1 Introduction

ACM’s SIGCSE is one of the first organizations focused on computing education [21]. In 1970, SIGCSE launched its Technical Symposium, which was initially focused as a forum for teachers of computing to share best practice with each other, exchange opinions, experiences, and course descriptions [21, 47]. Initial discussion topics, e.g., on programming paradigms and software engineering were enhanced by the presentation of teaching tools and educational technology towards the end of the first decade of SIGCSE [47]. Over the years, computing education research (CER) has significantly matured as an academic discipline, experience reports have been joined with more rigorous research, and more attention has been put to methodology, empirical evidence, and theory use [30]. In SIGCSE Technical Symposium, the scale of submissions has evolved from 18 accepted papers in the first year [50] to 161 accepted papers in 2018 and 297 papers in 2021. In recent years, SIGCSE Technical Symposium has attracted over 1000 attendees annually [21], and its publication profile and community have shifted remarkably from its inception. Out of all the central publication outlets that publish CER, SIGCSE Technical Symposium has, by far, published the largest share of articles [3], making it one of the most influential publication outlets of CER.

As the SIGCSE Technical Symposium has expanded its publication profile and diversified its community, it is relevant to investigate that scholarly community and the community’s publication and citation practices. In this chapter, we present a scientometric angle to the development of SIGCSE Technical Symposium in order to explore the collaboration networks, shifts in research focus, and citation practices. Scientometrics provides the possibility to go beyond simple counts to offer more mature and nuanced overviews of the temporal evolution of science. In this paper, state-of-the-art scientometric methods are used to offer an in-depth perspective on the evolution of the SIGCSE Technical Symposium. Our three research questions are:

  1. 1.

    How have authors and author networks shaped SIGCSE Technical Symposium and its community over time?

  2. 2.

    How has the publication profile of SIGCSE Technical Symposium evolved in terms of most-cited papers, keyword trends and keyword clusters?

  3. 3.

    How has SIGCSE Technical Symposium evolved from the viewpoint of international collaboration?

2 The Birth of SIGCSE

The need for large scale computing education efforts was born with the mass-production of fully-electronic, programmable computers just before the mid-1950s [47]. Around that time, in June 1954, the First Conference on Training Personnel for the Computing Machine Field was convened at Wayne University in Michigan, bringing together more than 150 interested people [26]. The early computing workforce was primarily trained by the burgeoning computer industry, but pressure was building up at universities to catch up with the computing education efforts [14]. At the turn of 1960s, already 150 universities offered some training in computing [15], and competition had started between major organizations in the computing field over their curriculum development efforts [47]. ACM started an education committee in 1960, but competitors were quick to establish their programs: In 1962, DPMA offered a professional examination in data processing and IFIP started panels on information processing [20, 47].

ACM formed a curriculum committee in 1962, but its first draft curriculum took 3 years and the final ACM Computing Curriculum (CC’68) took another three. However, once it was out, it quickly established an authoritative role in higher education: Just 4 years later, 78% of US computing education programs in universities reported that they considered ACM’s curriculum valuable for their computing education efforts [51].

ACM’s CC’68 had been a major undertaking involving many computing pioneers in a joint education effort, and the same year it was published, a number of pioneers signed a petition, at 1968 Fall Joint Computer Conference in Las Vegas, to establish ACM Special Interest Committee for Computer Science Education (SICCSE) [47]. Starting from 1969, the committee’s newsletter, SICCSE Bulletin, was devoted to computing education practice: “descriptions of new courses, novel approaches to established courses, problems and solutions, comments on the development of computer science education …” [22]. In 1970, the committee became an ACM special interest group (SIG), was rebranded as SIGCSE, and launched its first Technical Symposium a day prior to 1970 Fall Joint Computer Conference [1].

