Diversity or decolonization? Searching for the tools to dismantle the ‘master’s house’


Within the literature on decolonizing the curriculum, a clear distinction is frequently made between diversity and decolonization. While decolonization entails dismantling colonial forms of knowledge, including practices that racialize and categorize, diversity is a policy discourse that advocates for adding different sorts of people to reading lists and the staff and student body. As a team of staff and students, we are committed to decolonization, but we are also aware that within our discipline of political science, calls for diversity are more likely to be understood and accepted. We therefore bid for, and obtained, funding to conduct a quantitative review of our department’s reading lists in order to assess the range not only of authors, but also of topics and ideas. We found that male White authors wrote the majority of the readings, with women of colour authoring just 2.5 per cent of works on our curriculum. Our reading lists also featured disappointingly little theoretical diversity, with very little coverage of feminist, critical race or queer theory approaches, for example. We therefore used the standard methodologies and approaches of our discipline in order to point towards the silences and gaps that a decolonizing approach would seek to remedy. In this article, we explain our approach and findings. The project has been educational in the best sense and has disrupted hierarchical relationships between staff and students. It has helped us think more deeply about how data and research inform, and sometimes limit, change, as well as how the process of learning about how knowledge, including reading lists, is generated can support decolonization in itself.

This article is an account of a joint staff-student research project in the Department of Political Science at UCL, UK, where around five hundred master's students and a similar number of undergraduate students study at any one time. The project treads a slightly uneasy path between demands for straightforward diversity and calls for a deeper transformation associated with challenging the colonial heritage of the university curriculum. The project responds to, and hopes to further, campaigns for decolonization by providing evidence of the gaps, silences and missing voices on our own curriculum and by showing where transformation needs to occur. This was achieved by conducting a content analysis of the current curriculum and producing descriptive statistics to show the problems and exclusions. We went further than similar pieces of research, which have mainly focused on diversity of authors, and instead tried to develop a fuller picture of how far our curriculum needs to be decolonized in terms of the theoretical approaches, topics and examples it contains. However, in doing so, we found that we had tacitly to accept some key premises that we also want to disrupt. This article therefore not only straightforwardly reports our methods and findings, but also reflects on the usefulness and pitfalls of the research, including the collaborative nature of the work we did together.
We begin by explaining why, and then how, we went about the research, and what we found. We then ask how effective it is to challenge an overwhelmingly White and mainstream curriculum using the standard methods of that same curriculum. In conclusion, we explain that the project's most important contribution is to show how the practical experience of working on such a project can demystify the curriculum and enable students to develop their own critique, beginning the work of decolonization.

Decolonization and diversity in the higher education curriculum
In making her demand for representation, Lorde (2017) is excoriating about the tendency for inclusion to be made contingent on the ability to conform to dominant structures, and insists on differences in experiences and forms of knowledge as a key to making common cause for justice. Movements for decolonization similarly seek to go beyond the call simply for more 'inclusion'. Ideas about inclusion and diversity imply the grudging acceptance of the 'other', while failing to challenge the othering practices inherent to dominant knowledge regimes, which can be traced back to the colonial purposes and histories of the modern university (Albayrak, 2018). While diversity may well be needed to provide opportunities for minoritized students and academics, without structural change, the lives of those 'diverse' scholars can be made intolerably difficult, working within structures that promise to accept and tolerate them, rather than promoting the critiques that would challenge systematic oppressions and exclusions in the first place (Begum and Saini, 2019).
