COVID-19 AND THE INTERNATIONAL BACCALAUREATE: A COMPUTER-ASSISTED DISCOURSE ANALYSIS OF #IBSCANDAL

ABSTRACT

Covid-19 has occasioned ongoing shifts in discourse as language changes to reflect and shape new stages of the global pandemic and different voices weigh in on topics, such as infectious diseases and vaccine efficacy. This study looks at an instance of this that relates to the ‘global education industry’, where cancellation of the International Baccalaureate’s May 2020 high stakes examination instigated a wide-ranging discussion about the organisation. This was triggered by the publication of IB results for 174,355 students in 146 countries, many of which showed large discrepancies between predicted and final grades. Using computer-assisted discourse analysis and a corpus of tweets containing the hashtag #ibscandal, patterns of language use are analysed, providing valuable new insights into the impact on students in different national contexts.

Keywords:

• corpus linguistics
• critical discourse analysis
• International Baccalaureate
• Twitter
• global education
• #ibscandal

1. Introduction

In March 2020, due to Covid-19 and worldwide school closures, the International Baccalaureate (IB) organisation cancelled its high stakes diploma examinations for the first time in its 52-year history and in their place devised an alternate form of assessment based on an algorithm to calculate final grades (International Baccalaureate Organization, 2020a). On 6 July 2020, results for 174,355 students in 146 countries were published accompanied by congratulatory messages on the IB organisation website and Twitter feeds celebrating the Class of 2020 (International Baccalaureate Organization, 2020b) for their achievements during a difficult year of isolation, lockdown and remote learning (International Baccalaureate Organization, 2020c). At the same time, however, a variety of other sources reported issues with IB final results, which in many cases turned out to be lower than expected (Asher-Schapiro, 2020, July 21; Lough, 2020, July 8; Simonite, 2020, July 10). Since these results could affect university admission and/or scholarships, students, parents, teachers, academics and journalists demanded to know how grades were calculated and what statistical model was used. An online petition with the hashtag #ibscandal calling for ‘Justice for May 2020 IB Graduates’ collected 15,000 signatures within the first four days. This study examines reaction to these results on Twitter between July and September, 2020.

In similar circumstances with national education systems (e.g., A-levels), governments responded by overturning or nullifying results to benefit students. In contrast, as linguistic patterns on #ibscandal show, the IB organisation is constructed as being silent and uncaring while IB students, typically associated with rank and privilege, are constructed as the voiceless marginalised. This negative discursive construction of the IB is highly unusual and in stark contrast to its typical positive prosody. Furthermore, #ibscandal became the key space for a new source and different type of IB discourse, one that was not produced by the IB organisation.

The present study examines discourses that emerge in this new context. Drawing on a corpus of tweets containing the hashtag #ibscandal, keyword analysis is conducted using AntConc (Anthony, 2020). The corpus is divided into nine one-week sub-corpora, each compared to the remainder. Analyses reveal shifts in discourses that are intertextually linked to events in the wider world and provide a rare and important window into the impact of the global education industry on students in different national contexts.

Over its 52-year history, the IB’s popularity and numbers of schools around the world has continued to grow, with 5,489 schools in 158 countries to date that offer one or more of four programs: Primary Years, Middle Years, Career-related, and Diploma (International Baccalaureate Organization, 2021). With headquarters in Switzerland, assessment centre in the United Kingdom, and regional centres in the Netherlands, Singapore and the United States, the IB is transnational and a key player in the global education industry (Steiner-Khamsi and Draxler, 2018). The focus of this study is the IB Diploma Program (DP), originally created for university-bound students in their final two years of high school. At the end of their program, all DP students are required to undertake standardised examinations, held in May for northern hemisphere schools and in November for those in the southern hemisphere. The May 2020 examination session encompassed 174,355 students in 146 countries: the 10 with the highest number are shown in Table 1.

TABLE 1. Number of May 2020 IBDP candidates for top 10 countries (IBO, 2020)

Country	Number of candidates
United States	89,100
Canada	11,962
China (including Hong Kong & Macau)	6,983
Netherlands	5,070
India	4,662
United Kingdom	4,503
Spain	3,560
Ecuador	3,286
Mexico	3,113
Germany	2,163

Due to the pandemic in 2020, the IB organisation cancelled the May final examinations for the first time in its history and instead implemented an alternate assessment model based not only on individual student performance but also on the school’s record of previous IB results, thereby creating a ‘bespoke equation’ for every school (IBO, 2020a). The final results published on July 6 were met with consternation and questions from a variety of sources including students, parents, teachers, journalists and researchers in different parts of the world. This reaction provides a unique context for IB discourse. Given the dual nature of this organisation – a global brand that is operationalised in a local context – research about it tends to be small-scale, focussing on how the different IB programs are implemented in particular schools (e.g. Xi, 2015), countries, (e.g. Dvir et al., 2018), or regions (e.g., Lee et al., 2021). The May 2020 examination event provides a unique opportunity to understand the IB in its global role.