By 1970, SIGCSE membership had grown to over 600 members [32]. The early SIGCSE membership was primarily North American: in a member listing in December 1970, of SIGCSE Bulletin, out of 600+ SIGCSE members 80 were from outside the US, of whom 26 were from Canada. Accordingly, the early SIGCSE symposia gathered visitors primarily from North America (Fig. 3). For almost 30 years, reviewers were all North American (although it was shown that has no effect on review scores) [50]. Noticeable changes in the international profile of the symposium started to be seen in 1990s, which will be discussed more in Sect. 6 (Fig. 3). The same diversification was also noted by the SIGCSE Bulletin, which for the first time in 1997 received over 50% of its submissions from outside the US [34]. Two shifts can be discerned in the author composition of SIGCSE papers: roughly since the mid-1990s multiple-author papers have dominated over single-author papers, and international collaboration papers started to frequently appear in 2000s (Fig. 1). Papers have been overwhelmingly from the US, with more than 92% of attendees in recent years having an U.S. affiliation [42].

Fig. 1
A bar graph plots number of articles versus years for multiple authors from multiple countries, multiple authors from a single country, single author, and not indexed. The graph plots an increasing trend.

Number of papers presented and published in SIGCSE conferences, divided into papers with a single author, papers with multiple authors from a single country, and papers with multiple authors from multiple countries (authors affiliated with institutions from different countries). Gray color indicates unavailability of country metadata

Full size image
Fig. 2
A cluster diagram illustrates a huge network of S I G C S E authors.

Co-author network of SIGCSE authors with most collaborators. Node size indicates the number of unique co-authors, edge thickness indicates number of co-authorships, and colors indicate communities of researchers who frequently collaborate together (using Louvain modularity algorithm)

Full size image

2.1 Related Work

This is not the first paper to analyse the publications of the SIGCSE Technical Symposium. One of the earliest efforts to analyse the SIGCSE Technical Symposium is that of Valentine [49], and the classifications Marco Polo, Tools, Experimental, Nifty, Philosophy and John Henry. In Valentine’s classification, Marco Polo (“I went there and I saw this”) refers to an experience report, typically of trying a new curriculum or teaching method, Tools refers to research on educational tools, Nifty including research on innovative assignments, John Henry, describing papers that push the boundaries of pedagogy, Experimental referring to research with an experimental setup, and Philosophy debating issues on philosophical grounds [49]. Valentine’s analysis sparked an interest in the idea of CER as a research field or discipline, and in turn led to a number of other efforts, both nearly immediately [38], and in the years that followed [16, 18, 43]. In addition there has been increasing focus on what the field is about, on how to understand what research has been done, and what to prioritize in the future [37, 47].

Some relevant finger-posts in the debate on CER, and what conferences should be about include a panel debate at ITiCSE 2004 [18], where three views of CER were presented and analysed, namely, classroom practice reports, observing phenomena, and subject based learning research in collaboration with educational researchers, with a fourth perspective, reporting, which focused on the need for research ethics and honest reporting, arguing that all the other three forms were valid contributions, but needed to be honestly reported.

Fincher’s book on CS Education Research [16] also emerged the same year (2004), fueling an attempt to define a Core Literature for CER [38]. Since then there have been several attempts to develop research taxonomies [30, 43, 45] and analyse the research methods commonly employed [8, 12]. These endeavours provide the background to the current attempt to understand the evolution and significance of the SIGCSE Technical Symposium and its contributor communities. In particular the data presented in this chapter extends and contrasts with prior work on classification of publications and taxonomising the SIGCSE publication corpus [6, 49]. The main innovation in comparison with prior attempts is that our approach provides access to solid scientometrics and author network analysis, empirical data visualisations, and through these analyses a new longitudinal perspective on the development of the conference community, the shifts in themes and topics and significance of publications that appear in the conference proceedings.

3 Methodology

The metadata for all (1970–2021) SIGCSE Technical Symposium papers were retrieved from Scopus, as Scopus had the best quality metadata of SIGCSE Technical Symposium papers.Footnote 1 In the early years, proceedings were published as a special issue of the ACM SIGCSE Bulletin. Articles in each special issue of the bulletin that were not part of the conference proceedings were manually removed, resulting in a total of 4982 articles.

