Shuichiro Ogawa
日本語

Notes · updated 2026-08-13

End-User Development (EUC/EUD): Genealogy and Limits — A Map of 152 Sources

An integrative summary for a review paper on the history, social demands, and limits of end-user computing (EUC) and end-user development (EUD), collecting 152 sources through a lightweight scoping review — from Licklider in 1960 to the LLM era of 2026 — and mapping them into seventeen clusters. Full bibliographic details for every source appear in the “References” section below (with DOIs/URLs). The internal ledger with provenance tracking is at source/review/end-user-development/papers.md (repository-internal, not published). Prior notes: design-democratization-justification and democratization-recommodification-paradox for the critical examination of democratization rhetoric, vibe-coding-design-production and maker-to-editor-paradigm for how LLMs alter making, ai-cognitive-offloading-learning for effects on learning. Collection pipeline: three parallel source-researcher runs (profile: scholarly) → paper-screening-agent (first pass, 85 items), plus three parallel supplementary runs targeting the prehistory, the application domains, and non-anglophone research (second pass, 67 items). Protocol: .claude/rules/collection-protocol.md (zero fabrication, provenance tracking).

Survey Metadata

  • Collection date: 2026-08-13 / Count: 152 (85 from the first pass plus 67 from the second; 170 candidates in total at the exploration stage, 15 merged as duplicates, 2 excluded)
  • Period: 1960–2026 (Licklider 1960 is the lower bound among the primary sources of the prehistory; the 2020–2026 low-code and LLM period is also covered)
  • Why the collection ran in two passes: because the first pass (85 items) took 1980s information systems research as its starting point, four gaps remained — the prehistory preceding it, application domains outside HCI and information systems research, research outside the anglophone world, and the application of theory from the sociology of work. The second pass targeted only those gaps and added four clusters: the prehistory (H0, H5), the non-anglophone lineage (H6), the application domains (L5), and labour process theory (D5).
  • Venue distribution: roughly 45 items from ACM conferences and journals; roughly 20 from information systems venues (MIS Quarterly, ISR, CAIS, BISE, JMIS, Decision Support Systems); 27 scholarly books and book chapters (MIT Press, Springer, Harvard UP, Stanford UP, Princeton UP, Chicago UP, Yale UP, Routledge); 4 from a history-of-computing journal (IEEE Annals of the History of Computing); 6 from medical informatics venues (JAMIA, JBI, JMIR, Human Factors); 9 from Japanese-language journals; 2 German-language book chapters; 7 arXiv preprints.
  • Exclusions: two items. One was a misattribution to Panko of “A Critical Review of the Literature on Spreadsheet Errors” in Decision Support Systems 46(1); Crossref confirmed the correct authors as Powell, Baker & Lawson, and the record was deleted. The other was an Edinburgh doctoral thesis that surfaced as a candidate in the second pass; neither its submission year nor the exact form of its title could be settled and the full text was unreachable, so it was dropped as bibliographically unidentifiable. The same kind of check was applied to Benson’s (1983) DOI, Mørch’s (1997) chapter URL, Scaffidi et al.’s estimates, and the volume and issue of Prather et al. (details in the Provenance section of the internal ledger).
  • Machine verification of the bibliography: every reference carrying a DOI was queried against the Crossref REST API and checked for agreement on author surname and year of publication (results recorded in the Provenance section of the internal ledger).
  • Confidence notes: no predatory journals or retracted papers were identified within the range checked. Venues with reservations about their review standards (ISRN Software Engineering, IGI Global journals, the EuSpRIG conference, the PPIG workshop, and Japanese technical-committee reports and conference proceedings) and unreviewed preprints are retained with a non-peer flag. The primary sources of the prehistory (1960s–1970s) are handled separately as primary-historical, since the framework of peer review at the time differed from today’s. Items judged low-relevance are also kept rather than deleted, as part of the history of the research.

TL;DR

The norm that users should be able to write their own programs was not introduced by 1980s information systems research. It arose in the 1960s, in four separate places. Time-sharing as research infrastructure (E86, E87, E88); languages for teaching computing to non-specialists (BASIC in E90, Logo in E92); the normative claim for tools not monopolized by experts (conviviality in E95, the Dynabook vision in E94 and E98); and the institutional pressure of a shortage of specialists (E93 — the 1968 NATO conference fixed the phrase “software crisis” in the international vocabulary). Only the last of these came from the supply side; the other three placed value on user autonomy itself. What 1980s EUC research inherited was chiefly the last one.

The same phenomenon acquired two names because the observers stood on opposite sides. EUC (end-user computing) appeared in the vocabulary of information systems research in 1983, defined together with the question of how IS departments should manage a user population growing at 50 to 90 percent a year (E01). EUD (end-user development) was defined in 2006 in the vocabulary of HCI, as the activity of enabling users to modify, extend, and create software artifacts, positioned as a response to the shortage of professional programmers and to diversifying needs (E23). The former is a framework of control, the latter one of emancipation, and their objects largely coincide. That duality persists today: the literature treating citizen development as a governance problem (E32, E33) and the literature treating EUD as a field of empowerment (E27, E45) barely cite each other.