2. Theoretical and Analytical Framework

This study combines Corpus Linguistics (CL), the Discourse Historical Approach (DHA) and Social media critical discourse studies (SM-CDS) to search for linguistic patterns and identify discourses (i.e., systematic uses of language that point to underlying values and assumptions) (Baker et al., 2008; Wodak and Meyer, 2016). Studies using corpus-based (critical) discourse analysis to examine social issues have become well established since the seminal article by Baker et al. (2008) (Nartey and Mwinlaaru, 2019). The power of this approach lies in the analysis it makes possible of large volumes of real-life language use as it occurs in a given society, based on the principles that choice in language is not random, meaning is connected to use, and studying how language is used in a given society gives fresh insight into values and assumptions underlying social issues.

To understand how micro-level linguistic choices may be linked to macro-level social practice, the DHA offers a heuristic four-level model of context: (i) the language of the text; (ii) intertextual and interdiscursive relationships between texts; (iii) institutional frames of the specific context of situation; and (iv) broader socio-political and historical contexts. This analysis of context is then combined with analysis of discursive strategies to gain insight into underlying values and assumptions that vary in degree of intentionality depending on context. The DHA focuses on five types of discursive strategies: (i) nomination/referential (how are social actors represented?); (ii) predication (how are evaluative attributes linguistically realised?); (iii) argumentation (how are these claims justified?); (iv) perspectivisation (from what point of view are these claims made?); and (v) intensification/mitigation (are these explicit or implied?).

SM-CDS builds on well-established core principles of CDS to take into account new social media communication, considered part of ‘a new paradigm of communication’ (KhosraviNik, 2018, p. 582) that emphasises the social rather than technological aspects of interaction. It follows a ‘context dependent, critical analysis of communicative practices/content with a socio-political critique’ (p. 585) by proposing horizontal and vertical contextual analyses that focus on practices and links across platforms situated in the socio-political context of the users in society.

Twitter is not only an important source of public opinions and information but also a way for users to create ambient communities around shared topics using the affordances of the platform (Zappavigna, 2011), such as hashtags (#) and usernames (@) (Page, 2012). Due to its popularity and ubiquity, Twitter presents a crucial source of social information and Twitter studies cover a wide range of topics, such as online hate speech (Hardaker and McGlashan, 2016), Donald Trump’s tweets (Clarke and Grieve, 2019), vaping (Damiano et al., 2020), and election discussions in Canada (Vessey, 2021).

Many also use the affordances of the platform to challenge the power and inequality manifest in societal structures and institutions, known as ‘hashtag activism’. These activists use a hashtag symbol to signal solidarity with a particular movement, such as #MeToo or #BlackLivesMatter, as a way for ‘historically disenfranchised populations to advance counternarratives and advocate for social change’ (Jackson et al., 2020, p. xxxviii).

The present study builds on this body of research using computer-assisted discourse analysis to capture the dynamic aspect of tweets by examining how one topic or ‘discourse strand’ (Wodak, 2021) unfolded in real time, an approach that makes it possible to disclose how shifts in discourses were intertextually linked to events in the wider world. In other words, participants were not simply reacting to events but were also positioning themselves and events differently depending on what was going on. These shifts reveal values and assumptions over time, further highlighting and amplifying key issues surrounding the high-stakes examination results.

An important part of corpus linguistics research and the present study is the analysis of keywords, determined through word frequency and statistical tests. Scott (1997, p. 236) defines a keyword as any word that occurs ‘with unusual frequency’ in one corpus when compared to another corpus. Keywords act as a useful signpost for avenues of further enquiry (Baker, 2006). Such analysis can reveal patterns of language use that might not be noticed in a small sample of texts or be available through intuition, thus making this approach a valuable way into the data. Particularly relevant for this study is the inclusion in the analysis of tagging, with the Twitter affordances of hashtag (#) and mention (@) as keywords to reveal discursive strategies. Also a factor is the constraining aspect of Twitter’s 280 character limit, which highlights the importance of word choice. Tagging thus functions as a ‘meaning bearing form’ (KhosraviNik, 2018, p. 587) that indexes roles and relevance in the wider social context.