We take seriously the view that decolonization cannot be a 'metaphor' for diversity initiatives that leave colonial structures intact (Tuck and Yang, 2012). Rather, it must be a genuine engagement with discovering and challenging those parts of the curriculum that emerge from the colonial project of knowledge which sought to legitimize empire and control the colonized periphery (Bhambra et al., 2018: 5). In this vein, decolonial critiques have focused on the production of positivist knowledge as part of the heritage of colonial knowledge practices. As Dalia Gebrial (2018: 24) puts it, positivist approaches to knowledge appear simply to 'reveal facts … that are worth revealing, in a process removed from power'. If the facts that are worth teaching at university are simply 'out there' waiting to be imparted to students, then it does not particularly matter who the researcher is. By this logic, it may be desirable to 'diversify' the authors on a curriculum for reasons of fairness or representativeness -and perhaps to provide more role models for a diverse student body -but the knowledge produced by these diverse voices will essentially be the same. It is this indifference to difference that Lorde (2017) was emphatically against, insisting rather that one's positioning in social hierarchies precisely conditions one's ability to know different things and in different ways. Relatedly, thinking about how to wrench open the university curriculum has been preoccupied with showing that the apparently objective knowledge which it claims to impart is -on closer inspection -bound up in colonial practices and hierarchies which perpetuate gendered, racist and ableist thinking (Das, n.d.). Furthermore, as Tuck and Yang (2012: 2) point out, it is mobilized as a tool of oppression: not only do colonial 'perspectives and worldviews get to count as knowledge and research', but 'these perspectives -repackaged as data and findings -are activated in order to rationalize and maintain unfair social structures'. Academic research that privileges Western epistemologies and values also marginalizes and delegitimizes other practices and types of knowledge, making it more difficult for students who are carefully trained in them to access alternative ways of knowing and researching (Chalmers, 2017: 101;Emejulu, 2019: 204;De Sousa Santos, 2007: 47).
The project of changing reading lists and putting canonical writers in context is concomitant with a critique of positivism. The positivist creation of the supposedly neutral Western scholar (Dennis, 2018: 193) is deeply connected to the image of universality that allows 'the canon' to exist as such. As Coleman (n.d.: 2) argues, decolonization requires in part that canonical authors are 'put in their place'. Once the unmarked scholar is placed in a specific historical and temporal context, 'that place which is no place, the place from which they speak is exposed' (Dennis, 2018: 196-7). It is in such exposing, in supporting students to develop a critique of positivist and objectivist research, that decolonization can begin to occur.
Positivist and objectivist research is not necessarily the same thing as simply the use of quantitative methods. At least since W.E.B. Du Bois's (2007) pioneering work The Philadelphia Negro, published in 1899, showed in detail the living conditions of Black Americans in Philadelphia, statistical data have been at the service of improving the lives of people of colour, if used thoughtfully for that goal (White, 2011). In this vein, Agostinone-Wilson (2013) defends the use of quantitative methods in critical, including decolonial, research. She notes that while qualitative research is the methodology of choice for critical researchers, it was frequently used in service of empire (Agostinone-Wilson, 2013: 54-8), and advocates critical quantitative research to make actionable change (Agostinone-Wilson, 2013: 60; see also Barnes, 2018). Cokley and Awad (2013: 29) also engage, as we do, with Lorde, maintaining that it is not quantitative methods that she refers to as the 'master's tools', and that, when explicitly aligning with certain values, quantitative methods can challenge injustice and effect policy change (Cokley and Awad, 2013: 30). They argue that critiques of quantitative methods should lie with particular (mis)uses of them, not with the methods themselves, and that they can, in fact, be particularly useful due to their disproportionate influence on public policy (Cokley and Awad, 2013: 36). Phelps (2020: 7) agrees that 'to effect social change, a robust set of [quantitative] methodological tools is a tremendous asset'. For these authors, quantitative data support decolonization efforts.
The most significant contributions on questions of the usefulness of quantitative methods come from Indigenous scholars, most notably Walter and Andersen in their book-length discussion, Indigenous Statistics (2016). They stress that quantitative data, rather than being neutral, play a role in constituting reality (Walter and Andersen, 2016: 9).
We therefore need to focus on why and how particular questions get asked, how data are gathered and interpreted, and so on (Walter and Andersen, 2016: 10). Not only is Indigenous quantitative research possible, they argue; it has unique advantages in positioning Indigenous researchers as 'knowers' who are able to unsettle the categories that have been used to oppress and erase Indigenous peoples. The Western reverence for numbers and statistics means Indigenous quantitative data cannot be easily rejected, enabling the research to be more effective in achieving positive change. A good example for our purposes is the Decolonising SOAS Working Group (2018: 5), which advocates the use of quantitative data to evidence racialized inequalities in degree attainment and staff seniority.
