Threat Conceptions in Global Security Discourse: Analyzing the Speech Records of the United Nations Security Council, 1990–2019

Abstract

Since the end of the Cold War, the notion of global security, and presumed threats to it, has undergone considerable expansion and diversification. This process has been led by the United Nations Security Council (UNSC), where active deliberations concerning “threat(s) to the peace” have taken place among major international actors. Despite a sizable accumulation of scholarly arguments, however, the defining features of the structure and dynamics of the post-Cold War security discourse remain ambiguous. To address these ambiguities, this study investigates the entire body of Council deliberations over the past three decades. Based on an original dataset consisting of policy statements delivered at the UNSC, the study employs quantitative text analysis tools, including word embedding, to examine how council members have conceived the notion of security threat in terms of the various issues and entities discussed. It shows the security discourse at the UNSC to be highly stratified and reveals the persistent and pervasive influence of terrorism and the proliferation of weapons of mass destruction, which constitute the limited common grounds shared by the Council’s permanent members. These findings caution against the unconstrained use of certain theoretical constructs widely employed in other works, most notably, “securitization” and “interpretative community.”

Desde el final de la Guerra Fría, tanto la noción de seguridad mundial como las amenazas que, presuntamente, pesan sobre ella, han experimentado una considerable expansión y diversificación. Este proceso ha sido dirigido por el Consejo de Seguridad de las Naciones Unidas (CSNU), donde se han llevado a cabo deliberaciones activas sobre «amenaza(s) a la paz» por parte de los principales agentes internacionales. Sin embargo, a pesar de que se ha ido acumulando una cantidad considerable de argumentos académicos, las características definitorias de la estructura y la dinámica del discurso de seguridad posterior a la Guerra Fría siguen siendo ambiguas. Con el fin de abordar estas ambigüedades, este estudio investiga todas las deliberaciones del Consejo durante las últimas tres décadas. Basándonos en un conjunto de datos original que consiste en declaraciones de políticas pronunciadas en el CSNU, el estudio emplea herramientas cuantitativas de análisis de texto, tales la incrustación de palabras (word embedding), con el fin de estudiar la forma en que los miembros del consejo concibieron la noción de amenaza a la seguridad en términos de los diversos temas y entidades que se debatían. El estudio demuestra que el discurso en materia de seguridad dentro del CSNU está altamente estratificado y revela la influencia, persistente y generalizada, del terrorismo y la proliferación de armas de destrucción masiva, los cuales constituyen los limitados puntos en común compartidos por los miembros permanentes del Consejo. Estas conclusiones nos alertan sobre el uso ilimitado de ciertas construcciones teóricas que son ampliamente empleadas en otros trabajos, en particular, la «securitización» y la «comunidad interpretativa».

Depuis la fin de la guerre froide, la notion de sécurité mondiale, ainsi que les éléments qui la menaceraient, s'est considérablement élargie et diversifiée. Ce processus a été mené par le Conseil de sécurité des Nations unies (CSNU), qui a tenu des délibérations actives concernant les « menaces pour la paix » entre les principaux acteurs internationaux. Toutefois, malgré une importante accumulation d'arguments académiques, une ambiguïté subsiste concernant les caractéristiques définitoires de la structure et de la dynamique du discours d'après-guerre froide relatif à la sécurité. Pour l’éclaircir, cette étude analyse toutes les délibérations du Conseil de ces trois dernières décennies. En s'appuyant sur un ensemble de données original constitué de déclarations politiques prononcées au CSNU, l’étude emploie des outils d'analyse textuelle quantitative, notamment le plongement lexical, pour examiner comment les membres du conseil ont conçu la notion de menace de sécurité par rapport aux différentes problématiques et entités étudiées. Elle fait apparaître la stratification extrême du discours relatif à la sécurité au sein du CSNU, mais aussi l'influence persistante et généralisée du terrorisme et de la prolifération des armes de destruction massive, qui constituent les quelques points communs des membres permanents du Conseil. Ces résultats nous mettent en garde contre l'utilisation sans réserve de certains concepts théoriques largement employés dans d'autres travaux, notamment la « sécuritisation » et la « communauté interprétative ».

Introduction

On January 31, 1992, the first summit-level meeting in the history of the United Nations Security Council (UNSC) was concluded by its president, on behalf of the attending heads of state, with the following statement:

The absence of war and military conflicts amongst States does not in itself ensure international peace and security. The non-military sources of instability in the economic, social, humanitarian and ecological fields have become threats to peace and security. The United Nations membership as a whole, working through the appropriate bodies, needs to give the highest priority to the solution of these matters. (UN Document S/23500)

Since that meeting, the notion of global security, and presumed threats to it, has undergone considerable expansion and diversification, both inside and outside the Council chamber. Inside the chamber, various issues, entities, and objects, from humanitarian crises to HIV/AIDS, have been invoked for possible determination of the existence of a “threat to the peace” (Wellens 2003; Paige 2019), which is “the charter’s key threshold for council action,” including collective military measures (Bosco 2009, 49). Outside of the chamber, novel notions of security such as “human security” have been proposed and elaborated upon, along with an ever-widening range of security threats, from poverty to cyberattacks (UNDP 1994; 2022). Moreover, in the broader academic field of international security studies (ISS), different scholars have made varying contributions to this trend of expansion (Buzan and Hansen 2009), including the “securitization” of climate change, terrorism, and other matters (Buzan and Wæver 2009; Maertens and Hardt 2021; Martini 2021).

While conceptions of security and threats have clearly multiplied during the past decades, these conceptions have not had an equally strong or lasting impact on policy. In particular, in policy deliberations at the UNSC, issues such as international terrorism have repeatedly been taken up by delegates representing various countries and organs, whereas other matters such as human security have received only scant references from a limited number of (often non-permanent) members. These and other variations leave a considerable number of ambiguities concerning the defining features of the global security discourse articulated within the post-Cold War UNSC. The present study addresses these ambiguities. Specifically, focusing on the notion of “threat to the peace,” one of the most consequential notions in the Council’s policymaking, the study asks: What has this notion meant—both over time and among members—in the past three decades of Council debates? In addressing this question, the study pays particular attention to varying semantic associations between the threat notion on the one hand and the numerous issues, objects and entities discussed by the Council’s members on the other.