During the two months covered in this study, #ibscandal generated a total of 4,806 tweets. In contrast, the official IB Twitter accounts posted a total of 47 tweets during this same period: @iborganization = 35;@ib_dp account = 12. The clearly heightened activity of #ibscandal points to the important role it played during this time, as a place not just for voicing opinions but as a crucial source of information, often not available anywhere else, such as links to press reports or youtube videos relating to the IB grades. Participants included a wide range of stakeholders across geographical locations, which reflected the IB’s global reach, and thus provided a unique context in which much was being said about the IB but not by the organisation.

The following research questions guide this study:

1. What key themes emerge from #ibscandal each week?

2. How do these relate to the wider context?

3. What values and assumptions are associated with the IB?

3. Data and Method

Twitter data were collected from #ibscandal beginning with the first tweet to use this hashtag (Monday, 6 July 2020) and spanning two full months (Monday, 7 September 2020), for a total of 64 days. As is standard when constructing a corpus, data first had to be cleaned before use. In this case, unrelated tweets, such as advertisements were excluded, as were non-English tweets (Catalan, Dutch, Finnish, French, German, Greek, Norwegian, Spanish, Swedish, unidentified) and retweets. The final corpus consists of 4,278 original tweets and 117,253 words (Table 2).

TABLE 2. #ibscandal corpus divided by week

Week	Date	Days	Tweets	Tokens	% of corpus
1	July 6–12	7	880	20,900	18%
2	July 13–19	7	665	17,052	15%
3	July 20–26	7	443	10,927	9%
4	July 27-Aug 2	7	479	13,021	11%
5	Aug 3–9	7	273	9,087	8%
6	Aug 10–16	7	419	12,110	10%
7	Aug 17–23	7	549	16,357	14%
8	Aug 24–30	7	300	9,240	8%
9	Aug 31-Sept 7	8	270	8,559	7%

With respect to ethical considerations, the data used in this study are publicly available, and their use for research studies is sanctioned by Twitter’s Privacy Policy and Terms of Service. Twitter’s privacy policy states that ‘Most activity on Twitter is public, including your profile information, your display language, when you created your account, and your Tweets and certain information about your Tweets like the date, time, and application and version of Twitter you Tweeted from.’ and ‘Twitter is public and Tweets are immediately viewable and searchable by anyone around the world.’ (Twitter, 2021a). Twitter’s Terms of Service state that ‘By submitting, posting or displaying Content on or through the Services, you grant us a worldwide, non-exclusive, royalty-free license (with the right to sublicense) to use, copy, reproduce, process, adapt, modify, publish, transmit, display and distribute such Content in any and all media or distribution methods now known or later developed (for clarity, these rights include, for example, curating, transforming, and translating). This license authorizes us to make your Content available to the rest of the world and to let others do the same.’ (Twitter, 2021b).

To trace how discourses changed over time, the corpus was subdivided into nine weeks, the first eight comprising seven days and the ninth eight days. Although events do not actually tend to occur neatly in weeks, this division was selected as the most fruitful for the size of the dataset. In addition, the passing of time by participants was frequently marked by reference to numbers of weeks (Table 3).

TABLE 3. Concordance lines for weeks

Wrote my MP @JackieDP about 2	weeks	ago, not even an acknowledgement (I must say very unlike he
Another kick in the gut after 7	weeks	of playing the waiting game
ke! thrown back without a comment or change after 6	weeks	wait
After more than (long and exhausting) 6	weeks	my daughter review for Chem and Bio came back with no cha
ing back with no change in grade after waiting 7 long	weeks	How did you get to the results? Still a mystery

Using AntConc (Anthony, 2020), the log-likelihood statistical measure, Hardie’s log-ratio as the effect size measure, and a minimum frequency threshold of 10, separate keyword lists were obtained for each week. This was done by comparing each sub-corpus to the remainder as the reference corpus in order to identify the most salient differences between them, i.e., what was the most unique feature in each period (Baker et al., 2019). Individual words for each week were first qualitatively analysed in their immediate co-text through concordance (keyword in context), collocation (words that occur together) and cluster (recurring fixed phrases) and situated in their wider social context in order to understand more fully how they were being used. Keywords were then grouped by discursive strategies and semantic domains to help draw out trends and patterns that might not be seen at the individual word level. Due to limitations of space, only the top 10 keywords for each week are presented below, discussion focussing primarily on the largest group (shown in bold), with others brought in where necessary for greater understanding.