Further cleaning was applied to fix database inconsistencies, which were manually fixed and verified against information on the ACM website. Author keywords were cleaned and equivalent keywords were combined: for instance, “CS1”/“CS 1”, “K-12”/“K12”, and “OSS”/“open source software” were grouped. The data were analyzed using the Bibliometrix R package [4]. Scopus metadata were used to construct a temporal timeline for the evolution of keyword use. A similar visualization was created for country contributions, to plot the trend of country participation over the years. The overall frequency of country contributions was further plotted on a world map.

The author analysis included frequency, earliest and latest contributions, as well as the number of total citations. A network of co-authorships was constructed using fractional counting, which has become the preferred method for co-authorship link weighting over traditional counting [39]. In traditional counting, each co-authored paper counts as one link between each pair of co-authors. In fractional counting, this link is weighted by the number of authors in the paper. Thus, the more co-authors a paper has, the weaker the link among them. Louvain community detection algorithm [13] was applied to highlight frequently collaborating authors and significant SIGCSE co-authorship communities.

For clarity, the analysis included only the top 100 authors, ranked by the number of distinct collaborators. Another network was constructed for author keywords based on keyword co-occurrence. Louvain community detection algorithm was applied to highlight important themes within SIGCSE publications. For more details on the analysis methodology, refer to chapter “Scientometrics: A Concise Introduction and a Detailed Methodology for Mapping the Scientific Field of Computing Education Research” of this book [28].

4 Authors

Since the first SIGCSE Technical Symposium in 1970, 7349 unique author names have appeared in SIGCSE proceedings. Among the authors of SIGCSE papers, 5197 (70.8%) appeared just once, and 1040 (14.2%) twice, with a mean of 1.9 papers per author. For those authors who appeared more than once in SIGCSE, the mean number of publications is 3.9. Several authors stand out for a large number of contributions to the SIGCSE Technical Symposium series. All authors on the list of 20 most productive authors (Table 1) have authored or co-authored 22 or more papers in SIGCSE. The top positions on the list of most productive authors features well known computing educators. Daniel D. Garcia from University of California, USA, was involved in 47 papers within a timespan from 2002 to 2021, with a total of 195 citations for his SIGCSE publications in the whole Scopus database, earning the top position on the list of most productive authors in SIGCSE, followed by Owen Astrachan with 39 papers and 401 cites in Scopus within a timespan from 1990 to 2019, and Mark Guzdial with 37 papers and 726 cites in Scopus, within a timespan from 1994 to 2020.

Table 1 Twenty most productive authors in SIGCSE proceedings 1970–2021
Full size table

4.1 Collaboration

Figure 2 presents co-authorships in the papers published in SIGCSE Technical Symposium. The nodes represent those authors who have most co-authors (more than five unique collaborators). The edges that connect the nodes represent co-authorships between the authors. Unconnected nodes are active collaborators, but whose co-authors do not belong to the group of most active collaborators in Fig. 2.

Fig. 3
A scatter plot of countries versus years for frequencies 250, 500, and 750. The countries are Australia, Brazil, Canada, Chile, China, Czech Republic, Denmark, Finland, Germany, Ireland, Israel, Japan, Netherlands, New Zealand, South Africa, Spain, Sweden, Switzerland, U K, U S A. U S A has maximum frequency and Chile has least frequency.