In order to answer the above question, this study conducts a systematic investigation of the entire body of public deliberations in the post-Cold War UNSC. The quantitative text analysis methodologies used in the study show the highly selective but coherent nature of the global security discourse formed around the notion of “threat to the peace.” In particular, the study finds that, while the threat notion articulated at the UNSC has indeed been associated with an expanding roster of issues and entities, the notion has had rather stable elements at its core. These elements comprise terrorism and the proliferation of weapons of mass destruction (WMD), threat sources that were already identified in the January 1992 presidential statement quoted above (S/23500). Moreover, these issues, especially terrorism, almost exclusively constitute the limited common grounds for possible collective action by the Council’s five permanent members (P5), whose prospects for consensus have become increasingly shaky in the current situation of heightened geopolitical tension. Many other issues and entities have shown varying degrees of association with the threat notion; however, their relevance has depended mainly on the context (time periods, actors, and/or agenda items) in which they were raised and discussed. This stratified understanding of global security and threats raises serious doubts regarding other prevailing characterizations of the post-Cold War security discourse. It further cautions against the unconstrained use of some of the theoretical constructs widely employed in prior work focused on the UNSC, as well as in the wider ISS, most notably “securitization” and “interpretative community.”

In addition to presenting these findings, this study also makes two important contributions to the literature on Council politics and global security. First, it demonstrates a useful methodological framework that can help systematically uncover discursive dynamics from a large volume of speech texts produced by multiple international actors over many years. This framework benefits from the recent advancements in large-scale text analysis using machine learning (Grimmer and Stewart 2013; Evans and Aceves 2016). In particular, it employs “word embedding,” a technique that estimates numerical vector representations of words appearing in a given set of documents, or “corpus,” based on the relations of co-occurrences among these words. By applying this technique to a large corpus of Council deliberations, the study automatically extracts diverse concepts (e.g., “violence,” “Al-Qaida,” “climate change”) that are semantically associated with the threat notion (as represented by such terms as “threat” and “threats”), with varying intensity. Thus, the study reconstructs the actors’ understandings of the threat notion, as well as the evolution of these understandings over time.

Second, the study offers a newly created dataset on Council policy debates. The dataset comprises a large number of speech texts, each annotated with relevant pieces of information, such as the name and country of the speaker. These texts were extracted from the official meeting records (procès verbaux, or P.V.) of the UNSC, which contain the transcripts of all the public statements delivered by various delegates and officials at official open meetings.¹ By the end of 2021, the Council had convened 8,942 such meetings, starting from its first session on January 17, 1946. The dataset compiles a broad range of information, including meeting date, agenda, the president’s name and country, voting outcome, and output policy document (i.e., Council resolution and presidential statement) for all the meetings. Regarding speech texts, the dataset has so far covered more than 5,700 meetings held between January 1990 and December 2021, spanning the entire trajectory of post-Cold War Council politics. The extracted speech texts, nearly 90,000 in total, are available in a machine-readable, preprocessed format.

Threat Conceptions at the UNSC

Several groups of prior studies are pertinent to the question raised in the previous section. This section reviews two groups that are particularly relevant to the notion of security threat as articulated at the UNSC: one group from the literature on the UNSC itself, and the other from the wider ISS, especially studies that analyze the “securitization” of specific matters in the global security discourse. The section also points out various direct and indirect implications concerning the threat notion from these groups of studies for later evaluation.

UNSC and “Threat to the Peace”

Many textbooks on the UN have noted “the expansion of what the Council considers a threat to international peace and security” (Roberts and Zaum 2008, 56), especially after the end of the Cold War (Mingst et al. 2017; Weiss et al. 2017; Weiss and Wilkinson 2018). A broad array of issues, starting with civil wars and humanitarian crises but now including the illicit flow of small arms, drug trafficking, piracy, sexual violence, and climate change, have been raised and discussed as potential sources of “threat to the peace” (Wellence 2003; Paige 2019). The list of such emerging threats is so extensive that it has been described by some critics as a “theme park” of non-traditional security threats (von Einsiedel et al. 2016, 12). Because the Charter itself does not define what constitutes a “threat to the peace” and instead leaves the determination of the specifics of this notion entirely to the collective discretion of the Council members (Roberts and Zaum 2008, 14), many scholars, especially those from the field of international law, emphasize the “political,” even “arbitrary,” and “inconsistent” nature of the Council’s interpretation of the notion (Österdahl 1998, 103; Eckert 2001, 56; Wellence 2003, 21).

Beyond this general consensus on the seemingly unorganized widening of interpretation, few have addressed directly the threat notion unfolding at the UNSC (see below for an important exception), although several relevant strands of argumentation can be found. One of these strands concerns the diverging approaches of the P5 in Council politics, which includes the interpretation and articulation of the threat notion. Specifically, it is widely noted that China and Russia prefer a conservative and narrow interpretation of the UN Charter, showing general suspicion of, and often vehement opposition to, the presumed tendency of the other three Western members to readily involve the Council in “domestic affairs” such as humanitarian crises and human rights violations (Chan 2013; Wuthnow 2013; Fung 2019; Paige 2019; Schmitt 2019).

Conversely, in another strand of argumentation, certain scholars have noted the more cooperative aspects of Council politics. They highlight the integrative forces in the Council’s discursive dynamics, which might bring its members some semblance of community or even corporate identity (Hurd 2007). Johnstone’s conception of the UNSC and other international organs as “interpretative communities” (Johnstone 2003, 2011) is particularly influential in this regard (Chan 2013; Paige 2019). According to this conception, the UNSC largely functions as a deliberative setting where “states and, increasingly, non-state actors discuss, debate, and generate shared understandings about the terms of international life” (Johnstone 2011, 5). In this context, Johnstone (2011, 154) particularly notes the influence of the “humanitarian impulse” on Council deliberations, wherein the Council members have shown an increasing tendency to embrace norms such as the protection of civilians to address the physical harm to civilian populations posed by armed conflicts and other humanitarian crises. In a similar vein, Hanania (2021) observes the “Humanitarian Turn” in the post-Cold War UNSC in its growing involvement with protecting individuals.

Finally, influenced by Johnstone’s idea and using Council meeting records as primary materials, Paige (2019) conducts a detailed historical analysis of the P5’s justificatory discourse for or against the existence of a “threat to the peace.” The study examines more than 20 cases, from the Spanish question in 1946 to the more recent chemical weapons problem in Syria in the 2010s. This extensive analysis finds that, while the UNSC as a whole has “no discernible pattern or definition for what constitutes a threat to the peace,” each individual P5 member follows a distinct and consistent pattern regarding this issue (Paige 2019, 6). Somewhat unfortunately for the present study, however, these patterns largely concern the modalities of the P5’s justificatory arguments (e.g., emotive rhetoric versus formal legal reasoning) and not the content of the threats discussed by the members.