A decolonial and critical approach to quantification, therefore, pays particular attention at the level of conceptualization. Walter and Suina (2019) note, for example, that it is not good enough simply to add more survey items to a health survey focused on Indigenous deficit, but rather it is necessary to start from an awareness of the constitutive role of statistics, and to develop categories and conceptualizations that are meaningful to Indigenous communities from the outset. We support decolonization, and our project was aimed at opening up our curriculum to alternative knowledge practices. We are therefore committed to an ontology that assumes that statistics are among the practices that constitute reality and can impose racialized deficit narratives on marginalized and minoritized groups (Walter and Suina, 2019: 236). However, in practice, our study took 'gender' and 'race' to be static categories and aimed to produce a standard, 'factual' account of what we study and how. As we go on to discuss below, what is important here is what we learned from this about how the power that shapes everyday acceptance of these categories as 'facts' also shaped our ability to generate and communicate knowledge that might make a difference.
We started out by following in the footsteps of studies that have already used quantitative methods to ask related questions. Pflaeger Young et al. (2020), for example, demonstrate the relative under-representation of women in the discipline of political science in the UK, whereas Maliniak et al. (2013: 889) highlight the gender citation gap in international relations literature. Foster et al. (2013) investigated how many modules dealing with issues of gender and sexuality there were at the 16 'top-ranked' politics departments in the UK, finding that only 12 modules (out of 629) dealt with gender, and just 1 explicitly dealt with sexuality. Hardt et al. (2019) have put together an extensive database of PhD-level political science and international relations syllabuses from the United States, coded by the gender of the author, finding significant under-representation of women. A similar study by Colgan (2017) found that 82 per cent of readings in core graduate international relations syllabuses in the US were by men, although on syllabuses designed by women, this fell a little to 78 per cent. In line with these findings, in the UK, a study of all 2015/16 international relations syllabuses at the London School of Economics at undergraduate, master's and PhD level found that women had written around 20 per cent of the required and background readings assigned (Meibauer et al., 2018). Schucan Bird and Pitman's (2020) study, although small, used coding to compare reading lists in the sciences and social sciences on both gender and ethnicity, finding promising results on gender in the social sciences, where women and men were equally represented. However, they found that only 30 per cent of readings in the sciences were authored by women, and the majority of authors (90 per cent in the social sciences and 65 per cent in the sciences) were White. They use these findings to call for both a more representative and decolonized curriculum, but also for more clarity on what a representative or diverse reading list might look like in practice. We respond to that call by suggesting that descriptive representation, or the inclusion of more diverse authors, does not on its own enable students to challenge the mainstream forms of knowledge that emerge from imperial and colonial histories or to learn about counter-examples of voices and histories from the margins (see Shilliam, 2021, for numerous good examples of how this might look in our discipline). We therefore wanted to go further and look not only at diversity of authors, but also at theoretical approaches, geographical examples and topics, hence attempting to broach the work needed for decolonization beyond adding more diverse authors.
We therefore conducted a content analysis of a representative sample of our department's reading lists at undergraduate and postgraduate level. To the best of our knowledge, we are the first such project to investigate the characteristics of reading lists that go beyond author demographics.

Methods: Coding reading lists together
We worked as a team of eight student coders, with support from one member of academic staff (Cathy), to code all the compulsory readings from a sample of reading lists from 2018/19. Cathy instigated the project, recruited the team, provided advice and support, convened and chaired regular meetings, and explained the disciplinary context, alongside teaching coding and content analysis. The students did all the coding, including reliability checks, analysed the results and produced the tables. Students were remunerated for their work.
The collaborative approach was central throughout the project. We add our voices to those (for example, Decolonising SOAS Working Group, 2018;Dovey, 2020) who recognize the value of working jointly on projects to effect change in universities, particularly by challenging traditional power structures between staff and students. The research process was made possible through the range of expertise that joint working enabled: some students had expertise in quantitative methods, while others' lived experiences helped us contextualize what we were doing and communicate the stakes of the study more powerfully. Thus, 'expertise' was not limited to the figure of a teacher or academic, but rather Cathy supported the group to implement their own skills and ideas through regular meetings and ongoing discussions on the team's lively WhatsApp group. Decolonizing the university will be a multifaceted process going beyond reading lists, and the process of working in a collaborative and diverse environment empowered us as legitimate knowledge producers. For Cathy, working with the team enabled her to be part of a bigger piece of work than she could manage on her own, and for the project to benefit from skills (such as data visualization) with which she was not personally familiar. It was, though, a vulnerable experience from start to finish, both in terms of handing over control of the research process in ways that meant it was not always done exactly the way she would do it and in terms of responding to the usual academic criticism (including a revise and resubmit process for this article) in front of her students. It is a powerful experience to share responsibility and vulnerability in this way, demystifying the research process and destabilizing the idea that it is the teacher who 'knows' and the students who must learn.