Nor is the social demand voiced by the post-1980s literature singular. Nor were the four origins of the prehistory carried over intact; the correspondence is offset. At least the following four are distinguishable in this corpus. An economic demand, in which organizational development needs outran the supply capacity of IS departments (E01, E04, E05, E29); the Scandinavian demand for industrial democracy, in which workers determine their own tools (E39, E40); an educational demand that treats the ability to program as literacy (E46, E47, E48); and an industrial demand arising from the shortage of professional programmers (E23, E34). Their motives and their justificatory logics differ, yet they have circulated folded into the single word “democratization”. The critical literature (E52, E53, E54) shows that this folding is not specific to EUD but is a rhetorical structure that computerization movements have repeated with each technological wave.

On limits, the field has documented itself early and precisely. Learning barriers were classified into six kinds (E59), the cognitive properties of notations were given fourteen evaluative dimensions (E60), and why users decline to pay the learning cost was explained by an attention investment model (E61). On artifact quality, the reported level is at least one error in 94 percent of 88 audited spreadsheets (E67). And empirical work from the LLM era shows the same limits replayed at a new layer. Non-expert programmers cannot learn how to address a generative model (E74), and business users fail to detect serious flaws in generated code even when warned that errors are possible (E81).

The shape of the limits differs by the practitioner’s occupation. In accounting and finance the institutional response of auditing and internal control becomes the center of the problem (E124, E126); in medicine the phenomenon appears as compensatory work by which staff fill the gaps left by official systems (E129, E133); in scientific computing the artifacts produced become the very foundation of verifiable knowledge (E139, E142).

What this corpus lacks is any study that verifies over the long run whether what EUC/EUD promised actually happened. The literature on demands and the literature on limits are each substantial, but no study places them side by side and measures the distance covered. The path of importing theory from the sociology of work also remains open: two independent searches found no peer-reviewed study applying deskilling theory or jurisdictional theory to EUD or citizen development.

Historical Genealogy (H0–H6, 64 items)

H0. Prehistory: where the norm came from (1960s–1970s, 15 items)

The idea that users should operate computing machinery themselves appears first on the side of research infrastructure. Licklider (1960, E86) envisioned a symbiosis of human and computer, and Engelbart (1962, E87) proposed, as a framework for augmenting human intellect, a process in which users raise their own capability while reworking their own tools (bootstrapping). The CTSS of Corbató et al. (1962, E88) and the account by Fano & Corbató (1966, E89) made an environment in which multiple users touch the computer directly into something that actually existed.

From the side of education came languages for teaching computing to non-specialists. Kemeny & Kurtz (1968, E90) reported the Dartmouth time-sharing system and BASIC, and argued for the educational significance of students writing programs themselves. The final Logo report of Feurzeig et al. (1969, E92) is the first document to record a framework in which children learn mathematical thinking by writing programs.

The third is a normative claim about the ownership of tools. Illich (1973, E95) set against the industrial tools monopolized by experts the convivial tools that users can master and rework. The Dynabook vision of Kay (1972, E94) and Kay & Goldberg (1977, E98), Nelson’s (1974, E96) Computer Lib, and the Community Memory of Colstad & Lipkin (1975, E97) are attempts to translate that norm into the design and practice of computing.

Only the fourth comes from the circumstances of the supply side. The NATO conference report edited by Naur & Randell (1969, E93) fixed “software crisis” in the international vocabulary and recorded the perception that development could not keep up with demand. The backlog problem that Rockart & Flannery (1983, E01) place at their starting point had thus already been under discussion for fifteen years.

H1. The origins of EUC in information systems research (1980s–1990s, 7 items)

The starting point is Rockart & Flannery (1983, E01), who classified users into six types from a survey of 173 companies and argued that IS departments should build a “third environment” alongside the existing two. In the same year Benson (1983, E02) reported, from interview fieldwork, that EUC practice differed between mainframe and microcomputer settings. Cotterman & Kumar (1989, E06) classified end users into eight types along the axes of development, operation, and control, supplying a framework for deciding who may develop what.

What characterizes this lineage is that it welcomed growth while building in its limits. The growth-stage model of Huff, Munro & Martin (1988, E04) incorporated the arrival of a stage at which existing tools would no longer let most end users continue developing effectively. Brancheau & Brown (1993, E05) organized the decade into a framework of trade-offs between the resource constraints of IT departments and control.

H2. Tailorability and meta-design in HCI and CSCW (11 items)

The other lineage begins from descriptions of people actually reworking software. Nardi & Miller (1990, E08; 1991, E09) showed ethnographically that collaborative spreadsheet development is the rule rather than the exception, and MacLean et al. (1990, E10) proposed a button-based tailoring mechanism from the premise that no single system design can suit all users and situations. Mackay (1990, E11; 1991, E12) documented empirically the patterns by which customizations are shared, and the triggers and barriers that govern them.

Conceptual work builds on those descriptions. Mørch (1997, E15) separated tailoring into customization, integration, and extension, while Trigg & Bødker (1994, E14) traced how repeated tailoring activity gives rise to systematization within an organization. The meta-design and SER model of Fischer & Giaccardi (2006, E18) formalize design as an activity continuing into use. The irony in the title of Nardi (1993, E17), A Small Matter of Programming, captures this lineage’s self-understanding.

H3. The technical genealogy of end-user programming (4 items)

The edited volumes of Cypher (1993, E19) and Lieberman (2001, E20) established the lineages of programming by demonstration and programming by example, and Myers, Pane & Ko (2004, E21) derived design guidance for more natural languages and environments from empirical studies of how people think about programming tasks. In 2022 this lineage connects to generative language models (E22).