ISS and “Securitization”

Policy deliberations at the UNSC have also provided a nurturing ground for conceptual expansion and diversification outside the Council chamber. Mostly in parallel with the widening security discourse observed in the post-Cold War UNSC, security studies have also widened and deepened their perspectives (Buzan and Hansen 2009, chap. 7). Among these perspectives, the “securitization” theory espoused by the Copenhagen School, along with other critical and constructivist approaches, is particularly pertinent because of the strong emphasis on threat construction, discourse, and “speech act” (Buzan et al. 1998; Booth 2005; CASE Collective 2006).

Numerous scholars have used Council deliberations as evidence for securitization theory to describe and analyze a discursive process in which the issue of interest is presented as an existential threat and accepted as such. For example, some scholars see an extensive securitization of climate change, or even a “climatization” of security, in the global security discourse at the UNSC and other venues (Barnett and Adger 2007; McDonald 2013; Maertens 2021; Maertens and Hardt 2021). Others see a similar process occurring for public health issues (Rushton 2011; Burci 2014; Voss et al. 2022).

Global terrorism is another favored topic (Jackson 2005; Onuf 2009; Martini 2021). As illustrated by Buzan and Wæver (2009)’s use of “macrosecuritization” in this context, scholars typically stress the pervasive, even overarching nature of the threats posed by terrorism, especially after the 9/11 terrorist attacks on the United States. After closely following discursive dynamics at the UNSC during the past decades, e.g., Martini (2021) depicts a broadening process of securitization of international terrorism with a concomitant discursive construction of the identity of the international community (represented by the Council) as a whole.

Therefore, policy deliberations at the UNSC have inspired an impressive body of securitization arguments. Unsurprisingly, however, given the theoretical focus of the underlying framework, almost all of these arguments are issue-specific (see Buzan and Wæver 2009 for related discussion). In other words, these potentially competing claims of securitization have not yet been fully compared and adjudicated across different issue areas.

Implications Regarding Threat Conceptions

In sum, the preceding studies have produced observable implications regarding several aspects of the threat notion as articulated at the post-Cold War UNSC. To begin, the following constitutes a sort of the minimum consensus in the existing scholarship on both the UNSC and international security:

Implication 1An ever-expanding set of issues, entities and objects has been discussed in association with the threat notion over the years.

Beyond this, many researchers studying the UN, especially international law scholars, do not see any further pattern or trend regarding how the notion has been conceived (Österdahl 1998; Eckert 2001; Wellence 2003; Paige 2019). Accordingly, they would agree the following:

Implication 2No subset of issues, entities and objects has been consistently associated with the threat notion.

In contrast, other scholars observe discernible trends in Council deliberations. Those who emphasize the increasingly humanitarian nature of the post-Cold War security discourse at the UNSC (Johnstone 2003, 2011; Binder 2017; Weiss et al. 2017; Hanania 2021), e.g., would likely support the following:

Implication 3Issues, entities and objects that are distinctively related to humanitarian crises and civilian protection have been consistently associated with the threat notion.

Likewise, proponents of the securitization theory claim the dominant nature of the threat posed by a particular problem or entity. In the context of the post-Cold War UNSC, the following two propositions reflects such a perspective (Buzan and Wæver 2009; Maertens 2021; Maertens and Hardt 2021; Martini 2021):

Implication 4Issues, entities and objects that are distinctively related to global terrorism have been consistently associated with the threat notion.

Implication 5Issues, entities and objects that are distinctively related to climate change have been consistently associated with the threat notion.

At the same time, these scholars also presume the concomitant identity formation induced by the existential threat thus posed. Similarly, other scholars consider the UNSC to be an “interpretative community” (Johnstone 2003, 2011), a legitimate institution with a “corporate identity” (Hurd 2007), and “a diffusor of norms” (Hanania 2021). The strong collectivity connoted by these descriptions would not be conceivable without a considerable degree of conceptual agreement on such essential notions as “threat to the peace” among different Council members, especially the P5:

Implication 6Issues, entities and objects that have been associated with the threat notion are broadly shared among different Council members.

Conversely, numerous scholars who emphasize the highly divisive nature of Council politics (Frederking 2007; Malone 2006; Thakur 2017) would likely embrace the following:

Implication 7Issues, entities and objects that have been associated with the threat notion are shared among different Council members only to a limited extent.

Research Design

Toward a Systematic Investigation of Security Discourse

The implications featured above, which contain mutual contradictions and ambiguities, suggest the need for a much broader, more systematic analysis of the discursive dynamics that the UNSC has generated over the past decades. There have been notable developments that facilitate such an analysis. One is increased data availability. Employing a variety of quantitative and qualitative data, scholars have already subjected various aspects (e.g., resolutions and their adoption, agenda-setting) of Council politics to systematic investigations (Wallensteen and Johansson 2015; Binder 2017; Allen and Yuen, 2020, 2022; Binder and Golub 2020). More recently, a huge body of textual documents made publicly available online by the UN² has offered opportunities for wider and deeper inquiries, especially those regarding discursive dynamics (Baturo et al. 2017; Howard and Stark 2018; Schönfeld et al. 2019; Kentikelenis and Voeten 2020; Finke 2021; Mahmood et al. 2022).

In the present context of analyzing security discourse at the UNSC, an increasing number of scholars have extensively employed these documents, including Council resolutions and presidential statements, as well as speech transcripts (P.V.) (Medzihorsky et al. 2017; Kohlenberg et al. 2019; Hanania 2021; Martini 2021). In particular, “The UN Security Council Debates” dataset provided by Schönfeld et al. (2019)³ has spurred a variety of studies illuminating different facets of Council policy deliberations, such as the role of UN officials (Eckhard et al. 2023), engagement with weak states (Campbell and Matanock 2021), the P5’s posture on the Women, Peace, and Security (WPS) agenda (Glaser et al. 2022), rhetoric and the authorization of force (Scherzinger 2022), the responsibility to protect (R2P) (Scherzinger 2023), and the securitization of public health (Voss et al. 2022). Unfortunately, an investigation of the threat notion has remained absent from this expanding body of quantitative work on the UNSC. More importantly, the original dataset developed by Schönfeld and his colleagues covers the period of 1995–2020 at the time of writing, leaving the early post-Cold War years (1990–1994) out of its purview. Given the obvious importance of these years to the evolving security discourse at the UNSC (e.g., the decisive collective action in the Gulf War, the shocking genocide in Rwanda, and the peacekeeping debacle in Somalia all occurred during these missing years), this study uses an alternative, originally-developed dataset on Council deliberations, with temporal coverage that encompasses the entire post-Cold War period.⁴

Another recent development, the increasing methodological sophistication in large-scale text analysis, has further encouraged the quantitative treatment of these and other textual documents in the wider field of political science (King and Lowe 2003; Grimmer and Stewart 2013; Lucas et al. 2015; Wilkerson and Casas 2017). The existing studies on the UNSC cited above also deploy a broad array of quantitative tools, from simple word counting (Campbell and Matanock 2021; Martini 2021; Voss et al. 2022) to dictionary-based sentiment analysis (Scherzinger 2023) to more elaborate statistical modeling such as topic models (Hanania 2021; Eckhard et al. 2023) and spatial models (Medzihorsky et al. 2017). Again, this study has chosen a different approach, namely, word embedding.