As this pedagogical experience powerfully demonstrated to us, the curriculum is about much more than reading lists. It encompasses not only other forms of content (such as lectures), but also pedagogy and relationships. Nevertheless, students spend a great deal of time reading. For a standard British 15-credit module, students are expected to spend 130 hours in independent study, versus 20 hours in the classroom. Much of this study time is implicitly understood to involve reading. Reading lists also matter because they are the starting point of engagement in a discipline for those who will go on to write syllabuses of their own one day, reproducing the core disciplinary knowledge. There are many examples of groundbreaking work, crucial to the intellectual development of the discipline in its time, which were not cited or included in syllabuses, so that they ended up being forgotten or neglected (see, for example, the accounts in Vitalis, 2015;Shilliam, 2021). We therefore took the compulsory readings on our syllabuses as the focus of analysis.
First, we developed a coding scheme and associated set of decision rules. This was done jointly during regular team meetings in the summer term of 2019. We had a fairly good sense already of what we wanted to code for on the basis of prior research and our own theory-driven interests, so we constructed the coding scheme in a concept-driven, deductive way before looking at our data (see Schreier, 2012: 85). We were interested in aspects of diversity beyond authors' characteristics -such as types of topics studied -and in those cases we developed our coding rules from scratch.
After constructing an initial coding scheme, we conducted a pilot: all eight coders used this draft scheme to analyse the same reading list. This allowed us to check whether all coders had the same interpretation of the coding scheme. The reading list was well-known to Cathy, but not to the coders, which enabled some pedagogical input to clarify more complex categories. Any discrepancies were discussed during a team meeting, and coding instructions were made more specific. The resulting coding scheme included 5 sections and 24 variables (see Table 1). The scheme was translated into an Excel spreadsheet, and the non-open questions were programmed to allow only specified values.
For the purposes of this article, the most important variables we coded for each reading were: the gender and race of the author(s); whether or not the topics of race, gender, sexuality or disability were discussed; the geographical location of any case studies or significant examples; and theoretical approach. For papers with multiple authors, each author was categorized by race and gender, and we also coded for their position in the list of authors. Gender and race of authors were inferred from publicly available information including names, pronouns, photographs and interviews, where available (following Schucan Bird and Pitman, 2020). For case studies, we counted those instances when the reading contained a separate section with a geographical example or when the authors referred to examples used as 'case studies'. We coded the geographical location, aggregating to the country level.
The last category, on theoretical approach, requires some discussion. As explained above, it is important for students to be able to critique mainstream knowledge. A significant minority of students consistently report on student evaluations and in staff-student committees that our curriculum in the department is narrow compared to the discipline as a whole, and to what is on offer at other institutions, particularly so far as it enables students to critique or understand alternatives to positivist empirical work and analytical political theory ('analytical political theory' is generally contrasted with 'critical' or 'continental' approaches (see Blau, 2017: 6-7)). Looking at the various working groups and sections of the discipline's professional associations, for example, we find groups dedicated to: Colonial, Postcolonial and Decolonial; Critical Studies on Terrorism; Gendering International Relations; Interpretivism in International Relations; Post-Structural Politics; Marxism; Left Radicalism; Feminist, Post-Colonial, Queerly Interventions and many others (see BISA, n.d.; PSA, n.d.; EISA, 2021). Our hunch was that the theoretical approaches characterized by these groups are not well represented in our reading lists.