H4. The formation of EUSE and the field’s self-definition (5 items)

The definitional chapter by Lieberman et al. (2006, E23), the ACM Computing Surveys review by Ko et al. (2011, E24), and the mapping study of 165 papers by Barricelli et al. (2019, E25) mark three moments of the field’s self-description. Batalas et al. (2021, E26) problematize the dispersion of EUD definitions itself, showing that the field has not settled its own extension.

H5. Historical research on the prehistory (9 items)

The primary sources of the prehistory have been given a different image by later research in the history of computing. Bardini (2000, E101) rereads Engelbart’s vision through a framework of coevolution, and Turner (2006, E104) traces, from the junction of the counterculture and Silicon Valley, the cultural provenance of the idea that individuals become autonomous through computing. Ensmenger (2010, E105) shows how technical authority concentrated in the occupation of the programmer, and how the discourses of software crisis and backlog were bound up with the politics of that authority.

This lineage also imposes constraints on the historical account a review paper can give. Rankin (2018, E107) excavated the educational computing of Dartmouth and Minnesota and argued against the received view that attributes the origins of personal computing to a handful of inventors. The three faces of HCI that Grudin (2005, E103) distinguished — computer operation, information systems management, and discretionary use — likewise make it difficult to draw EUD as a single linear lineage.

H6. The non-anglophone lineage (13 items)

Searching outside the anglophone world, the first finds are contemporaneous German-language sources. Müller (1982, E110) and Jahner (1982, E111) discussed databases and languages for the Endbenutzer (end user) a year before Rockart & Flannery.

In the Japanese-language world the situation differs. Studies taking EUC as their subject in peer-reviewed journals are confirmed in 佐藤 (1994, E112) on a management-stage scale and 高原 et al. (2002, E113) on a development environment, but both imported and applied US EUC management theory. From the 2010s onward the term almost disappears from Japanese academic literature, and in its place reports of individual in-house development cases (E117, E119, E120, E121) are scattered across technical-committee meetings and conference proceedings whose review standards cannot be confirmed. For the terms widely used in practice — 野良システム (“stray systems”) and 属人化 (“person-dependency”) — no peer-reviewed literature was found. Most recently, 花原 (2026, E122) examined in-house development with generative AI in university library operations and pointed to risks of safety and technical debt.

The multi-tier subcontracting structure specific to the Japanese software industry does exist as quantitative research in industrial economics (E115). No study, however, was found that explicitly ties it to EUC/EUD as a demand factor, so making that connection becomes work for the review paper itself.

Social Demands (D1–D5, 41 items)

D1. Organizational and economic demands (11 items)

The first driver of EUC is the fact that development demand outran the supply capacity of IS departments (E01, E04, E05). That configuration recurs under new names. Behrens (2009, E29) and Fürstenau et al. (2017, E30) describe systems born outside official IT departments as an ambivalent phenomenon; the latter theorizes the persistence of shadow IT, from a savings bank case, as a feedback loop between the social construction of risk and interdepartmental power relations. Low-code and citizen development in the 2020s restate the same supply-demand imbalance in a new vocabulary (E33, E35, E36, E37), and Viljoen et al. (2024, E32) organize quality degradation, shadow IT, and technical debt into a governance problem from 30 interviews with citizen developers and professionals.

The population estimate most often cited from this lineage needs care. Scaffidi, Shaw & Myers (2005, E38) estimated from US labor statistics that in 2005 more than roughly 45 million people used spreadsheets and databases, and projected for 2012 a total of 90 million end users, of whom more than 55 million would use spreadsheets or databases and more than 13 million would self-identify as programmers. The figure “55 million” tends to circulate with its year stripped off, but it is a projection for 2012, not an observation.

D2. Demands of democracy, participation, and labour (7 items)

The Scandinavian participatory design tradition traced by Ehn (1988, E39) and Kensing & Blomberg (1998, E40) carries a justificatory logic distinct from productivity: industrial democracy led by trade unions. Bratteteig & Wagner (2014, E41) show from two large projects that this participation is in practice a matter of distributing power and decision-making, and often remains formal. Le Dantec & DiSalvo (2013, E42) recast participation as the ongoing formation of infrastructure (infrastructuring) rather than a one-off engagement. The computational empowerment of Iversen et al. (2018, E43) and Dindler et al. (2020, E44) is an attempt to connect this lineage to education.

D3. Educational and cultural demands (6 items)

Papert (1980, E46) and diSessa (2000, E47) positioned programming as a medium of expression rather than a tool of efficiency, and Wing’s (2006, E48) computational thinking became the vocabulary of K-12 policy (E49). Vee (2017, E51) uses the apparatus of literacy studies to trace historically how programming was institutionalized, moving from military and administrative infrastructure into firms and individual use. This lineage contains its own critique. Denning (2017, E58) points to the vagueness of the definition of computational thinking and the difficulty of measuring it, and Guzdial (2015, E50) argues that differences in why everyone should learn computing govern learning goals and pedagogy, yet are reduced to a single frame.

D4. Critiques of democratization rhetoric itself (7 items)

Iacono & Kling (1996, E52) theorized the mechanism by which social movements promoting computerization construct narratives of liberation through the rhetoric of technological utopianism. Barrett et al. (2013, E53) developed this into a rhetorical analysis, and Han et al. (2025, E54) show that the same framework applies directly to today’s AI discourse. Irani (2019, E55) and Costanza-Chock (2020, E56) show, through long-term fieldwork and critical examination respectively, that discourses of empowerment and participation coexist with economic value extraction and with the reproduction of structural inequality. Placing these four side by side, the question “is this democratization the real one?” meets, first, the answer that it at least shares the rhetorical structure of previous waves.