As the next subsection details, word embedding is a powerful tool for uncovering the detailed semantics of a particular notion (e.g., “threat”) in a given discursive context as well as its variation across multiple contexts (Garg et al. 2018; Kozlowski et al. 2019). It has recently been making significant inroads into political science and international relations (Pomeroy et al. 2019; Rice et al. 2019; Rheault and Cochrane 2020; Rodman 2020; Rodriguez and Spirling 2021). For example, in a highly relevant study, Rathbun and Pomeroy (2022) apply word embedding to a corpus of 8,640 public speeches delivered at the annual United Nations General Debate (Baturo et al. 2017), and reveal what they call the “moralization” of harm and threat. The present study employs a similar methodology for a slightly different purpose (representing the meaning of the threat notion itself, not its connection with morality) and in a considerably different context (the Security Council rather than the General Assembly).

Representing Threat Conceptions with Word Embeddings

Word embedding refers to a class of unsupervised machine-learning models that learn high-dimensional vector representations of words for a given corpus (Mikolov et al. 2013; Pennington et al. 2014). Each word is “embedded” or assigned a series of numbers denoting its coordinate in vector space, so that two words that tend to co-occur with a similar set of words in the text are likely to be closer together.

In the present context, word embedding is used as a simple device for the automated retrieval of relevant concepts that shared, at one time or another, discursive contexts with the notion of “threat to peace” during the past three decades of Council deliberations. More specifically, the investigation to be conducted first quantifies the strength of the semantic association between the token “threat” and each of the remaining tokens (words and phrases) with a measure of similarity between the corresponding vector representations (“embedding vectors,” or simply “embeddings”). It then derives those tokens, often called “nearest neighbors,” that have the highest similarities to “threat.” Lastly, it examines how the composition of the nearest neighbors of “threat” varies across different periods and across different countries by way of simple set operations. Although these steps are a fairly rudimentary, straightforward application of word embeddings, similar usage in other fields has produced insightful findings. For instance, Rodriguez and Spirling (2021) illuminate nuanced partisan differences in the representation of contentious notions (e.g., “abortion,” “tax,” and “justice”) by legislators in the context of legislative debates.

Nevertheless, some caution is required regarding the exact meaning of “semantic association.” For example, when applied to the entire corpus of Council speeches, word embedding typically produces nearest neighbors for the term “threat” that include: “threats,” “danger,” “serious_threat,” “international_peace,” ”terrorism,” ”challenge,” ”terrorist,” ”security,” ”challenges,” “risks,” and so on (see table 1 below). It is immediately evident that the associations of these phrases with “threat” are far more diverse than expressions such as “similar words” might indicate. Some of these phrases (e.g., “threats,” “danger,” and “serious_threat”) are indeed synonymous with, or semantically highly similar to, the term “threat.” Others (“terrorism,” “terrorist”), however, are rather distinct entities or issues that are threatening. Still others (“international_peace,” “security”) are the entities or objects that are being threatened. In other words, word embedding does not directly uncover some innate meaning of a given notion. It simply provides a certain constellation of concepts that are variously associated with that notion. The specific conception of the notion is revealed only after adequate reading and reasoning concerning the related concepts (see Kozlowski et al. 2019, 914). This is unsurprising, given that word embedding is unsupervised machine learning, which requires post-hoc human interpretation and validation.

Notwithstanding these caveats, the above procedure based on word embedding can be expected to give a manageable description of the evolving security notion centered around “threat to peace” in its entirety, across many agendas and over many years. It can also uncover conceptual overlaps and differences among threat conceptions that have been articulated by different actors and espoused in different time periods.

Moreover, in comparison with widely-used word-frequency analysis (Campbell and Matanock 2021; Martini 2021; Voss et al. 2022), or its more elaborate variants such as “keyword-in-context” counting as performed by Scherzinger (2023),⁵ word embedding can yield more reliable findings that depend less on a scholar’s particular choice of words. That is, it estimates a vector representation of each word considering all the possible associations, direct as well as indirect, among all the words found in the same corpus, not just a limited number of chosen words. Besides ensuring the robustness of derived insights against word choice, this helps capture more nuanced aspects of the threat notion than simple word-counting permits. First, the proposed procedure can detect concepts that are closely associated with the threat notion, even if these concepts are not frequently mentioned along with the term “threat.” In the later analysis, phrases such as “violent_extremism” and “transnational_organized_crime” represent these concepts. These phrases were found to have a strong semantic association with the threat notion because other concepts closely related to them (e.g., “terrorism”) frequently co-occur with the term “threat.” Second, the procedure can also isolate and exclude concepts that are not distinctively associated with the threat notion, even if these concepts are frequently mentioned along with the term “threat.” In the following analysis, some tokens that frequently co-occur with “threat,” including “civilians” and “sanctions,” were found to have only a weak association with that notion. This is because these concepts are frequently discussed in many other discursive contexts in the UNSC, not just in those concerning the security threat. In sum, word embedding is highly instrumental in extracting threat conceptions from a large number of speech texts in a sufficiently informative, but adequately discriminating manner.

Data and Methods

Text Data and Their Acquisition

English transcripts of Council meetings (S/PV) are the primary data source for the investigation. These data were obtained from the Official Document System of the UN via hyperlinks embedded in the summary tables provided on the website of the Dag Hammarskjöld Library.⁶ Although all the available transcripts, from the first meeting in January 1946 to the most recent one in December 2021 (as of the end of 2021), were obtained this way, the analyses were limited only to meetings held during the 30-year period from 1990 to 2019.⁷ That is, this study covers the official Council meetings from the 2,904th (held on January 11, 1990, the first meeting of the year) to the 8,697th meeting (held on December 20, 2019, the last meeting of the year). Among these 5,794 meetings, 253 were closed to the public or held without any substantive deliberations, thus leaving only brief communiqués. These unrecorded meetings were excluded from the analyses.