There is no incontestable way to capture narrowness of the curriculum, and it was also not our philosophical aim to construct impermeable boundaries between 'mainstream' and 'alternative' knowledges, which is impossible. Rather, we wanted to know whether we would find evidence that students are reading work within a narrow range of approaches, when they would like the option of encountering more theoretically diverse work influenced by thinkers as different as, for example, Kimberlé Crenshaw, Judith Butler, Karl Marx, Michel Foucault, Cynthia Enloe, Frantz Fanon, Mittie Maude Lena Gordon, Linda Tuhiwai Smith or the Union of Physically Impaired Against Segregation. We therefore defined 'non-mainstream' broadly to include any work that critiqued or went beyond the standard positivist or analytical work, and named critical race, feminist, Marxist, postcolonial, decolonial, queer theory, poststructural approaches and the social model of disability as possible examples. These examples were explicitly named for three main reasons. First, they capture some widely used approaches in the social sciences that have enabled marginalized, minoritized and/or colonized people to analyse and critique practices of exclusion and oppression. Second, they are the sorts of ideas that students in our department tend to mention when they advocate for a more expansive curriculum. Third, it was plausible that we would find some readings starting from these approaches on the syllabus, based on our group's personal experiences. Had we found any readings discussing Indigenous statistics or African philosophy, say, we would certainly have placed them in the 'non-mainstream' category, but it was not a surprise that we did not. There were, of course, some borderline cases, such as liberal feminist work, and readings about whether political or economic research can be value-free. In those cases, out of fair-mindedness, we coded these as 'non-mainstream', and therefore our final numbers may overestimate the amount of 'non-mainstream' work we found, at least in the judgement of more radically minded readers. The coding of 'non-mainstream' work has been controversial. We have been criticized for the fact that the examples of 'alternatives' we identified had little in common and do not form a coherent body of work. It is difficult to respond this, as a diverse range of possible approaches is what we are hoping to show was lacking. We do not know how to look for theoretical diversity in a different way, and this criticism has tended to feel like a way of derailing the conversation, rather than offering practical alternatives. Others worried that we were simply advocating importing ideas from other disciplines, which our students would be welcome to study if only they changed degree and took world literature, say, instead. However, it is important to note that academic disciplines themselves emerge from colonial histories (Shilliam, 2021: 18) and that policing the boundaries between them has in the past served to consign work by important thinkers in political science, such as the Black scholars Alain Locke and Ralph Bunche, from politics departments (where they would be read by future lawmakers) to African American studies departments, where they are appreciated and understood, but confined to the margins (Vitalis, 2015: 13-16). In any case, as discussed above, we are trying to show that work that is being done in our discipline is not represented on our reading lists. Finally, some interlocutors said that they did not particularly find anything worthwhile in the alternative approaches we listed or that there were other more useful ways of teaching about race, gender, sexuality, disability and so on. This is clearly a question for academic debate, which we cannot solve here, but which therefore could usefully be the topic of discussion in classrooms. Our aim, then, is to give an empirically grounded sense of the narrowness of our reading lists that may give pause for thought to students and staff who care about theoretical diversity.
We next needed to select an appropriate sample of reading lists to code (Abbott and McKinney, 2013: 319). We considered coding all the department's reading lists, but the pilot showed that this would be too time-consuming. We therefore concentrated on the department's 28 compulsory modules, because making a module compulsory signals to students that it contains foundational material and also makes the reading for that module mandatory for all students on a programme. The department runs 11 programmes in total, 10 at postgraduate and, in the relevant year, 1 at undergraduate level. Each programme has at least 1 core compulsory module (more for undergraduates), and students also take compulsory research methods modules. Of the compulsory modules, 3 were not coded, either because the reading lists were not available or because they did not specify required academic readings. In total, we coded 8 undergraduate and 20 postgraduate compulsory modules. We also randomly selected 20 per cent of the department's optional modules to get as accurate a sense as possible of the range of readings to which the average student might be exposed. In total, we coded 3 undergraduate and 9 postgraduate optional modules. All reading lists were accessed from module websites or, in a few cases, from module lecturers directly. In total, 40 modules and 1,360 non-unique readings were analysed. This does not give us a complete picture of the department's readings, but it does give a powerful indication, particularly given the stark results outlined in the section below.