D5. Labour process theory and the sociology of professions as a gap (10 items)

Seen from the side of labour, two theories should be usable for EUD. Deskilling through the separation of conception and execution (Braverman), and the process by which professions contest jurisdiction over work (Abbott). Yet two independent searches found no peer-reviewed study applying either of them to EUD or citizen development.

What exists is adjacent territory. Kraft (1977, E143) applied the deskilling thesis to programmer labour, and Ensmenger & Aspray (2002, E148) re-examined it historically and showed that deskilling and reskilling proceed in parallel. Ensmenger (2001, E147) followed from archival sources how the professionalization movement among programmers failed, and Barrett (2004, E149) systematized the labour process of software development as empirical research. Including the sceptical rejoinder (E150), the objects of all these analyses are employed professional programmers, not non-specialists in business departments.

The work that treats the structurally closest question in a different vocabulary is the control theory of Munro et al. (1987, E144). It discusses how authority is redistributed between the IS department and end users, which, transposed into the vocabulary of the sociology of work, comes close to a description of jurisdictional contest. The authors themselves, however, cite neither Abbott nor Braverman. In 2026 a theoretical proposal for introducing labour process theory into HCI appeared (E152), but within the scope of its abstract it touches on neither EUD nor citizen development.

Outside this corpus, democratization-recommodification-paradox connects Braverman’s deskilling and Bourdieu’s symbolic capital to democratization by AI. That note, however, is a record of debate rather than peer-reviewed research, and its citations of the theories carry [unverified] markers. Using it in a review paper would require separate work returning to the primary texts.

Limits (L1–L5, 47 items)

L1. Cognitive limits (5 items)

The six learning barriers Ko, Myers & Aung (2004, E59) identified from observing Visual Basic .NET novices — design, selection, coordination, use, understanding, and information — are the reference point of this lineage. The cognitive dimensions of Green & Petre (1996, E60) supplied properties of notations such as abstraction gradient and viscosity as evaluative axes. Blackwell’s (2002, E61) attention investment model explains end users’ avoidance of programming as a rational judgment rather than a deficit of ability, which makes it an internal objection to democratization rhetoric. Beckwith et al. (2006, E62) argue that these barriers are not evenly distributed across users, and that software design itself contributes to gender differences.

L2. Artifact quality (8 items)

Spreadsheets are at once EUD’s success case and the domain where evidence of its limits has accumulated most densely. Panko (1998, E64) synthesized the finding that cell-level error rates match those of simple clerical tasks while nearly every spreadsheet contains an error at the sheet level; his 2015 restatement (E67) reports, from field audits across 13 studies, a mean cell error rate of 5.2 percent and at least one error in 94 percent of 88 audited spreadsheets. Powell, Baker & Lawson (2008, E65) criticized the field on the grounds that although the prevalence of errors is agreed upon, knowledge of their kinds, mechanisms, and detection remains thin. On the intervention side, experiments show that visible test-coverage indicators improve detection efficiency and correct overconfidence (E69, E70).

L3. Organizational and governance limits (2 items, 6 counting sub-cluster assignments)

Alamin et al. (2022, E73) identified 40 challenges facing low-code developers from roughly 33,000 Stack Overflow posts, and Silic et al. (2025, E72) report, from an employee survey (n=140) and executive interviews (n=10), a pathway by which unsanctioned generative AI use erodes an organization’s knowledge base through overreliance. The shadow IT literature in D1 (E28, E30, E31) supports this line from the organizational side.

L4. New limits in the LLM era (12 items)

Generative AI widens EUD’s reach and at the same time relocates its limits. Sarkar (2023, E57) formulates this as the generative shift hypothesis while examining why code as a form of representation is still required. What the empirical work shows is that the limits did not disappear but surfaced at another layer. Liu et al. (2023, E74) identify an abstraction gap in which non-expert programmers cannot learn how to address a generative model. In the experiment of Perry et al. (2023, E76, n=47), participants using an AI assistant wrote significantly less secure code and were overconfident that their code was secure. Vaithilingam et al. (2022, E78) observed that Copilot did not reliably improve task completion time or success rate, yet many participants continued to use it. Virk & Liu (2025, E81) showed that business users failed to detect serious flaws in generated data analyses even when explicitly warned that errors were possible.

Read through Blackwell’s attention investment model (E61), what generative AI lowered is the cost of learning, not the cost of verification. As long as verification cost has not fallen, the range of what non-experts can build widens without widening the range whose correctness they can vouch for themselves.

L5. How the limits take shape in each application domain (20 items)

Outside HCI and information systems research, the limits take a different shape in each occupation.

In accounting and finance, the center of the problem is not the existence of errors as such but who verifies and who bears responsibility. Powell et al. (2008, E124) designed a staged auditing protocol for spreadsheets in operational use, and Coster et al. (2011, E126) reported, from a survey of 38 US companies undergoing SOX compliance, the actual state of which controls are (and are not) implemented at each stage of the lifecycle. Caulkins et al. (2008, E125) recorded, from interviews with practitioners, how spreadsheet errors propagated into actual decisions.