The proceedings of each meeting are recorded in a single document file. Each file contains the transcripts of speeches delivered by speakers during the corresponding meeting, the names of these speakers, and the countries or agencies they represented, along with other information, such as the agenda and date of the meeting. Although these files themselves do not constitute a structured database, several formatting conventions of Council meeting records enable one to locate and extract relevant pieces of information while discarding irrelevant ones (e.g., page numbers, document symbols, unnecessary punctuation, etc.). Online Appendix A details the text extraction process.

One potentially confounding aspect of Council deliberations for their systematic treatment is that a sizable portion of recorded speeches are procedural statements given by the Council’s monthly rotated presidents. These statements, for instance, include opening and closing remarks, vote announcements, and ceremonial greetings. Meanwhile, a president can also give a non-procedural policy statement as a representative of his or her country. There are several patterned expressions for the latter type of statements (see Online Appendix A), which were exploited to screen out non-policy procedural statements. With this additional procedure, a total of 50,565 (supposedly substantive) speeches, comprising 41,736,274 words, were obtained (85,626 speeches comprising 44,707,577 words, including the discarded procedural statements).

Online Appendix B provides several illustrations concerning how the extracted speeches are distributed across different actors and years. They confirm the general tendencies of Council meetings already indicated in a prior work that used a similar dataset (Schönfeld et al. 2019). One such tendency is the steadily expanding volume of Council debates, with an increasing number of speakers appearing in the Council chamber over years. This tendency had actually accelerated throughout most of the 2010s (then was suddenly disrupted by the COVID-19 pandemic). Another notable tendency is the sizable, but not necessarily overwhelming, volume of speeches contributed by P5 delegates over the past three decades. In fact, the combined number of P5 speeches is 10,525, barely over 20% of the total. The most vocal countries in terms of the number of speeches are the Russian Federation (2,231), the United States (2,139), the United Kingdom (2,093), France (2,071), China (1,991), Japan (958), Germany (763), Brazil (679), South Africa (676), and Indonesia (629).⁸

Text Preprocessing

A series of preprocessing measures were applied to the obtained text data. These measures include removing punctuation and other non-alphanumeric symbols, converting uppercase characters to lowercase ones, and converting the spelling of words from British to American English. Besides these fairly common and simple procedures, a more elaborate set of techniques for natural language processing (NLP) was used to make the later analysis context-sensitive. The most important technique was named entity recognition (NER).⁹

As already noted, by utilizing word embeddings, the analyses detailed below aim to extract from the acquired speech texts concepts variously associated with the threat notion. One problem that arises is that a single word does not necessarily correspond to a single concept. In the context of Council deliberations, the expression “Democratic Republic of Congo,” for example, constitutes one concept (the country referred to by this expression), and treating these four words separately (“Democratic,” “Republic,” “of,” and “Congo”) is semantically meaningless when they are collocated in this order in the same text. NER helps address this gap between a word and a concept in a non-arbitrary, reproducible manner (i.e., independent from particular concept-phrase mappings chosen by a particular scholar). NER is an NLP task that detects, extracts, and classifies named entities (e.g., “Yemen,” “Gulf of Guinea,” and “Kofi Annan”) embedded in unstructured text. This study used an open-source, highly advanced NLP engine (spaCy)¹⁰ to perform NER on the Council speech texts. With this procedure, the named entities mentioned in these texts were systematically extracted, and multiple words (“democratic,” “republic,” “of,” and “congo”) that were judged to constitute a single named entity were combined into a single phrase (“democratic_republic_of congo”).

However, not all concepts are represented by named entities. A concept can be denoted by an unnamed but patterned expression that consists of more than one word. In the context of the UNSC policy deliberations, examples include “human rights,” “sexual violence,” “transnational organized crime,” and so on. In order to detect such patterned expressions, another preprocessing procedure was applied to the speech texts as a complement to NER. This procedure builds on two statistical techniques for NLP. One is part-of-speech (POS) tagging, which employs a statistical model and a dictionary to estimate a POS (e.g., “noun,” “verb,” and “adjective”) for each word appearing in a given text.¹¹ The other is phrase detection. Using a statistical algorithm, phrase detection connects frequently collocated words into a single phrase.¹² Iterative application of phrase detection to the Council speeches, while using POS tagging as a screening device for selecting phrases that function as nouns, indeed generated a large number of familiar expressions, including “human_rights,” “sexual_violence,” and “transnational_organized_crime.”

These procedures produced a total of 139,813 named and unnamed expressions (including single noun words), each of which is presumed to denote some meaningful concept. The most frequent expressions are “council” (155,503 times), “security_council” (130,647), “united_nations” (129,400), “security” (94,782), “general” (93,458), “peace” (93,150), “efforts” (91,159), “support” (90,747), “people” (80,587), and “secretary” (79,876).

The preprocessed speech texts comprise a total of 34,330,270 tokens (words and phrases), with 170,977 unique tokens. The entire set of these speeches, or the corpus, is available from the ISQ Dataverse, which contains raw and preprocessed versions of all the extracted speech texts. Each speech text is time-stamped, annotated with the corresponding speaker's name, country/agency, and status (monthly President or not, UN Secretary-General or not, etc.), and coded as a procedural or policy statement (see above). The dataset also includes a list of 5,729 relatively frequent conceptual expressions that were selected from approximately 140,000 tokens obtained in the way described above. These expressions, which have appeared 100 times or more during the three decades of Council deliberations, were extensively utilized in the subsequent analyses.

Analytical Procedure

There are several methods for performing word embedding, including the “word2vec” algorithms (Mikolov et al. 2013). This study employed GloVe (Global Vectors for Word Representation), an embedding algorithm proven to effectively perform various linguistic tasks (e.g., word analogy) on a wide variety of corpora of different sizes (Pennington et al. 2014; Garg et al. 2018; Kozlowski et al. 2019). GloVe efficiently computes a word-word co-occurrence matrix from an input corpus and then applies a log-bilinear regression model to this matrix to estimate the numerical elements of a vector denoting each word in the corpus.¹³

GloVe generated vector representations of all the tokens (words and phrases) contained in the Council speech corpus. The strength of the semantic association between two tokens was then computed using cosine similarity, a widely-used similarity measure. To illustrate, suppose that C is the Council speech corpus or some subset of it. Let i and j denote two tokens appearing in C, and w_i and w_j represent their corresponding embedding vectors. Then, the cosine similarity between i and j for C is given as follows:

The product in the numerator is an inner or dot product of two vectors. The measure takes a value of -1.0 (diametrically opposite) to 1.0 (identical).