4. Findings and Analysis

The first tweet to use this hashtag was on July 6 when IB results were made public: ‘IM IN EXACTLY THE SAME SITUATION #ibscandal’ (July 6). This ‘tweet zero’ (Hardaker and McGlashan, 2016) helps to illustrate the importance of contextual analysis as without that, it becomes difficult to understand why this person is shouting (the use of block capitals) and what situation is being referred to.

Week 1: July 6-12

The Week 1 sub-corpus is the largest and is connected to the publication of the IB Diploma results and the launching of #ibscandal as part of an online petition calling for justice. Table 4 shows the top 10 keywords in relation to this wider contextual event, which reveal that the key discursive strategy to emerge in Week 1 was nomination, i.e., the strategic naming of social actors used to amplify the #ibscandal message through numerous news and political outlets.

TABLE 4. Top 10 keywords for week 1 (July 6–12)

@bbcradio2, @bbcradiolondon, @bbcradiomanc, @cnnbrk, @sigridkaag, @conservatives #shameonyou, #ibresultsday, #myp, #pyp

Six of these keywords involve nomination strategy that tags actors using the @username designation: two are political (Dutch and British); four are news outlets, three of them British (@bbcradio2, @bbcradiolondon, @bbcradiomanc). Concordance lines for each keyword reveal additional news outlets. Although at first glance the news outlets seem to be primarily British, concordance lines revealed a broader global context as well as two types of uses: addressees at the beginning of a message (Example 1) or tagged at the end (Example 2):

Example 1: @AJEnglish @BBCAfrica @BBCBreaking @cnnbrk @CNNAfrica @ndtv you may have missed it but you need to sniff and pick on the trending #ibscandal (July 8)

Example 2: Admit your mistakes and make things right … how would you feel if it was your children being treated like this? @BBCNews @SkyNews @BBCNewsAsia @BBCr4today @BBCBreakfast @radiolancashire @bbc4live @BBCR1 @BBCRadio2 @BBCRadioManc @BBCRadioLondon #ibscandal (July 9)

As noted by Wonneberger et al. (2020), mentioning actors automatically signifies their inclusion in a post, thus creating mutual awareness.

Week 2: July 13–19

The Week 2 sub-corpus is the second largest, with the salient contextual event a statement published by the IB on July 15 to explain their awarding model (Table 5).

TABLE 5. Top 10 keywords for week 2 (July 13–19)

@drrlofthouse, @lordjimknight, @teacherdevtrust, @kategreensu, @youtube, @govuk, @dailymailuk, #accountabilitymatters, #showmethealgorithm, warwick

In relation to this contextual event, keywords indicate the dominant discursive strategy to be perspectivisation, i.e., the issue with results is presented from a professional (teacher) development point of view. Seven of these keywords are usernames, with four related to education in the UK: (i) professor of teacher education; (ii) chief education advisor at an international company serving ‘the world’s largest community of teachers’; (iii) Teacher Development Trust, a UK organisation specialising in professional development for schools and colleges; and (iv) a politician whose role is that of shadow education secretary. Collocation and concordance lines reveal additional mentions of individuals involved in professional development and teacher training as well as the head of the Russell Group of universities in the UK (e.g., University of Warwick). As with Week 1, nomination occurs as a way to both tag and address actors directly.

Concordance lines for the keyword @youtube link to three videos posted by IB teachers (Canada and UK) explaining in detail why the IB grading model was flawed. The two hashtags #accountabilitymatters and #showmethealgorithm reveal the use of predication strategies to intensify the IB’s apparent lack of transparency in not revealing its method of grade-calculation despite the organisation’s further explanation.

Week 3: July 20-26

The salient event that serves as the context for Week 3 is investigation into the IB results by the Norwegian government initiated on July 20 (Table 6).

TABLE 6. Top 10 keywords for week 3 (July 20–26)

#ibthieves, #ibcrooks, #ibmafia, #ibcriminals, #machinelearning, @tsimonite, wks, bias, ibsca, eurs

In relation to this contextual event, keywords point to a discursive shift from nomination to predication strategy, with four using the hashtag feature to indicate strong evaluative terms linked to criminality, further intensified by occurring in clusters tagged to different newspapers. As noted by other researchers (e.g. Page, 2012; Zappavigna, 2012), the searchable aspect of hashtags serves to amplify a message beyond its immediate context and thus potentially increase attention from a wider public. The keyword #machinelearning co-occurs with #ibalgorithm and is connected to a Reuters article about problems of bias in the IB grading model (Global exam grading algorithm under fire for suspected bias July 21), which becomes a point of discussion to draw attention to other forms of bias incorporated in it, as shown in the following example:

Example 4: Avi, thank you. Also strong bias against small schools and gender bias prob too. But thanks. (July 21)

The #machinelearning hashtag also occurs in tweets referring to investigations by Norway, Finland and the UK into the IB grading model:

Example 5: #MachineLearning #ibalgorithm … 2wks since #ibscandal Regulators in FI, NO, UK start to investigate (July 20)

Example 6: So what has #IBO been refusing to tell everyone … Norway sets out the questions that MUST be answered #ibscandal #MachineLearning (July 22)

Week 4: July 27-August 2

The relevant event for Week 4 is the IB’s response on July 31 to Norway’s questions about the grading model, which may have transgressed European General Data Protection Regulation (GDPR) laws. However, this topic is backgrounded as the top 10 keywords focus on accountability (Table 7), once again showing nomination as the dominant strategy.

TABLE 7. Top 10 keywords for week 4 (July 27-August 2)

@ibalumninetwork, @ibscandalglobal, @iborganization, @kauricapital, @theirc, dwight, staff, board, #ibfraud, #boycottib

Five of the top 10 keywords are usernames, all related to the IB via its alumni network and Board of Governors, the latter also linked to their respective organisations (Kauri Capital, International Rescue Committee). Concordance lines reveal a discursive strategy of nomination, i.e., calculated ways of naming (Hansson, 2015), and argumentation through the use of ‘topoi’ (Reisigl and Wodak, 2001) or ‘premise-conclusion shortcut’ (Gabrielatos and Baker, 2008) that directs the reader to accept conclusions based on the underlying premise. For example, the recurring pattern of introductions to individuals and their professional status (Example 7) points to the ‘topos of responsibility’, which states that ‘because a state or a group of persons is responsible for the emergence of specific problems, it or they should act in order to find solutions of these problems’ (Reisigl and Wodak, 2001, p. 78).

Example 7:

MEET #IBO [role], [name], [role] at Kauri Capital, Singapore

MEET Ms [name], [role], Dwight School, New York City, USA and [role] IB Heads Council

MEET Ms [name], [role] the Verdala International School, Malta (July 29)

This type of repetition points to a strategy aimed at drawing attention to IB organisation governance and the individuals who are responsible for oversight.

Week 5: August 3-9

Week 5 marks one month since IB results were published. Although the key event for this week is the publication of Scotland’s examination results on August 4 (#sqaresults), which exhibit the same problems as the IB results, it is overshadowed by continued focus on the IB’s response to Norway.

Table 8 shows the top 10 keywords for week 5, of which six use discursive strategies of argumentation and de-legitimation to describe different IB activities as a way to discredit it. For example, the keyword jenkins refers to the IB official responsible for replying to Norway’s complaint under the GDPR. Concordance lines reveal heavy criticism of his responses, referring to him as ‘Mr CopyPaste’. The keyword complaint is found in a recurring phrase @iborganization answer to Norway GDPR complaint (Aug 3) with references to the IB lacking transparency for not explaining its grading model and being only interested in making money. This idea is further emphasised by the keyword profit, which draws on a dominant discourse about the IB as a non-profit organisation, as shown in Table 9.

TABLE 8. Top 10 keywords for week 5 (August 3–9)

jenkins, annual, fraud, director, complaint, profit, @sijboltnoorda, @iselaconsuergrag, @nikbishop, #sqaresults

TABLE 9. Concordance lines for the keyword profit

Surplus in 2019%1000 higher than in 2012. $71.5 million for a “non	profit”	@IBscandalglobal #KauriCapital @theIRC
because of its booming revenue? Don’t think so. We were sold “non	profit”	not Wall Street @IB scandalglobal
The “Non-Profit @iborganization had a	profit	of $71.5 million in 2019. Why? Where did
Big Business disguised as a flower-power non-profit. It’s a for	profit	SCAM with complete disregard for its clients
financial-overview/ The @iborganization is a for	profit	corporation masquerading as a non-profit
a non-profit hold $180 million in surplus? Because it’s not a non-	profit	It’s just a rather commonplace Fraud

As can be seen in the concordance lines, repeatedly describing the IB in terms of finance serves as an intensification strategy to draw attention to an aspect of the organisation that tends to get backgrounded, as it is typically described in terms of its humanitarian mission rather than as a business (Fitzgerald, 2017).

Week 6: August 10-16

Week 6 is dominated by Scotland’s apology for its problematic results (Aug 10) and England’s publication of A-level results (Aug 13).