In medicine, end-user development appears less as an outlet for creation than as compensatory work by which staff fill the functional gaps of official systems. Koppel et al. (2008, E129) systematized nurses’ workarounds to barcode medication administration systems, and Balka (2010, E131) reported the “ghost charts” maintained outside the official record. The ethnography of Mörike et al. (2022/2024, E133) describes how frontline staff themselves build shadow systems mixing digital and analog media. Precisely because this is a high-risk domain, failures of that compensatory work bear directly on patient safety. This domain, however, is not only a record of failure. REDCap, by Harris et al. (2009, E130), continues to be used widely as a platform on which clinical researchers who are not programmers can build their own data collection forms.

In scientific computing, the limits extend to the foundation of knowledge itself. Hannay et al. (2009, E135) and Pinto et al. (2018, E138, 1,574 R users) reported that many researchers develop software through self-teaching and self-devised practices. The consequence appears at the level of reproducibility. Wang et al. (2020, E139) showed that many of 936 notebooks on GitHub could not be re-executed, and Samuel & Mietchen (2024, E142) analyzed 27,271 notebooks linked to biomedical publications and reported that, of the 10,388 Python notebooks that declared their dependencies, 1,203 ran without error. The danger O’Brien (2025, E82) saw in overreliance on AI code assistance is an extension of a structure that predates LLMs.

Implications for the Review Paper

The three lines of inquiry are not independent. History (H) shows that EUC and EUD described the same phenomenon through the opposed frames of control and emancipation, and that the vocabulary of emancipation goes back to four origins in the 1960s. Demands (D) show that those four were folded into the single word “democratization” while their motives and justificatory logics remained different. Limits (L) show that cognitive and quality constraints bear on every one of the demands, while the way the constraints appear differs by the practitioner’s occupation.

The usable axis for a review paper is therefore not “did EUD succeed?” but the way of posing the question as “against which demand did which limit bite, and how hard?” Quality and governance limits (L2, L3, L5a) bear on the organizational demand (D1); cognitive limits (L1) bear on the educational demand (D3). Where professionals build software outside their main occupation (L5), the locus of the limits lies not in individual ability but in the structure that assigns them development work without training or institutional support. And what bore on the democratic demand (D2) cannot be answered within the range of this corpus. The power analysis of participation (E41) and the critiques of democratization rhetoric (E52–E56) suggest that the demand itself may never have presupposed a measurement of attainment.

Placing the prehistory (H0) at the starting point sharpens the question one degree further. Of the four origins in the 1960s, only the software crisis lineage came from the circumstances of the supply side. The remaining three (research infrastructure, education, and the autonomy of tools) place value on users building things themselves, irrespective of whether development demand is met. As long as EUD’s attainment is measured only by the clearing of the backlog, those three fall outside the scope of evaluation from the start.

Gaps (Unaddressed Threads)

  1. No empirical study directly tests the unfulfilled promise. D1–D4 address why EUD was demanded and how that discourse was constructed, but this corpus contains no study that tracks longitudinally and quantitatively whether it was realized. No peer-reviewed follow-up verifying the realization of the 2012 projection in Scaffidi et al. (E38) was discovered in either search. Given the prehistory (H0), that verification needs to be carried out after separating “which promise” into the four origins.
  2. Deskilling theory and the theory of professional jurisdiction have not been applied. The second pass re-searched the labour process theory and occupational jurisdiction lineages in order to verify this judgment independently, but did not reach any peer-reviewed study applying them to EUD or citizen development (D5). What exists is adjacent literature on professional programmers (E143, E147–E150) and IS control theory that treats an isomorphic question without the vocabulary of the sociology of work (E144). Building that bridge could itself constitute the review paper’s theoretical contribution.
  3. Coverage outside the anglophone world remains thin. The second pass added German-language (E110, E111) and Japanese-language (E112–E122) work, but no empirical research with English abstracts could be reached in the Chinese- or Korean-language spheres. For the Japanese-language sphere, no peer-reviewed literature was found corresponding to the terms frequently used in practice (野良システム, 属人化), so the asymmetry between academic and practitioner discourse is itself an observation. Verifying that asymmetry requires a different kind of corpus, one that includes practitioner venues.
  4. No study traces the discontinuity between the prehistory and the 1980s. Between the primary sources of H0 and the EUC research of H1, no study tracking citation relations could be found. Even granting that it is natural for Rockart & Flannery (1983, E01) not to cite Licklider or Illich, by which route the norms of the 1960s were converted into the management theory of the 1980s cannot be traced from this corpus.
  5. No EUD research on education or public administration was found. The second pass filled in accounting, medicine, and scientific computing (L5), but repeated searches produced no peer-reviewed literature meeting the criteria on the construction of teaching-material systems by teachers or the in-house development of administrative systems by public employees. The Japanese-language case reports (E120, E121) touch on university administrative work at most.
  6. The citizen developer’s own perspective is thin. Apart from E34, no research on skill formation and career paths from the practitioner’s side could be found, and D1 leans toward the organizational governance viewpoint. This gap is the first place where the labour process theory of D5 could be connected.