In the analyses reported below, for a given set of speeches, cosine similarity was computed between the token “threat” and each of the conceptual expressions (noun words and phrases) that were derived by the preprocessing procedure described above.

Because word embedding does not generally work well for low-frequency words, the computation was limited to those expressions, 5,729 in total, that appear in the entire corpus 100 times or more. Although the following results are based on the similarities calculated between these expressions on the one hand and a single word, “threat,” on the other, Online Appendix C shows illustrations obtained from different specifications, wherein other relevant tokens such as “threats” (plural) and “international_peace” were also incorporated into the computation. These specifications gave similar results to those presented below.

In this study, the “nearest neighbors” of the token “threat” are those words and phrases whose cosine similarities with that token exceed some threshold value. In order to ensure comparability among different sets of speech texts, all the similarity values in each set were first converted to their standard scores (z-scores). Nearest neighbors were then chosen using a common cutoff value such as +2.0 (i.e., two standard deviations above the mean). For the sake of readability, the presentation below (except tables 1 and 2) is mostly limited to the nearest neighbors obtained with a +2.5 cutoff. However, the findings for other cutoffs are reported in Online Appendices D and E. Moreover, because the estimation of embedding vectors can be perturbed depending on the random seeds used for initialization, the estimation was iterated 10 times for each setting. The nearest neighbors reported in this study are “robust” since these tokens were observed in the semantic neighborhood of “threat” in at least 9 of the iterations for a given setting.

Word embeddings are relative to the corpus concerned and should thus depend on the context (time, author, agenda, etc.) of each corpus. Therefore, this study iterated the procedure described so far for different subsets of the speech corpus. Specifically, the corpus was divided according to country (e.g., a set of speeches delivered by US delegates), time period (e.g., speeches delivered during the 2010s), or both (e.g., speeches delivered by US delegates during the 2010s). In order to examine how the threat notion at the UNSC could vary with topics under discussion, the study also partitioned the corpus according to agenda. However, excessive division would result in too small subsets, which would render reliable statistical estimation untenable. Therefore, the partition in this respect was based on the official agenda categories (e.g., “Africa,” “Middle East,” and “Thematic Issues”), rather than much finer agenda items (e.g., “Reports of the Secretary-General on the Sudan and South Sudan,” “The situation in the Middle East, including the Palestinian question,” and “Protection of civilians in armed conflict”).¹⁴

Finally, GloVe works with several learning parameters. Among them, the number of dimensions of the embedding space was set to be 300. The size of the “context window,” which is another important parameter, was set to be 12 in both reading directions. In other words, if a word is located up to 12 words ahead or behind another word, then these two words are considered to have co-occurred. These parameter values were assigned following the recommendations of prior work (Rodriguez and Spirling 2021). Online Appendix C reports the results obtained for different learning settings, which show no major difference from the results reported in the next section.

Results and Interpretations

Temporal Consistency and Change

Table 1 lists the conceptual expressions (out of 5,729 relatively frequent ones) that are at least to a certain degree associated with the token “threat” for three succeeding periods in the post-Cold War era, namely, the 1990s, the 2000s, and the 2010s. Depending on their standardized similarities (denoted as z) with “threat”, these expressions were categorized as either “moderate association” (2.0 ≤ z < 2.5), “strong association” (2.5 ≤ z < 3.0), or “highly strong association” (z ≥ 3.0). Besides the minimum frequency requirement (i.e., 100 times or more in the entire corpus) and the POS-tag screening (i.e., only noun words and phrases), the listed expressions were essentially machine-retrieved from a large number of speeches delivered during the corresponding time period.¹⁵

Although temporally bounded, each subset of the corpus of policy statements delivered by the speakers represents the whole range of countries and organs, and includes all the agenda items under the Council’s consideration during the period concerned. Therefore, it is not surprising that the underlying associations shown in the table are considerably complicated and defy any straightforward interpretation. Nevertheless, several distinct features are notable. First, as evident especially from the third column (“Moderate association”) of table 1, the term “threat” has indeed shared discursive contexts with a steadily increasing, but constantly changing roster of named and unnamed concepts. Examples include “landmines,” “balkans” (1990s), “aids” (that is, HIV/AIDS), “sexual_violence” (2000s), “piracy,” “sahel” (2010s), and so on. The term “climate_change” also appears in the second and the third rows, demonstrating an increase in its association with the threat notion over the corresponding periods (from the third column to the second).

Second, conversely, some concepts show almost invariant association with “threat.” Looking at the first column (“Highly strong association”), these include “danger,” “serious_threat,” and “challenge,” whose semantic connection to the threat notion should be evident almost in any context. Furthermore, they also include “international_peace” and “security.” This might also be unsurprising as it indicates that the actual usage of the term “threat” has consistently followed the Charter’s expectation that “threat to the peace” is something posed to “international peace and security” (although “regional_peace” is also noticeable in the table).

Third, several groups of concepts have stably, if not invariantly, strong associations with the threat notion. One such group, including “weapon,” “mass_destruction” (together constituting “weapon of mass destruction” in the original text), and “proliferation,” evidently represents one of the Council’s persistent concerns, namely, the proliferation of WMD. Another group, which consists of “terrorism,” “terrorist,” “international_terrorism,” and many other related concepts, plainly points to global terrorism as a potential source of the threat. As the table depicts, both of these groups have had consistent presence in Council deliberations regarding the threat notion during all the three periods, including the 1990s. However, their associations with the notion have evidently strengthened from the 2000s onward, constituting an unambiguous trend in otherwise highly complicated dynamics. This has particularly been the case for terrorism, wherein an ever-expanding array of entities and notions such as “al-qaida,” “isil,” “violent_extremism,” and “foreign_terrorist_fighter” have been conceptually mobilized to highlight its increasingly threatening nature.

Figure 1, which was derived from table 1, illustrates these observations. The figure is a Venn diagram that depicts all possible logical relationships among three sets, represented by the three overlapping closed circles. These sets correspond to the nearest neighbors of “threat” for the three periods, that is, the 1990s, the 2000s, and the 2010s. For each of these neighbors, z ≥ 2.5 (either “highly strong” or “strong” association in table 1).

The three-digit binary notations such as “100” and “110” indicate whether a token in the corresponding region is a member (1), or not (0), of the set denoted by each digit. Accordingly, the concepts appearing in the “100” region, e.g., “middle_east,” “civilian_population,” and “kosovo,” had strong semantic associations with the threat notion in the 1990s, but have not so thereafter, whereas “crisis,” “regional_peace,” and “aggression,” all found in the “110” region, were strongly associated with the notion both in the 1990s and the 2000s, but not in the 2010s, and so on. Among others, the diagram confirms that, along with almost self-evidently related concepts such as “danger” and “serious_threat,” several more non-trivial ones, including “terrorism,” “weapon,” and “mass_destruction,” constitute the most temporally stable component (“111”) of the post-Cold War security discourse having unfolded at the UNSC.