Once we had collected our sample of reading lists, the coding was done manually by our team of eight coders in July and August 2019. We each coded all the required readings on five or six reading lists as first coders. In addition, in order to check reliability, we each acted as second coder. Second coders each analysed one week's worth of material on a reading list first coded by someone else. This was done to ensure that our findings are reliable, replicable and not dependent on subjective impressions of individual coders. The rate of agreement between the coders was above 90 per cent for most variables, indicating a high level of confidence in our findings. There were a few relevant areas where intercoder agreement went below 90 per cent. We had 87 per cent agreement on whether a reading included race as a topic and 86 per cent on the case study variable, with disagreement usually about whether a particular example was substantial enough to count as a 'case study'. Not surprisingly, our intercoder reliability was lowest on the variable judging whether readings were mainstream or non-mainstream. Despite quite detailed joint work to clarify this variable, and detailed coding instructions, we only reached intercoder reliability of 79 per cent. This is clearly because making the judgement about whether a reading is 'mainstream' or not is a difficult intellectual exercise, particularly if the coders have had relatively little exposure to critical and non-mainstream theories. We can therefore take our findings for this variable as a broad indication only, but given the low number of nonmainstream readings (despite our tendency to include readings as non-mainstream where plausible), we nevertheless think that the finding is valuable.
It is important to note that our qualms about the objectivist nature of content analysis were not purely theoretical, but were central to some of the difficulties we had in coding. For example, we were interested in the race of authors, and particularly in whether or not our curriculum is overwhelmingly made up of readings by White authors. However, processes of racialization are not natural or given, but are rather complex social constructs that differ in time and place, and are always conditioned by the operation of power (Lewis, 1998). It is clearly not possible to know from bibliographic information on a reading list either the complexities of how particular authors are racialized or how they might personally identify in terms of their racial identity. We therefore coded on a binary 'White/Person of colour' variable, which was an uncomfortable experience -but perhaps not as awkward, unreliable or logistically difficult as trying to use university websites to guess at more detailed identity categories, or contacting academics to request this information! We also coded on a 'male/female' gender binary, endeavouring to code all authors as they identify. To the best of our knowledge, none of the authors on the syllabus identifies outside of the gender binary. This was likewise uncomfortable, not least because we used these tools in the full knowledge that gender is also a complex and ambiguous social construction which is produced, as much as described, by attempts to make it a binary variable (Butler, 2007), and which is, furthermore, a colonial construction (Lugones, 2016). Similarly, aggregating case studies to the country level naturalizes borders and states in ways that critical scholars in the discipline have sought to problematize (Shapiro, 1989: 15;Parker and Vaughan-Williams, 2009). We discuss the implications below.

Results: The 'facts' about our curriculum
Our results were troubling, if not surprising. Tables 2 to 5 provide a summary of our relevant findings. For convenience, we are here reporting primary authors only, as there    is a convention in the social sciences that the main contributor to an article is named first. We also looked at patterns of collaboration among authors of readings, but this information does not significantly change the overall finding that White men dominate the syllabus, so for reasons of space we do not discuss that here. It is particularly notable that being assigned a reading authored by a woman of colour is extremely rare in our department: just 2.5 per cent of primary authors. Just under a quarter (24.9 per cent) of primary authors are women, and people of colour comprise just 7.3 per cent of primary authors in total. For comparison, 70.4 per cent of primary authors are White men. These findings are disappointing in comparison with Schucan Bird and Pitman's (2020) analysis of a social science syllabus, which was gender equal, but it is in line with the studies by Colgan (2017) and Meibauer et al. (2018).
The diversity of authors is strongly associated with the characteristics of the module convenor. Whereas male lecturers assign readings by female primary authors in 17 per cent of cases, this share is more than two times higher for female lecturers (37 per cent). The difference is highly statistically significant at the 5 per cent level. We observe similar differences by race, as authors of colour comprise 15.8 per cent of all primary authors in readings assigned by lecturers of colour, but only 6.7 per cent in those assigned by White lecturers. Although we recorded only 4 lecturers of colour compared to 37 White lecturers, the difference in the ethnic diversity of authors assigned is statistically significant at the 5 per cent level, with p-value = 0.008.
Of the total 1,360 readings analysed, 135 readings (9.9 per cent) addressed any of the topics race, gender, disability or sexuality, 38 (2.8 per cent) of which looked at a combination of two or more of these topics. Predominantly, these were made up of readings based on race (85 readings, 6.3 per cent) and gender (79 readings, 5.8 per cent). Only 0.81 per cent of readings focused on disability. Furthermore, 10 out of the 11 readings that did, came from the same reading list (International Law and Human Rights). This module is a core component of the Human Rights MA, and it is not open to students on any other degree programme. Similarly, only 0.81 per cent (11 readings) of readings addressed sexuality, 7 of which were sourced from the same module (Gendering the Study of Politics: Theory and Practice). This is an elective module taken by around forty students each year. As Foster et al. (2013) have also pointed out, this may communicate to students that sexuality is an 'optional' topic for the study of politics, which is very unfortunate.