The 20 most active countries in SIGCSE proceedings by the number of papers published as determined by the affiliation of the first authors. The size of the circle indicates the number of first authors from a given country each year. In most early years, only the US and Canada exceed the threshold for visualization

Full size image

Some clearly identifiable clusters have formed around the authors of SIGCSE. The pink cluster at the center is formed around known SIGCSE contributors, editors, and award winners. These include Daniel D. Garcia (University of California), Owen Astrachan (Duke University, Durham, NC), and Henry M. Walker (Grinnell College, Grinnell, IA), many-times chair and bulletin editor in SIGCSE, as well as receiver of the Lifetime Service to Computer Science Education award. The pink cluster also features David Reed (Dickinson College), the leader of the first New Educators Workshop in SIGCSE, as well as Julie Zelenski and Nick Parlante (Stanford University, Stanford, CA, USA) and Michael Kölling (King’s College, London, England), known for his work on programming education tools, and receiver of the SIGCSE Outstanding Contribution to Computer Science Education Award in 2013, and SIGCSE Test of Time award in 2020.

The green cluster at six o’clock centers around Beth Simon (University of California at San Diego, USA; an active member of the SIGCSE community, chairing numerous editions of SIGCSE Technical Symposium), Leo Porter and Christine Alvarado (University of California San Diego), also active members of SIGCSE, having served, for instance, as editor of SIGCSE Bulletin among the years. The cluster also features Daniel Zingaro, Andrew Petersen, Michelle Craig (University of Toronto, Ontario, Canada), and CER researchers Briana B. Morrison (University of Nebraska Omaha, Omaha, NE, USA), and Steven A. Wolfman (University of British Columbia, Vancouver, BC, Canada), with connections to Brett A. Becker (University College Dublin, Belfield, Ireland) and Andrew Luxton-Reilly (University of Auckland, Auckland, New Zealand).

The brown cluster features Mark Guzdial (Georgia Tech, Atlanta, Georgia), a long-time member and Outstanding Contribution Award winner in 2019, and Barbara Ericson (Georgia Tech, Atlanta, Georgia), among others, while the Dark Green cluster at one o’clock features Kristy Elizabeth Boyer and Tiffany Barnes (North Carolina State University). The gray green cluster at two o’clock is centered around Susan Rodger (Duke University), a long-time member and chair in SIGCSE, and Stephen Edwards (Virginia Tech, Blacksburg, VA, USA), Outstanding Contribution to Computing Education Award winner in 2021 and prominent member of the SIGCSE community.

The dark pink cluster at five o’clock features long time active members and chairs in SIGCSE, Renée McCauley (College of Charleston), and Sue Fitzgerald (Metropolitan State University, St. Paul, MN, USA), while the light blue cluster at two o’clock is centered around Lillian Cassell (Villanova University, Philadelphia, PA, USA) and John Impagliazzo (Hofstra University, Hempstead, NY, USA) who is a 2007 lifetime service to computer science education award winner of SIGCSE. Other remarkable computing educators and SIGCSE community members are featured in other, smaller clusters shown in Fig. 2. The clusters in SIGCSE are dominated by US-centered networks with few connections to other countries, and feature members, chairs, award winners, and extraordinary computer science educators and researchers, some of which are also part of the most productive authors in SIGCSE (see Table 1).

Interestingly, some other highly active and awarded members of the SIGCSE community from outside the USA are not present, including Judy Sheard (Monash University, Australia) and Alison Clear (EIT, New Zealand), Lauri Malmi (Aalto University, Finland)—winner of the Outstanding Contribution to Computer Science Education Award—, Mats Daniels (Uppsala University, Sweden)—SIGCSE Outstanding Service Award and Lifetime Service to Computer Science Education Award winner—, and Raymond Lister (UTS, Australia).

5 Papers

Learning how to program is, of course, a topic of persisting interest for computing education research. The most cited paper in SIGCSE [31] is about using Scratch in K-12 computing education. With 273 citations in Scopus, the paper has become a popular reference in K-12 computing education. The second most cited paper [11] is about object-orientation in CS1 and has significantly contributed to the debates of programming languages and paradigm choices in CS1. The next two most cited papers are about pair-programming in CS1 [33], and success factors in CS [53]. Both of these papers [33, 53] have become popular references in CS education with 247 citations in Scopus each. Other top papers include a broader selection of papers on varying topics; automatic grading, gender differences, game-based approaches, learning styles, and aptitude in computer science.