Convergence and Divergence among the P5

Focusing on cross-national commonalities and differences of the threat conceptions among the Council’s most important members, the P5, table 2 shows the conceptual expressions that each permanent member has had associated with the threat notion over the entire study period. In the same way as the previous table, this table also categorizes these expressions into the three tiers of semantic neighborhood in accordance with their standardized similarities with the threat notion. Temporally more disaggregate versions, i.e., similar tables derived separately for each of the three decades, are provided in Tables D1–3 in Online Appendix D.

Conceptual relationships induced by 5, rather than 3, distinct sets often become too complex to concisely describe. Figure 2, by way of two Venn diagrams, highlights some aspects of these relationships. These three-set Venn diagrams describe convergence and divergence of the threat conceptions of two groups of permanent members, namely, the UK, the United States, and France on one hand, and the United States, Russia, and China on the other. Each member’s threat conception is again represented by the respective nearest neighbors obtained with z ≥ 2.5, which are listed in the first and second columns of table 2.

Several observations can be drawn from these diagrams. First, some of the concepts that constitute the temporally stable component of the threat notion mentioned above (see figure 1) are also widely, but somewhat thinly, shared among the P5 members. As can be seen from the “111” regions in the respective diagrams, these concepts comprise “international_peace,” “terrorism,” “challenge,” “terrorist,” and “security.” Although figure 2 shows the threat conceptions derived from the temporally aggregated sets of speeches, a similar set of tokens almost constantly appear among the nearest neighbors of the P5’s threat notions obtained on a much finer temporal scale (see Online Appendix Tables D1–3). It can thus be argued that these tokens represent a common, if highly abstract, conceptual basis for the P5’s shared understanding of the threat notion.

Second, however, the conceptual overlap becomes considerably heterogeneous beyond the limited common ground suggested above. For example, figure 2 reveals that the three Western members (the UK, the United States, and France) share some concerns, for instance, those about the proliferation (denoted by “proliferation,” “weapon,” and “mass_destruction”), more emphatically than the remaining two. On the other hand, it is equally obvious that the threat conceptions of the latter members, China and Russia, do not necessarily show a comparable degree of conceptual convergence. To the contrary, in this respect, the United States and Russia apparently share a much broader, more specific array of concepts such as “isil,” “al-qaida,” “drug_trafficking,” and “syria” in their respective articulation of “threat to the peace.”

Third, the P5’s threat conceptions largely diverge further beyond. Each tends to emphasize different geographical locations (e.g., “sahel” for France), subject matters (e.g., “drug” for Russia), and organizational entities (“lra,” namely, the Lord Resistance Army (LRA), for the UK), even when discussing a common, generally agreeable, subject such as global terrorism. While any number of characterizations can probably be made about each member’s threat conception, one noticeable instance regarding China illustrates the point. Chinese delegates have used the term “threat” in much the same way as their counterparts in many contexts (e.g., stressing the threat posed by al-Shabaab to international and regional peace). In other contexts, however, their usage considerably diverged, and it often had strongly negative connotations as well. In table 2, these connotations are conveyed by distinctive words such as “pretext” and “excuse.” For many Council watchers, these words have been employed in a largely predictable manner, e.g., when China opposed to “interfering in a country’s internal affairs under the pretext of alleviating humanitarian crises or the frequent wielding of military threats or intervention by invoking Chapter VII of the Charter” (S/PV.3954 (December 16, 1998)).

Threat Conceptions in Different Agenda Groups

The three Venn diagrams shown in figure 3 depict the evolving dynamics of the threat notion that were derived separately for three major agenda categories, namely, “Africa,”¹⁶ “Middle East,”¹⁷ and “Thematic Issues.”¹⁸ The overlapping circles in the diagrams denote the nearest neighbors obtained using the 2.5 z-score cutoff. Further information is provided in Tables E1–3 in Online Appendix E. In comparison with the more aggregated temporal evolution displayed in figure 1, the steady trend of conceptual expansion of the threat notion is even more pronounced for these three categories. This is particularly the case for thematic debates, a relatively recent convention that was introduced to the Council in the late 1990s and has since rapidly broadened its purview.

There are also considerable, but largely predictable, differences among these trajectories. For example, some named entities (e.g., “boko_haram” for Africa, “al-nusra_front” for the Middle East) and unnamed subjects/objects (e.g., “piracy” for Africa, “rocket,” indicating rocket attacks launched from the Gaza Strip, for the Middle East) have a distinctively regional nature. In particular, the conspicuous constellation of tokens such as “civilians,” “attack,” “impunity,” and “rome_statute” found in the African case (see also Online Appendix Table E2) confirms that Council deliberations concerning African agendas have a markedly stronger tendency to associate the humanitarian consequences of conflicts and crises with the threat notion. Conversely, some concepts, notably “climate_change,” become more pronounced when the analysis is exclusively focused on thematic debates.

However, behind these apparent differences, more or less parallel dynamics can be recognized. That is, the conceptual expansion noted for all three agenda groups has more or less been driven (with some time lags) by the unambiguous trend of terrorism to dominate the security discourse at the UNSC. Specifically, regardless of agenda categories, i.e., not just for the thematic category that comprises the agenda items exclusively devoted to terrorism (e.g., “Threats to international peace and security caused by terrorist acts”), a sizable portion of the nearest neighbors retrieved from the most recent speech texts (“011” and “001” regions in figure 3) are the concepts (e.g., “violent_extremism”) or entities (e.g., “al-shabaab”) frequently mentioned in this context. As in the preceding analysis, this common thread also seems to provide a minimum conceptual basis for the P5’s common understanding of the threat across different agendas (see Figures E1–3 in Appendix E), although reliable cross-national comparison is difficult here due to the far less availability of relevant speeches.