We found 298 case studies from 74 countries and 6 continents. By far the most commonly analysed countries are the United States (76 cases; 25.5 per cent) and the United Kingdom (54 cases; 18.1 per cent). The only other 2 countries with more than 10 case studies are India (14) and Germany (13). In contrast, there is a paucity of case studies from Africa, as well as Eastern Europe and Southeast Asia. The concentration of case studies in the Global North is not encouraging, although the department's three modules on international development were, as chance would have it, not included in the sample. Most of the modules we looked at could have included a more diverse set of examples without entirely changing focus and scope, with just one (Politics of Economic Policy in Post-Industrial Democracies) plausibly constrained to Global North examples in its scope. (The question of whether or not it is ever very useful to think about the post-industrial democracies in isolation from their colonial pasts and their ongoing relationships with other countries is not something we agree on as a group, but is worth considering.) It is also important to note that studying the colonial peripheries from a Eurocentric perspective was central to colonial power, and therefore adding more diverse case studies in and of itself does not necessarily decolonize the curriculum. Given the emphasis on mainstream approaches in the curriculum discussed below, it is clear that even where a range of geographical areas are covered, this is not usually from local or critical perspectives. (See Roelof, 2020, for a good example of a broad Introduction to Politics syllabus that focuses primarily on examples and case studies from the Global South and also on disrupting Eurocentric knowledge.) Despite the discussion above on the difficulties associated with coding for theoretical diversity, and the fact that intercoder reliability means we are least confident in our data for this variable, the findings are stark enough to be worth reporting. For relevant readings, we found that 82.5 per cent adopted a mainstream approach and 17.5 per cent took a more critical, non-mainstream approach. Of the 211 nonmainstream readings we found in our sample, 91 (43 per cent) were concentrated in just four modules. This finding demonstrates the narrowness of our curriculum.
These results mattered to the team personally. As student coders who worked long hours typing 'White' and 'male', the experience brought up viscerally strong emotions. It is no longer possible to deny the fact that our curriculum does not provide the tools to challenge colonial forms of knowledge, and does not cover the full range of possible authors, examples and debates in our discipline. Our long-held suspicion that something is amiss was made real and corroborated. This was both distressing and empowering, helping us understand the power dynamics in the university in ways that were pedagogical: a true learning experience. The feelings of alienation that we had experienced were suddenly backed up with incontrovertible evidence, providing a powerful sense of relief. These data, this proof, made it possible for the feelings that have been held unspoken about the curriculum to be voiced, and they validated our experience of disengagement and alienation.
The overwhelmingly positivist leanings of the curriculum intersected with the lack of diversity among the authors in important ways. A curriculum matters for more than symbolic purposes: as Edward Said (1978) has shown, power and colonial knowledge are inextricably linked. The curriculum promotes objectivity and neutrality as a set of core values, but it also privileges a particular set of voices and ways of seeing the world. It thus implicitly makes a claim that this curriculum with these authors and these examples is the objective view of the world.
There is a real irony in the fact that a piece of positivist, mainstream research that categorizes in ways we oppose was also the tool that helped us understand our situation and validate our emotions. This experience shows the power of the positivist curriculum: our feelings can only be heard and made real, even to ourselves, through the very methods that reproduce oppressive practices, such as naturalizing race, gender and states as static, mutually exclusive categories. Thus, while questioning the dominance of mainstream approaches, our research nonetheless uses mainstream tools and, in some sense, must subscribe to a system of knowing which perpetuates the Western scientific paradigm that simultaneously and necessarily marginalizes other world views.
As such, the project is riven with ambiguity. On the one hand, by pointing towards the shortcomings of our reading lists, we hope to communicate to an audience that is committed to the power of data and numbers that there is a problem: knowing that exactly 92.7 per cent of primary authors on UCL political science reading lists are White, for example, shows to students, staff and wider society the extent and the gravity of the problem. On the other hand, our aim is not to encourage the mere presence of 'the other'. By doing research to point towards the silences and gaps in our reading lists, we also learned in a visceral way about the power relations inherent in knowledge production.