Table 10 shows the top 10 keywords, of which four relate to Scotland’s apology over downgraded exam results and the government’s decision to award new grades based on teacher estimates. There are recurring references to Scotland, SQA [Scottish Qualifications Authority], Scottish government, and parliament, who overturns or reverses exam results. Scotland accepts the original teacher estimates of grades instead of the ones issued by the flawed statistical model they had used, thereby making individual student appeals unnecessary.

TABLE 10. Top 10 keywords for week 6 (August 10–16)

scottish, scotland, sqa, #sqaresults, #alevelresults, predictive, processed, suffer, eu, admit

Despite the emerging consensus on flawed algorithms and their impact on individual students, potentially further discriminating against and disadvantaging already marginalised groups, there appears still to be no action on the part of the IB, whose algorithm has the same problems. The keyword predictive is found in the recurring phrases predictive algos and predictive algorithms in tweets that list the failure of such models in terms of unfairness and biases, using negative lexis such as warning, dangers, danger, unfairness, bumbling bureaucrats, garbage in garbage out, problems, failure, and audit. IB students are referred to metaphorically as neglected child left to suffer alone, with questions about how long students need to suffer before the IB takes action.

As England’s A-level results are published amidst similar outcry and intense media and political attention, there is a plea not to forget IB students in the same situation:

Example 8: Please do not forget the thousands of UK resident British students that have waited 6 WEEKS for an empathetic and considered response from the IB organisation. We fear they will be left behind (Aug 16)

Week 7: August 17-23

Week 7 is dominated by events in England, where the government followed Scotland’s lead and rejected the flawed A-level results (Aug 17) and, at the same time, the IB announced its revised grading model (Aug 17).

As shown in Table 11, of the top 10 keywords for Week 7, seven relate to grades, with a focus on A-level results being overturned in favour of teacher predicted grades. This is in contrast to the IB’s revised model that privileged externally marked coursework (internal assessments (IAs)) over teacher grades (predicted grades (PGs)). Dominant references point to a lack of fairness and equal treatment compared to A-levels, constructing a discourse of discrimination and disadvantage around IB students, who appear to have been forgotten.

TABLE 11. Top 10 keywords for week 7 (August 17–23)

changes, forgotten, internal, coursework, pgs, unjust, forget, unlawful, @luisreygonni, @bbceducation

Examination of concordance lines for internal (assessment) reveals negative attributes such as unreliable and marked down, erratically, externally, heavily, incorrectly, inaccurately, severely, thus suggesting a lack of confidence in the IB’s revised grading model. The discursive strategy of argumentation is again evident in the explicit use of lexical items, such as unjust, disparity, disadvantage, discriminated, not fair to construct a discourse of discrimination around the topoi of humanitarianism (universal human rights) and justice (equal rights for all) (Reisigl and Wodak, 2001, p. 78), as can be seen in Example 9.

Example 9: How can one accept such discrimination? One set of 18-year olds but two grading systems: A-levels based on PG but IB students graded on IAs (marked erratically). Please ensure equal treatment for IB and A-levels. (Aug 20)

Week 8: August 24-30

The Week 8 sub-corpus is dominated by events in England and the resignation of top-government officials over the flawed A-level results (Aug 25–26), blamed on a ‘mutant algorithm’ (Aug 26) by the British prime minister (Table 12). Interestingly, the focus appears to have shifted to regulation in the UK, which is unexpected given the global nature of the organisation and that the #ibscandal dataset is maximally representative (i.e., not restricted by country).

TABLE 12. Top 10 keywords for week 8 (August 24–30)

@amanda_spielman, @rtayloropendata, @lambertvictoria, @kier_starmer, @iconews, @commonsed, seven, comply, btec, cags

As Table 12 shows, nomination strategies again dominate in Week 8, with six of the top 10 keywords being usernames. Two of these reference positions of authority over regulations governing education in England (Chief Inspector Ofsted [Office for Standards in Education]), Chair of Ofqual [Office of Qualifications and Examinations Regulation]), one is the education editor for a national British newspaper, the Daily Telegraph, and one is the leader of the UK Labour Party. Two other mentions are the Information Commissioner’s Office (@iconews) responsible for data protection in the UK and the Commons Select Committee for Education (@commonsed) which oversees the Department for Education in England. The discursive pattern that emerges here points to the link between different social institutions and the regulation of education in England in particular and the UK more broadly. Concordance lines show a pattern of appeals for regulators to take some action in terms of fulfilling their responsibilities, as can be seen in Examples 10 and 11.