Unverified Items

  • E02 (Benson 1983): the number of interviewees (67 users, 19 IS professionals) is carried over from the exploration stage and has not been re-checked against the original [unverified]
  • E13 (Henderson & Kyng 1991): exact page range and the DOI of the original edition (Design at Work, 1st ed., Erlbaum) [unverified]
  • E20 (Lieberman ed. 2001), E35 (Prinz et al. 2021): no DOI; only a book page and a permalink [unverified]
  • E37 (Ajimati et al. 2024): DOI and full author list unconfirmed [unverified]
  • E39 (Ehn 1988): multiple editions (1988/1989/1990) were confirmed, but catalog records disagree on the publisher-ISBN correspondence, leaving the first edition’s bibliography unsettled [unverified]
  • E52 (Iacono & Kling 1996): the DOI’s existence was confirmed, but the full-text page was unreachable [unverified]
  • E59 (Ko et al. 2004): participant count not stated in the abstract and unconfirmed [unverified]
  • E66 (Panko & Aurigemma 2010): DOI unconfirmed (ScienceDirect URL only) [unverified]
  • E67, E68 (Panko 2015/2016; Croll 2009): the review standard of the EuSpRIG conference is unconfirmed [unverified]
  • E77 (Sarkar et al. 2022, PPIG): no formal DOI found; PPIG’s review standard also unconfirmed [unverified]
  • E83 (Meske et al.), E85 (Ge et al.): unreviewed preprints [unverified]
  • E27 (Paternò 2013, ISRN Software Engineering), E31 (Kopper et al. 2018, IJITBAG): venues with concerns about review standards, accepted conditionally [unverified]

The items added by the second pass (prehistory, non-anglophone lineage, application domains, labour process theory) are as follows.

  • E91 (Licklider & Taylor 1968): the original text in Science and Technology was unreachable, so the bibliography rests on confirmation through secondary citation databases [unverified]
  • E99 (McCarthy 1992/1983): the year varies between the draft (1983) and the journal publication (IEEE Annals 14(1), 1992), and the journal version has not been reached [unverified]
  • E101 (Bardini 2000): ISBN unconfirmed (year of publication, publisher, and Open Library work ID confirmed) [unverified]
  • E108 (Haigh & Ceruzzi 2021): DOI not obtained (ISBN, publisher, and year confirmed from multiple independent sources) [unverified]
  • E114 (富澤 & 明星 2009): only the abstract from the national conference is available; the full text is unconfirmed [unverified]
  • E115 (Minetaki & Motohashi 2009), E144 (Munro et al. 1987), E146 (Star & Strauss 1999): exact page ranges [unverified]
  • E116 (岡部 et al. 2014): no DOI could be confirmed, substituted with the CiNii CRID [unverified]
  • E117, E119 (中所 2014/2016): not peer-reviewed, being IEICE technical reports
  • E120 (神馬 et al. 2023), E121 (小西 et al. 2024): whether they were peer-reviewed could not be confirmed [unverified]
  • E123 (Grossman et al. 2007), E125 (Caulkins et al. 2008), E132 (Saleem et al. 2011), E135 (Hannay et al. 2009), E136 (Prabhu et al. 2011), E141 (Cosden et al. 2022/2023): bibliographies confirmed, but some details of sample size or method have not been reached [unverified]
  • E126, E127 (EuSpRIG 2011/2012): as with E67 and E68, the conference’s review standard is unconfirmed [unverified]
  • E151 (Meilvang 2020): Sociology was inferred from the SAGE DOI prefix, but the journal name display could not be confirmed [unverified]
  • E152 (Qin & Cheon 2026): an unreviewed preprint whose CHI 2026 acceptance is stated only on the arXiv page and is unconfirmed on the ACM side [unverified]

In the machine verification of the bibliography, the 110 items carrying a DOI were queried against the Crossref REST API and checked for agreement on author surname and year of publication. The only remaining mismatches were four items differing in character forms of umlauts and apostrophes (E30, E82, E110, E133), all of which are the same works. Kay (1972, E94) has no bibliographic record registered in Crossref, but the DOI resolves and reaches the corresponding page in the ACM Digital Library. The DOIs of the seven Japanese-language sources (registered with JaLC) were likewise confirmed to resolve.

References

Primary sources of the prehistory (H0)

Origins of EUC in information systems research (H1)

Tailorability and meta-design in HCI and CSCW (H2)

Technical genealogy of end-user programming (H3)

Formation of EUSE and the field’s self-definition (H4)

  • E23. Lieberman, H., Paternò, F., Klann, M. & Wulf, V. (2006). End-user development: An emerging paradigm. In End User Development, Springer HCI Series 9, Ch.1. https://doi.org/10.1007/1-4020-5386-X_1
  • E24. Ko, A.J. et al. (2011). The state of the art in end-user software engineering. ACM Computing Surveys, 43(3), art.21. https://doi.org/10.1145/1922649.1922658
  • E25. Barricelli, B.R., Cassano, F., Fogli, D. & Piccinno, A. (2019). End-user development, end-user programming and end-user software engineering: A systematic mapping study. Journal of Systems and Software, 149, 101–137. https://doi.org/10.1016/j.jss.2018.11.041
  • E26. Batalas, N., Lykourentzou, I., Khan, V.J. & Markopoulos, P. (2021). Reconsidering end-user development definitions. IS-EUD 2021 (LNCS 12724), 19–35. https://doi.org/10.1007/978-3-030-79840-6_2
  • E27. Paternò, F. (2013). End user development: Survey of an emerging field for empowering people. ISRN Software Engineering, 2013. https://doi.org/10.1155/2013/532659

Historical research on the prehistory (H5)

The non-anglophone lineage (H6)