Not surprisingly nearly half of the top papers (Table 2) focus on teaching programming or introductory computer science: [11] is about object-orientation in CS1, while [33] and [35] centered around pair programming in CS1, and the topic of [27] was a gamified approach in CS1. Further on, success factors in programming were investigated by Bergin and Reilly [7], while other topics that centered around CS1 were learning styles [48], modeling of learning [40], factors of persistence [5], and errors in Java programming in CS1 [23]. Other common topics included K-12, Scratch, and computational thinking (CT) [29, 31, 41, 52]. One paper in the top cited list deals with gamification, an investigation of Iosup and Epema [24]. A number of tools papers are also represented among the most highly cited SIGCSE Technical Symposium publications, including: detecting plagiarism [54], program visualisation in Python [19], and tools for grading [25]. Other highly cited topics included success factors in CS [53], choice of major [10], and gender differences [9].

Table 2 Twenty SIGCSE papers with most citations in scopus
Full size table

The impact of a publication on the community can be considered from a range of perspectives. Pears et al. [38] proposed a classification system which incorporated both sustained cumulative citation and scholarly estimation of impact, proposing a system that divided papers into qualitative categories seminal, influential and synthesis. If we consider this perspective a majority of the papers in the list are relatively old (were published prior to 2010), and can be considered established and influential. Interestingly relatively few are authored by the productive and highly collaborative, “central”, members of the community depicted in Fig. 2; the exception being Barker. One can wonder why publications by the central and most collaborative members of the SIGCSE Technical Symposium community are not more highly represented in the list of highly cited works.

There are four more recent papers (less than 10 years old) with high citations that might be considered both influential and seminal, as they seem to be popular works upon which to ground further efforts in the areas of program visualisation, modelling learning of programming, computational thinking and gamification. These are all areas of strong interest in the current CER research agenda internationally, and can probably be rightfully claimed as some of the Symposium’s more recent seminal works.

6 International Collaboration

SIGCSE Technical Symposium was launched in the United States, and has always been strongly US-oriented. Figure 3 shows frequencies of contributions per country in each year of SIGCSE, as determined by the first author affiliation. Prior to 1990, only few papers originated from outside of North America (US or Canada). The countries with most contributors are United States (11,003 author appearances), Canada (429 author appearances), United Kingdom (245 author appearances), and Australia (169 author appearances). The number of contributions from other countries has increased over the years, and nowadays an increasing number of papers in SIGCSE originate from European, South American, African, Oceanian, and Asian countries. But although the symposium’s papers today originate from many corners of the globe (Fig. 4), contributors from Africa are rare, and Asia and South America are greatly underrepresented. It seems plausible to argue that continuity of publication in the Symposium is linked to strong SIGCSE chapters and activities. This hypothesis is supported by the observation that Finland, Sweden, Australia, New Zealand, UK and Israel demonstrate continuity in participation. The aforementioned countries also are characterised by having active CER groups and PhD programmes, as well as, in most cases, access to local SIGCSE conferences such as ITiCSE (mostly in Europe), ICER (UK, Australasia, US), CompEd (held outside of North America and Europe), as well as Koli Calling (Finland, Sweden) and ACE (Australasia/Oceania).

Fig. 4
A world map highlights the S I G C S E in various continents. The frequency gradient ranges as follows, 1, 7, 55, 400, and 3000. North America has maximum frequency range of 3000 Asia Pacific has least frequency as 1.

Distribution of SIGCSE papers across the globe

Full size image

A typical paper in SIGCSE Technical Symposium does not seem to be amenable to international collaboration. Until 2000s, papers with authors from more than one country were few and far between (Fig. 1), and only some 5.4% of papers ever published in SIGCSE contain authors from multiple countries. A large portion (28.7%) of papers in SIGCSE have been authored by a single person, while out of all papers with multiple authors, some 92% were authored by authors from a single country, with only 8% including authors from multiple countries. Out of all papers, 6% could not be indexed with regards to author country. A minor increase in multi-country papers can be observed in the most recent two decades of SIGCSE Technical Symposium (Fig. 1).