Example 10: @RTaylorOpenData please ask @ofqual to address the plight of IB students – the only UK students still being assessed on basis of an algorithm instead of CAG. (Aug 25)

Example 11: @ofqual where is our regulator? Why our children future was on the mercy of downgrading #Algorithm. Why #IBstudents in this #ibscandal STILL WAITING FOR FAIR results we demand IB u-turn (Aug 26)

By calling upon those charged with education oversight to pay attention to the IB situation, the argumentation discursive strategy combining the topoi of responsibility and law becomes evident.

Week 9: August 31-September 7

Although the Week 9 sub-corpus comprises eight days, it is the smallest in size, with the event of note being Norway’s (Sept 3) rejection of the IB’s revised grading model (Table 13).

TABLE 13. Top 10 keywords for week 9 (August 31-September 7)

ibgroup, legal, join, action, method, illegal, vocational, treated, @keir_starmer, @commonsed

Of the top 10 keywords, six cluster around legal action against the IB. Parents, students and teachers are invited to sign up, join the fight, register, contact, and email the address that is provided. This action is legitimised as the only way to hold the IB accountable and obtain justice, as can be seen in Examples 12.

Example 12: Legal action is the only way to get justice at this stage. Parents and students: please email [] to indicate willingness to take legal action so that @Ofqual, @ICOnews and @iborganization treat IB students fairly. (Aug 31)

Furthermore, the IB’s revised grading model rejected by Norway is described as illegal for using undisclosed metrics in grade calculation (Table 14):

TABLE 14. Concordance lines for the keyword illegal

@iborganization, you still used an	illegal	algorithimic bias to skew the grades you ‘marked’ yourselves
@CommonsEd please investigate the unfair and	illegal	grading scheme of @iborganization. @Keir_Starmer please
With the IA grading method clearly being	illegal	and reports that IAs were not marked properly, nor marked by
Legal action moving forward with IB	illegal	Algorithm still being used @educationgovuk @ofqual @ibor
Still using unknown factors, erratic results.	Illegal	Read our gdpr-lawyers’ letter to No DPA, IBO and @Kunnsk

Horizontal Findings

As noted by KhosraviNik (2018), an important part of SM-CDS is the analysis of both horizontal (user patterns) and vertical (socio-political) contexts of the data. Two main patterns were found. The first pattern was nomination strategies, i.e., referential invocations using @username across nine weeks, which made visible a hierarchy of responsibility and governance beginning with a general strategy of getting the message out through general news outlets and political representatives, moving on to teacher and professional groups, followed by people responsible for IB governance, and finally education regulators and government officials. The nominated actors suggest an implied or assumed hierarchy, as people keep going higher up the chain of command to try and get some response.

The second pattern was predication through hashtags, i.e., those functioning as evaluative (e.g. #shameonyou) rather than topic-based (e.g. #ibresultsday). Week 1 has one evaluative hashtag, Week 2 has two, Week 3 has five and Week 4 has two. As discussed above with nomination, the predication pattern becomes visible over time, with negative evaluative attributes directed at the IB and intensifying as time moves on. From Week 5 onwards, a discursive shift becomes evident that can be linked to the examination results from Scotland and England. Although these were released one month after the IB results, they contained the same algorithmic biases that yielded inequitable grades. Swift action taken by governments in favour of their students left IB students in an ambiguous situation with no equivalent response on their behalf. Although hashtags in Weeks 5 and 6 are topic-related (#sqaresults, #alevelresults), concordance lines revealed comparative statements between IB, A-level and Scottish students.

In sum, horizontal analysis of user patterns showed nomination and predication strategies being realised through the use of the Twitter affordances @username and #hashtag.

5. Discussion

The aim of this study was to examine discourses surrounding the IB as they emerged in a new context triggered by the Covid-19 pandemic’s impact on education systems worldwide, which resulted in school closures and cancellation of high stakes examinations. This new context involved diverse stakeholder voices from around the world focussed on the same topic (discourse strand), thus providing a unique space by which to gain fresh insights into the IB and the values and assumptions associated with it. Specifically, the study examined shifts in discourses over time to capture its dynamic nature and relationship to the immediate and wider social context.