  • E110. Müller, G. (1982). Relationale Datenbanksysteme für den Endbenutzer. In Betriebs- und Wirtschaftsinformatik. Physica-Verlag. https://doi.org/10.1007/978-3-642-68601-6_5
  • E111. Jahner, E. (1982). Erhöhung des EDV-Leistungsangebots für das Finanz- und Rechnungswesen durch Endbenutzer-Sprachen. In Betriebs- und Wirtschaftsinformatik. Physica-Verlag. https://doi.org/10.1007/978-3-642-68704-4_13
  • E112. 佐藤修 (1994). エンドユーザ・コンピューティング管理段階測定尺度について [On a measurement scale for the management stages of end-user computing]. 『オフィス・オートメーション』, 15(3-4), 214–219. https://doi.org/10.20627/officeautomation.15.3-4_214
  • E113. 高原康彦, 柴直樹, 高木徹 (2002). Prolog をベース言語とするエンドユーザ開発のためのオブジェクト指向データベースの設計と実現 [Design and implementation of an object-oriented database for end-user development based on Prolog]. 『経営情報学会誌』, 11(1), 69–89. https://doi.org/10.11497/jjasmin.11.1_69
  • E114. 富澤浩樹, 明星聖子 (2009). 文学研究活動に着目した EUC/EUD に関する考察 [A study of EUC/EUD with a focus on literary research activity]. 『経営情報学会全国研究発表大会要旨集』, 2009f, 75. https://doi.org/10.11497/jasmin.2009f.0.75.0
  • E115. Minetaki, K. & Motohashi, K. (2009). Subcontracting structure and productivity in the Japanese software industry. The Review of Socionetwork Strategies, 3(2). https://doi.org/10.1007/s12626-009-0008-8 [unverified] (page range)
  • E116. 岡部建次, 永田大, 宮崎茂次 (2014). 表計算ソフト上で自由に動く自律エージェントシステムの汎用化とエンドユーザコンピューティング化 [Generalizing an autonomous agent system that runs freely on spreadsheet software and turning it into end-user computing]. 『日本経営システム学会誌』, 30(3), 209–220. https://cir.nii.ac.jp/crid/1520853832365463680 [unverified] (DOI)
  • E117. 中所武司 (2014). マッチングシステムを例題としたエンドユーザ主導開発方式に関する考察 [A study of an end-user-driven development method using a matching system as an example]. 『電子情報通信学会技術研究報告』, 114(292), 1–6. https://ndlsearch.ndl.go.jp/books/R000000004-I025980412
  • E118. 澤田浩之, 徳永仁史, 古川慈之 (2015). 高度な専門知識不要の IT システム開発ツール: MZ Platform [An IT system development tool requiring no advanced expertise: MZ Platform]. Synthesiology (English edition), 8(3), 147–157. https://doi.org/10.5571/syntheng.8.3_147
  • E119. 中所武司 (2016). エンドユーザ主導開発のためのドメイン特化型技術の適用性に関する考察 [A study of the applicability of domain-specific technologies to end-user-driven development]. 『電子情報通信学会技術研究報告』, 116(284), 25–30. https://ndlsearch.ndl.go.jp/books/R000000004-I027790857
  • E120. 神馬豊彦, 清水研三, 阿部慶太朗, 佐藤亙人 (2023). 業務担当者による電子契約システムを活用した支払先口座登録業務自動化システムの開発とその効果 [Development and effects of a system for automating payee account registration built by business staff using an electronic contract system]. 『大学 ICT 推進協議会年次大会論文集』, 2023, 497–503. https://doi.org/10.24669/axies.2023.0_497 [unverified] (whether peer-reviewed)
  • E121. 小西民恵, 高橋亨輔, 六車俊紀, 谷﨑勇太, 油谷知岐ほか (2024). 派生開発による建物修繕依頼システムの内製開発 [In-house development of a building repair request system through derivative development]. 『学術情報処理研究』, 28(1), 200–206. https://doi.org/10.24669/jacn.28.1_200 [unverified] (whether peer-reviewed)
  • E122. 花原稔 (2026). 生成 AI を用いたバイブコーディングは大学図書館業務の改善に寄与するか [Does vibe coding with generative AI contribute to improving university library operations?]. 『大学図書館研究』, 130. https://doi.org/10.20722/jcul.2211

Organizational and economic demands (D1)

Demands of democracy, participation, and labour (D2)

  • E39. Ehn, P. (1988). Work-Oriented Design of Computer Artifacts. Arbetslivscentrum (Lawrence Erlbaum edition, 1990). ISBN 978-9186158453
  • E40. Kensing, F. & Blomberg, J. (1998). Participatory design: Issues and concerns. Computer Supported Cooperative Work, 7(3–4). https://doi.org/10.1023/A:1008689307411
  • E41. Bratteteig, T. & Wagner, I. (2014). Disentangling Participation: Power and Decision-Making in Participatory Design. Springer. https://doi.org/10.1007/978-3-319-06163-4
  • E42. Le Dantec, C.A. & DiSalvo, C. (2013). Infrastructuring and the formation of publics in participatory design. Social Studies of Science, 43(2). https://doi.org/10.1177/0306312712471581
  • E43. Iversen, O.S., Smith, R.C. & Dindler, C. (2018). From computational thinking to computational empowerment: A 21st century PD agenda. PDC ‘18. https://doi.org/10.1145/3210586.3210592
  • E44. Dindler, C., Smith, R.C. & Iversen, O.S. (2020). Computational empowerment: Participatory design in education. CoDesign, 16(1). https://doi.org/10.1080/15710882.2020.1722173
  • E45. Fischer, G., Giaccardi, E., Ye, Y., Sutcliffe, A.G. & Mehandjiev, N. (2004). Meta-design: A manifesto for end-user development. Communications of the ACM, 47(9), 33–37. https://doi.org/10.1145/1015864.1015884