Conclusion: Finding the right tools
In summer 2020 -about a year after we compiled the original coding -there was a sudden uptick of interest in our project from students and staff in our department. This was related to the Black Lives Matter protests, which brought conversations about the problem of the White and colonial university to public attention. It was useful and strategic to have the data at hand to help us to challenge our department to make a change that will be measurable. The department has now made a commitment to repeat the study in 2021/2, and this signals a desire for things to improve, particularly in terms of diversity of authors and topics. As Sara Ahmed (2012) has argued, 'diversity' is a discourse that people use in order to do things, and some of the things we have managed to do tend in the right direction. We therefore agree with those scholars who suggest that quantitative methods do not have to be the 'master's tools'. It depends on how you use them and what you use them to do.
Yet, while there appears to be a growing acceptance that representation of women and people of colour on the curriculum needs to improve, it has been much more controversial and difficult to make the case for real decolonization and a wider range of theoretical approaches. This shows, in Audre Lorde's (2017: 110) words again, that 'only the most narrow parameters of change are possible and allowable'. Yet many of us had life-changing experiences when we first read critical race, feminist, queer or poststructural theories, and we began to understand the processes that simultaneously maintain the fiction of objectivity and also produce our own marginalization and exclusion. It is therefore discouraging that our university education limited the knowledges to which we were exposed: we would have liked to have been given the tools to do our own work of dismantling and critique. Yet this change cannot be achieved through student demand alone, because students who have not read or been taught these alternative works do not yet know how transformative they might be. It is therefore not the use of quantitative methods per se that we want to criticize, but rather the positivist assumptions -dating back to the colonial histories of social scientific research -that these methods reflect, and do not constitute, the reality of gender, race, state borders and so on.
We all (students and staff) learned not only how to do a research project, but also how colonial histories and power relations mark all knowledge production. Struggling over the difficulties of coding 'gender' and 'race', while noticing that these categories tended to be taken for granted by our colleagues and fellow students, was instructive in terms of what gender and race are and how they are maintained by research practices. Meanwhile, struggling to be taken seriously in our desire to include more theoretical diversity on the curriculum has helped us understand the processes of power that produce disciplinary boundaries, reading lists and syllabuses. For students, this helped to demystify the curriculum and gave concrete form to our suspicion that it is not something neutral or inevitable. We had conversations about how reading lists come to be put together, who writes them and what pressures those people are under. The project helped the students understand, for example, that casualized lecturers may have few choices about what or how they teach, that lecturers may emulate the curriculum that they were taught themselves because they think it has a stamp of authority, and that students may have more power than they think to demand a change, but need to learn what the change is that they would like to see. We are now able to think more deeply about how data and research inform change and to pursue joint discussions about working towards that change politically and over the longer term.
Such different figures as Mahmood Mamdani (2019: 17) and Kehinde Andrews (2018) have argued that we cannot abandon the university or lock ourselves out of its resources and opportunities. Rather, they suggest that we must use the privileges it affords to rethink it from the inside or transform it, putting its tools at the service of struggles for liberation. In this vein, Andrews (2018: 139) suggests that 'to truly dismantle the master's house means to overturn and not redeem it … The struggle becomes how to subvert the tools not how to abandon them'. Our project made those tools available to students, not only to use them in officially sanctioned ways, but also to put them to the work of critique that we hope to have demonstrated here. There is no agreement in our team about whether we want to reform or overturn the university, how thoroughgoing the demolition would be, and what the replacement structure might look like, although recent endeavours to start a Free Black University look to some of us like a promising move.
We were hoping to dismantle homogeneous spaces where 'colleagues notably wince at the term "decolonise"' (Begum and Saini, 2019: 198) and to create an environment in which calls for decolonization are understood and accepted. We wanted this to start with reading lists, so that we could learn about the ideas that have been transformational, as well as seeing more authors who look like us, and learning about the topics that affect our lives. We know that our research is insufficient to fully enact the change we want to see, and we feel ambiguous about the tools we have used, but not regretful, because doing this work together gave us (at least partly) the education we were longing for.
Ultimately, university classrooms are 'a microcosm of and impetus for broader societal transformation' (Le Grange, 2016: 3). It is only by disrupting the power relations between staff and students so that we can jointly develop and critique the curriculum that we will ultimately dismantle the master's house.