Findings through keyword analysis revealed unique dominant discourses for each week. This was achieved by comparing each sub-corpus to the remainder as a reference corpus in order to highlight what differed in each week and to understand the shifting nature of discourse, that is, that it is not a fixed or static entity but fluid and responsive. Each week showed a particular preoccupation, beginning in the first week with a concerted effort to get the message out as widely as possible through a variety of news outlets, culminating after two months with a call to join forces in legal action against the IB. As with many previously taken-for-granted societal structures that Covid-19 has shown to be problematic (e.g., health and education systems that exacerbate inequality), #ibscandal points to deficiencies within the governing structure of the IB organisation, further emphasised in the contrast between the daily information updates and diverse voices (students, parents, teachers, researchers, journalists, lawyers) evident in the far more active #ibscandal compared to the relative inactivity of official IB channels. This is an inversion of the usual mode of IB operations in which the organisation controls the messaging in a proprietary way, constructing an image of a uniform and united ‘IB World’ that is the same regardless of context. Here, the organisation emerges as more complex and multi-faceted than previously thought.

Findings also showed an unclear or absent regulatory oversight of the IB in different countries. For example, while governments in Scotland and England stepped in to address their problematic results (although the action taken in England did not meet with universal approval and may have caused other problems), there appeared to be no equivalent overall national regulator over IB grades. Although it is clear that the IB organization is responsible for its grades, what emerged in the data as being less clear was the question of jurisdiction in terms of the organization itself, whether Switzerland as the headquarters or the UK as the assessment centre. As such, students were either left to appeal individually through their schools or universities or to take up legal action. One source for the lack of clarity (revealed by the Norwegian investigation) is the IB’s global structure, since its offices are located in different countries, making it unclear whose jurisdiction the May 2020 results came under, whether it was Switzerland as the headquarters or the UK as the assessment centre. In other words, contrary to the widely perpetuated notion of the IB as a single standardised entity, a more fragmented and disparate picture comes to light, with accompanying regulatory challenges.

The publication of Scottish and A-level results a month after IB results had a clear impact in terms of how the IB organisation and students were constructed. Repeated references to human rights and equality emphasised the importance of all students being treated the same, with IB students constructed as a minority group in danger of being excluded. This discourse of discrimination situating IB students as the out-group is highly unusual, as they are typically perceived as privileged and elite (Fitzgerald, 2018). In addition, the shift from global to UK perspective, particularly since the IB results in question affected 174,355 students in 146 countries, as well as the absence of other voices, such as from Canada and the US, which comprise 58% of the results, suggests that postsecondary entry requirements in other countries may not be quite as reliant on examinations as the UK.

Findings also showed the damaging impact of grading algorithms on individual students as thousands of posts on the platform related similar experiences to do with unexpected results. This was in stark contrast to statements made by institutional representatives defending the use of algorithms as a way to ensure equity through standardised results that avoided grade inflation. The power of individual voices was amplified through #ibscandal and the ‘aggregation of personal stories’ (Clarke, 2016), a fundamental aspect of hashtag activism.

The final pattern of note had to do with questions around commonly held beliefs about the IB as an educational organisation. For example, the IB’s well-known status as a non-profit foundation was discursively undermined by repeated references to its financial holdings and wealth accrued through high program fees. Furthermore, the lack of response from the IB juxtaposed to the swift action taken by governments in the same situation showed the IB as uncaring and lacking compassion (qualities central to its mission statement), thus deemed not to be abiding by its own principles, as shown by evaluative statements framed in terms of corruption, incompetence and lack of care.

6. Conclusion

The Covid-19 pandemic exposed deficiencies in our social institutions, revealing gender, racial and class divides that had remained hidden and taken for granted. The IB results debacle shone a spotlight on the organisation, revealing cracks within, drawing scrutiny and questions previously unknown due to a decentralised structure and tight control over the brand and its message. However, as a global education player working closely with national governments, these issues are not just limited to the organisation but point to larger, more fundamental concerns at the national level, as was revealed by the involvement of Norway and the focus on the UK.

The #ibscandal hashtag was launched on the same day as the IB results were published, and formed the title of the online petition calling for justice. As mentioned above, the meaning of words is intimately connected with the way they are used. To understand more about the kinds of value judgements or ‘semantic load’ (Mautner, 2009) associated with scandal, collocates were obtained using a 22 billion word reference corpus of general English (enTenTen18). The top 10 lexical collocates were watergate, corruption, analytica, involving, enron, rocked, bribery, cheating, iran-contra, sex, indicating a strong pattern linked to notions of cover-up and mismanagement. As noted by Tognini-Bonelli (2001, p. 111), ‘words which are co-selected do not maintain their independence’, suggesting that the word scandal was used as a discursive strategy.

Acknowledgements

Many thanks to Tony McEnery, Michael Fitzgerald, and two anonymous reviewers for their very helpful feedback on earlier versions of this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Social Sciences and Humanities Research Council of Canada under Grant number 756-2019-0176.