in the European Union. Version 1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/ARXIV.2107.14099\u003c/span\u003e\u003cspan address=\"10.48550/ARXIV.2107.14099\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Preprint, arXiv\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTan SY, and Araz Taeihagh (2021a) Adaptive Governance of Autonomous Vehicles: Accelerating the Adoption of Disruptive Technologies in Singapore. Government Inform Q 38(2):101546. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.giq.2020.101546\u003c/span\u003e\u003cspan address=\"10.1016/j.giq.2020.101546\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTan SY, and Araz Taeihagh (2021b) Governing the Adoption of Robotics and Autonomous Systems in Long-Term Care in Singapore. Policy Soc 40(2):211\u0026ndash;231. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14494035.2020.1782627\u003c/span\u003e\u003cspan address=\"10.1080/14494035.2020.1782627\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe AI Act of the European Parliament and of the, Council (2024) /1689 Pub. L. No. (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://eur-lex.europa.eu/eli/reg/2024/1689\u003c/span\u003e\u003cspan address=\"https://eur-lex.europa.eu/eli/reg/2024/1689\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTieku TK (2019) Ruling from the Shadows: The Nature and Functions of Informal International Rules in World Politics. Int Stud Rev 21(2):225\u0026ndash;243. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/isr/viz012\u003c/span\u003e\u003cspan address=\"10.1093/isr/viz012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuring AM (2009) Computing Machinery and Intelligence. In \u003cem\u003eParsing the Turing Test\u003c/em\u003e, edited by Robert Epstein, Gary Roberts, and Grace Beber. Springer Netherlands. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-1-4020-6710-5_3\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4020-6710-5_3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUllmann-Margalit E (1977) \u003cem\u003eThe Emergence of Norms\u003c/em\u003e. 1. published in paperback. Clarendon Library of Logic and Philosophy. Clarendon Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUtech, Myron R (1963) The Social Action Approach of Max Weber. Thesis, Bowling Green State University\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanDeveer S (2004) Ordering Environments: Regions in European International Environmental Cooperation. In \u003cem\u003eEarthly Politics: Local and Global in Environmental Governance\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanDeveer S (2011) Networked Baltic Environmental Cooperation. J Baltic Stud 42(1):37\u0026ndash;55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01629778.2011.538516\u003c/span\u003e\u003cspan address=\"10.1080/01629778.2011.538516\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiener A (2009) Enacting Meaning-in-Use: Qualitative Research on Norms and International Relations. Rev Int Membr 35(1):175\u0026ndash;193. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0260210509008377\u003c/span\u003e\u003cspan address=\"10.1017/S0260210509008377\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiener A (2014) A Theory of Contestation. SpringerBriefs in Political Science. Springer\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWodak R (ed) (2014) \u003cem\u003eMethods of Critical Discourse Analysis\u003c/em\u003e. 2. Aufl., Repr. Introducing Qualitative Methods. SAGE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evon Wright GH (1963) \u003cem\u003eNorm and Action: A Logical Enquiry\u003c/em\u003e. Reprinted. International Library of Philosophy and Scientific Method. Routledge \u0026amp; Kegan Paul [u.a.]\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZ\u0026uuml;rn M (2018) \u003cem\u003eA Theory of Global Governance: Authority, Legitimacy, and Contestation\u003c/em\u003e. First edition. Oxford University Press\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Governance of Artificial Intelligence, Norm Emergence, AI Governance Norms, Technology Governance, Norm Identification.","lastPublishedDoi":"10.21203/rs.3.rs-9383807/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9383807/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eEfforts to establish a global Artificial Intelligence (AI) regulatory regime are occurring in various locations through transnational debates and governance proposals. Yet, our understanding of the dynamics of a global AI governance regime is unclear. This study identifies emerging global norms governing the development and deployment of AI technologies. I examine (a) 345 policies, white papers, and executive orders published by fifty-nine states; (b) 873 AI governance-related debates, conferences, and discussions hosted by or on the platforms of fifteen international organizations between January 2000 and January 2022; and (c) 87 AI governance frameworks proposed by 54 non-state actors. Utilizing G. H. von Wright\u0026rsquo;s norm analysis framework, I identify 1,028 prescriptive norms that permit, obligate, and prohibit AI development and deployment practices. I then categorize them using the Ideal Type Analysis method, suggesting five constitutive norms emerging: \u003cem\u003e\u0026lsquo;AI-fication,\u0026rsquo; \u0026lsquo;datafication,\u0026rsquo; \u0026lsquo;trustworthiness,\u0026rsquo; \u0026lsquo;anthropocentrism,\u0026rsquo;\u003c/em\u003e and \u003cem\u003e\u0026lsquo;enviro-centrism.\u0026rsquo;\u003c/em\u003e These emergent norms will influence AI technologies in two ways. First, they will regulate human behavior and create new actors through socialization processes regarding how AI technologies are developed and used. Second, since norms function as secondary rules, these five AI governance norms will define and influence the dynamics of an AI governance regime globally.\u003c/p\u003e","manuscriptTitle":"Governing Artificial Intelligence (AI) Through Creating Governance Norms: Identifying the Emerging Global AI Governance Norms.","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-07 11:56:26","doi":"10.21203/rs.3.rs-9383807/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ee39f21f-62d0-45c6-94d7-f7d62fbfceea","owner":[],"postedDate":"May 7th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"100280373391649993484150777603489196592","date":"2026-05-04T17:53:58+00:00","index":37,"fulltext":""},{"type":"reviewerAgreed","content":"77055388497897651496477844027376613877","date":"2026-04-30T18:07:05+00:00","index":32,"fulltext":""},{"type":"reviewerAgreed","content":"121868385509753377436987855877431329244","date":"2026-04-29T09:36:33+00:00","index":28,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-07T11:56:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-07 11:56:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9383807","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9383807","identity":"rs-9383807","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}