7 Keywords and Themes

Keyword analysis reveals emergence, evolution, rise and fall of topics and trends during the history of SIGCSE conference. For almost 30 years in the beginning of SIGCSE, postal mail was the method of paper and review submission [50]. Electronic submission systems started to evolve from 2000, initiated by then Program Chair Henry Walker [50]. Prior to 2003, keywords were not used at all, or they were not used consistently [36]. Thus, analysis of keywords is only possible from 2003 onward. Figure 5 shows yearly occurrences for all 20 keywords that have appeared in top five keywords during one or more years of SIGCSE. Figure 5 shows popularity and emergence of keywords, such as the steady popularity of CS1, slight decline of pedagogy, and increase of K-12, computational thinking, and gender and diversity. The frequency of keywords object-oriented programming and Java clearly attenuates soon after 2015 and appear to be dying out.

Fig. 5
A dot plot of popular keywords versus years from 2003 to 2021. The frequent occurrence is 10, 20, 30, and 40. C S 1 has maximum frequency and broadening education has least frequency.

Changes in popular keywords (2003–2021)

Full size image

This is reflected in the keywords’ evolution, too, and while also the processing of papers was manual and done via postal mail prior to 2003 [50], the top keywords overall were not present in those years.

An understanding of the interrelationships between CER topic areas can be obtained by conducting a network analysis of keyword clusters. The network in Fig. 6 shows clusters of keywords, linking together keywords that are commonly found together in the keywords block of the publication. The light blue cluster centers around computational thinking (CT) linking the area to contexts such as, K-12 education, using Scratch and robotics in computing education, with a large twist in gender diversity. The orange cluster centers around introductory programming (CS1, CS2) and preliminary programming course (CS0), showing a strong interest in the research community in related pedagogies, such as active learning and collaborative learning, and known debates around teaching CS1: object orientation, choices of programming language, assessment, and motivation. The mint green cluster centers around algorithms and data structures and their visualisation, while the dark green cluster centers around collaboration and pair-programming. The pink cluster shows that there are strong interests in connecting and exploring combinations of topics of relevance to software engineering, simulation, interdisciplinary education, curriculum, cybersecurity, and accessibility in computing education.

Fig. 6
A cluster diagram illustrates a huge network of the keyword co-occurrence.

Keyword co-occurrence

Full size image

8 Discussion

Our first research question asked: How have authors and author networks shaped SIGCSE Technical Symposium and its community over time? Here our co-authorship analysis combined with our data on frequency of publication per country clearly demonstrate that the SIGCSE community remains extremely US centric with most collaborations being within a number of strong personal networks based around prominent researchers and their close colleagues and former students. Some isolated links to other networks can be detected and have been discussed earlier. Overall the SIGCSE Technical Symposium community is characterised by longstanding US-based collaborative groups. We also observe that some research communities, for instance Africa and Asia are particularly under-represented.

Our second research question asked: How has the publication profile of SIGCSE Technical Symposium evolved in terms of most-cited papers, keyword trends and keyword clusters? The analysis of keywords show that learning how to program is a topic of persisting interest in SIGCSE, which is well known [6]. Themes such as program visualisation, computational thinking and gamification are areas with strong interest within the CER community and can be considered as more recent seminal work in SIGCSE. Changes in topics post-2003 show increases in K-12 computing education, computational thinking and gender diversity, and dying out of object-oriented programming and Java. Clusters of keywords have centered e.g. around introductory programming and computational thinking and K-12.

A significant part of top cited papers in SIGCSE are about introductory programming. A 2004 meta-analysis of SIGCSE papers published between 1984 and 2003 showed that first year CS instruction, including CS1 and CS2, formed a significant portion of SIGCSE papers already in 1984–2003 [49], while recent research systematically categorised some 481 CS1 papers published during the first 50 years of SIGCSE from 1970 to 2018 with the following categories: teaching (105 papers), students (78 papers), first languages & paradigms (45 papers), tools (38 papers), CS1 content (38 papers), collaborative approaches (36 papers), CS1 design & structure (60 papers), learning & assessment (81 papers) [6]. Introductory programming is a top topic in SIGCSE, as it is in other publication forums of computing education, too [46].