Discussion and Conclusions

The interpretations given in the previous section, which were derived carefully with a manual-reading of related speech texts, offer useful insights to both the literature on Council politics and that on global security. As a preface, it should be stressed that the findings above are, for the most part, consistent with existing observations about the Council. For example, as already mentioned, the widely noted ambivalence of China toward the Council’s role in general, and its widened interpretation of the “threat to the peace” in particular (Chan 2013; Wuthnow 2013; Fung 2019; Paige 2019), is at least partially revealed by the obtained constellation of threat-associated concepts (table 2). Similarly, it is not particularly surprising to learn that drug- and Afghan-related notions (e.g., “drug,” “drug_trafficking,” “afghan,” and “afghanistan”) are among the most salient aspects of Russia’s threat conceptions, considering that the trafficking of heroin and opium produced in its southern neighbor, Afghanistan, is one of the security concerns persistently raised by Russia in the Council chamber over the years (Trenin 2016). Other known geographic and thematic focuses of Council members are also evident from retrieved tokens and their relatively strong associations with the term “threat,” e.g., “climate_change” for the United Kingdom and France, “sahel” for France, and “north_korea,” “iran” for the United States (von Frederking 2007; Wuthnow 2013; Einsiedel et al. 2016; Paige 2019; Maertens and Hardt 2021). These correspondences, to some extent, credit the validity of the foregoing analyses.

With this validity, it is now possible to credibly assess competing views—summarized as Implications 1–7 above—of the threat notion articulated at the post-Cold War UNSC, and to present a much clearer picture of its conception over time and across actors, thus answering the question raised at the beginning. First, largely consistent with Implication 1, the key notion of “threat to the peace” has evidently become enriched. The foregoing analyses confirm that an increasing number of issues, entities and objects have indeed become associated with the threat notion across different agendas over the past decades. On the one hand, this expansive trend partially reflects an increasing volume of Council deliberations themselves (see Online Appendix B), wherein an expanding array of speakers have appeared in the Council chamber, raising and discussing a broadening range of issues. On the other hand, the threat notion articulated there has not been equally open to all these issues. Some screening and selection have certainly been occurring. Accordingly, a highly structured, even stratified, collective understanding of the security threat has been uncovered (table 1 and figure 1). Such an understanding clearly contradicts Implication 2, which emphasizes the inconsistent nature of the post-Cold War threat conception.

Given this understanding, several arguments highlighting specific aspects of the post-Cold War security discourse also lose their forces. For instance, climate change, which has often been treated as a manifest example of securitization at the UNSC (Rasmussen and Beck 2012; Maertens 2021; Maertens and Hardt 2021), has been an issue actively debated in the Council chamber since the United Kingdom first introduced it as an independent topic for thematic debates in 2007. This is reflected in the increasing salience of the term “climate_change” in the entire constellation of threat conceptions (figures 1 and 3). However, such a trend falls far short of what some call the “climatization of security” at the UNSC (Maertens 2021; Maertens and Hardt 2021). Besides occasional discussion in other contexts (e.g., the 2016 debates on the Sahel region; see S/PV7699 (May 26, 2016)), climate change remains a topic for specific occasions (thematic debates on this topic) and for specific proponents (the United Kingdom and France among the P5). Accordingly, Implication 5 receives only tenuous support from the foregoing analyses.

Similarly, the post-Cold War “Humanitarian Turn” in the UNSC (Hanania 2021) does not seem to translate into any comparable transformation of the threat notion, which is inconsistent with Implication 3. Humanitarian words indeed abound in Council deliberations. For example, the cumulative counts of “women” (58,537), “children” (41,042), “violence” (36,459), “civilians” (34,874), and “protection” (34,874), words that have often been mentioned in this context, all surpass the count of the term “terrorism” (31,950) in the study period. However, except for some notable circumstances, especially those concerning African issues (figure 3 and Online Appendix Table E2), the revealed associations of these “humanitarian” concepts with the threat notion are far weaker than the terrorism-threat connection, even though, as explained, some of these concepts frequently co-occur with the term “threat.”

Conversely, global terrorism and, to a lesser extent, the proliferation of WMD are far more strongly and distinctively associated with the threat notion. The reported results clearly show that these “A-list threats” (Carter 2001, 5–6) more or less underlie the three decades of security discourse regarding “threat to the peace” at the UNSC and have largely driven its evolution, particularly since the 2000s (figure 1). These findings unambiguously support Implication 4, which can be directly derived from several existing accounts (Buzan and Wæver 2009; Martini 2021). In particular, the findings closely correspond to Martini (2021)’s extensive account of the expanding domain of global terrorism in the Council security discourse. Unlike Martini, who focused largely on the thematic debates on terrorism (“Threats to international peace and security caused by terrorist acts”), the present study covers a whole range of agenda items and has found that a similar process of expansion has also been observed in other, regional agenda groupings such as the Middle East and Africa (figure 3). Global terrorism might be far more pervasive in the security discourse at the UNSC than her analysis suggested.

Another difference from the studies of Martini and others is more critical. These scholars, under the influence of securitization and other critical theories, often situate security discourse, especially its construction and representation of the threat (presumed as the Other), in the wider process of identity construction of “international community” (presumed as the Self), stressing their mutual constitution. However, the foregoing analyses produced few indications that such a process has been occurring. With regard to the core members of the supposed “international community,” the P5, their threat conceptions do converge, particularly around terrorism, but the convergence still remains largely abstract and patchy at best (figure 2). This finding privileges Implication 7 over Implication 6, if not decisively. In any case, as far as security discourse is concerned, there is a certain limitation to viewing active policy deliberations as producing a strong communal or corporate identity (Hurd 2007), including the pervasive view of the Council as an “interpretative community” (Johnstone 2003, 1; Paige 2019). In the present context, such a community, if any, still lacks a solid discursive basis.¹⁹

Footnotes

Author Biography

Takuto Sakamoto is a professor of Human Security at the Graduate School of Arts and Sciences, the University of Tokyo. For more on his work, see researchmap.jp/takuto_sakamoto.

Notes

Author's note: The author highly appreciates the tireless research assistance provided by Tomoyuki Matsuoka over the past couple of years. He also acknowledges the valuable contributions of Mihoko Iijima, Liu Jinwu, Yuri Goto, Nozomu Takaguchi, and Li Qingqian to this study. The author would further like to thank the organizers and attendants of the following events for the valuable occasions they offered him for presenting the study and obtaining feedback: the 2nd Annual Conference on Politics and Computational Social Science at Georgetown University; the 2019 Annual Convention of the Japanese Association of International Relations in Omiya; European Research Center for Political Culture (ERCAM) Workshop at the University of Bucharest; Social Fabrics Research Lab (FABLAB) Seminar on Big Data Analysis in International Relations at the West University of Timişoara; and the 2020 Annual Meeting of the Academic Council on the United Nations System. Finally, the author acknowledges the financial support (Grant Numbers: 18H03621; 22H00815) provided by the Japan Society for the Promotion of Science (JSPS). Data statement: The data underlying this article are available on the ISQ Dataverse, at https://dataverse.harvard.edu/dataverse/isq.

References