Our final research question concerns How has SIGCSE Technical Symposium evolved from the viewpoint of international collaboration? Attendance to SIGCSE has greatly diversified when compared to the symposium’s early decades, when the participants were predominantly from the US and Canada. Yet even in 2021, countries and whole continents are greatly underrepresented on the map of contributions to SIGCSE. For example, voices from Africa, a home to 1.2 billion people, are seldom heard, even when the virtual format would enable online presentations. In 1970, when the symposium started, the world looked different: a number of industrialized countries were far ahead of others in the computerization race and in computing education activities, many non-English speaking countries had their own forums of the SIGCSE kind, the academic world was much more segregated than it is in 2022, and publishing in national languages was more often the norm in earlier times. But in 2022, there is a need for ACM’s prime (flagship) international computing education conference to discuss what the symposium could do to better serve the needs of computing education for all, in all countries, and not only in a few high-income countries.

The SIGCSE symposium was initially a forum for teachers of computing to share their best practices [21]. While research papers were also part of SIGCSE, the focus was initially on supporting practitioners [21]. Computing education research became more common post-2005 as the field started to evolve into its “mature era” [21]. The year 2005 also marks a turning point in computing education research where the discipline started to expand from mostly experience reports only to more rigorous computing education research [44].

8.1 Limitations

This research has several limitations. First, while Elsevier’s Scopus is accurate and maintained well [17], it is not flawless. The data has multiple issues with missing fields, inconsistencies in keywords, citation counts and references not being perfectly recorded. Manual checkups were needed in cases, but even with extensive cleaning, we can claim to have reached a representative, but not a comprehensive, sample of research. It is well known that citation counts differ between engines such as Scopus, Web of Science, Google Scholar, or ACM Digital Library. Comparing the reliability between the citation counts in different engines is beyond this work. Lastly, another limitation of our work is that we used first authors’ affiliation to analyze countries’ productivity, to avoid overrepresenting countries in which there are many authors per paper. However, this approach fails to capture all countries’ contributions.

9 Conclusion

We observe a continuing US-centric publication focus, and an urgent need to enable broader participation from the developing world. While a strong European participation starts to appear along with emerging Asian countries, e.g., China and India, there is almost near absence of half of the world. This is of course made worse by visa issues, subscription fees, and costs of travel that many scholars in the global south can not afford. In line with ACM’s initiatives for increasing global participation—founding the Global Computing Education Conference (CompEd) in 2019, held outside of North America and Europe—, SIGCSE can do more to bridge the global divide of knowledge, e.g., invest more in travel support, issue fee waivers for certain countries, endorse young scholars from the global south, or even celebrating the conference outside of the US. In addition, our thematic analysis indicates an over-emphasis on programming and the learning of programming in the historical publication data. These are worrying trends, and should be addressed as the Technical Symposium proceeds into the next decade. Our analysis also shows the weak appearance of learning theories, a fluctuating trend for pedagogy as well as learning methods e.g., active learning. As computer science education grows, more alignment with learning theories would help improve our teaching and learning. In fact, we believe that SIGCSE may be an important venue for discussions of innovative pedagogies and theories that are germane to twenty-first century computing education. Novel educational trends like educational data mining and learning analytics [2, 3], have not made it to the top 20 keywords in SIGCSE, which raises questions about how the computer science education symposium aligns with novel trends that are pioneered by computer scientists. A positive trend that continues to be strong is gender and diversity which gives hope that our research can inform practice into a more equitable and diverse future. Our analysis has kept us wondering, has SIGCSE been a driver of innovation of computing education research? or just a mirror of the community interests?