Educational and cultural demands (D3)

Critiques of democratization rhetoric (D4)

Labour process theory and the sociology of professions (D5)

Cognitive limits (L1)

Artifact quality (L2)

Organizational and governance limits (L3)

  • E72. Silic, M., Silic, D. & Kind-Trüller, K. (2025). From shadow IT to shadow AI: Threats, risks and opportunities for organizations. Strategic Change. https://doi.org/10.1002/jsc.2682
  • E73. Alamin, M.A.A., Uddin, G., Malakar, S., Afroz, S., Haider, T. & Iqbal, A. (2022). Developer discussion topics on the adoption and barriers of low code software development platforms. Empirical Software Engineering. https://doi.org/10.1007/s10664-022-10244-0

New limits in the LLM era (L4)

  • E74. Liu, M.X., Sarkar, A., Negreanu, C., Zorn, B., Williams, J., Toronto, N. & Gordon, A.D. (2023). “What it wants me to say”: Bridging the abstraction gap between end-user programmers and code-generating large language models. Proc. ACM CHI 2023. https://doi.org/10.1145/3544548.3580817
  • E75. Pearce, H., Ahmad, B., Tan, B., Dolan-Gavitt, B. & Karri, R. (2022). Asleep at the keyboard? Assessing the security of GitHub Copilot’s code contributions. IEEE S&P 2022. https://doi.org/10.1109/SP46214.2022.9833571
  • E76. Perry, N., Srivastava, M., Kumar, D. & Boneh, D. (2023). Do users write more insecure code with AI assistants? ACM CCS 2023. https://doi.org/10.1145/3576915.3623157
  • E77. Sarkar, A., Gordon, A.D., Negreanu, C., Poelitz, C., Ragavan, S.S. & Zorn, B. (2022). What is it like to program with artificial intelligence? PPIG 2022. https://arxiv.org/abs/2208.06213
  • E78. Vaithilingam, P., Zhang, T. & Glassman, E.L. (2022). Expectation vs. experience: Evaluating the usability of code generation tools powered by large language models. CHI 2022 Extended Abstracts. https://doi.org/10.1145/3491101.3519665
  • E79. Barke, S., James, M.B. & Polikarpova, N. (2023). Grounded Copilot: How programmers interact with code-generating models. PACMPL, 7(OOPSLA1), 85–111. https://doi.org/10.1145/3586030
  • E80. Prather, J. et al. (2024). “It’s weird that it knows what I want”: Usability and interactions with Copilot for novice programmers. ACM TOCHI, 31(1), 1–31. https://doi.org/10.1145/3617367
  • E81. Virk, Y. & Liu, D. (2025). Non-programmers assessing AI-generated code: A case study of business users analyzing data. IEEE VL-HCC 2025. https://doi.org/10.1109/vl-hcc65237.2025.00044
  • E82. O’Brien, G. (2025). Threats to scientific software from over-reliance on AI code assistants. Nature Computational Science, 5, 701–703. https://doi.org/10.1038/s43588-025-00845-2
  • E83. Meske, C., Hermanns, T., von der Weiden, E., Loser, K.-U. & Berger, T. (2025). Vibe coding as a reconfiguration of intent mediation in software development: Definition, implications, and research agenda. arXiv:2507.21928. https://arxiv.org/abs/2507.21928
  • E84. Fawzy, A., Tahir, A. & Blincoe, K. (2026). Vibe coding in practice: Motivations, challenges, and a future outlook. A grey literature review. ICSE-SEIP 2026. https://doi.org/10.1145/3786583.3786866
  • E85. Ge, Y. et al. (2025). A survey of vibe coding with large language models. arXiv:2510.12399. https://arxiv.org/abs/2510.12399

Application domain: accounting information systems and finance (L5a)

  • E123. Grossman, T.A., Mehrotra, V. & Özlük, Ö. (2007). Lessons from mission-critical spreadsheets. Communications of the AIS, 20, art.60. https://doi.org/10.17705/1cais.02060 [unverified] (number of cases)
  • E124. Powell, S.G., Baker, K.R. & Lawson, B. (2008). An auditing protocol for spreadsheet models. Information & Management, 45(5), 312–320. https://doi.org/10.1016/j.im.2008.03.004
  • E125. Caulkins, J.P., Morrison, E.L. & Weidemann, T. (2008). Spreadsheet errors and decision making: Evidence from field interviews. In End-User Computing. IGI Global, 856–874. https://doi.org/10.4018/978-1-59904-945-8.ch061 [unverified] (number of interviewees)
  • E126. Coster, N., Leon, L., Kalbers, L. & Abraham, D. (2011). Controls over spreadsheets for financial reporting in practice. Proc. EuSpRIG 2011. arXiv:1111.6887. https://arxiv.org/abs/1111.6887
  • E127. Ferreira, M.A. & Visser, J. (2012). Governance of spreadsheets through spreadsheet change reviews. Proc. EuSpRIG 2012. arXiv:1211.7100. https://arxiv.org/abs/1211.7100
  • E128. Leon, L., Przasnyski, Z.H. & Seal, K.C. (2015). Introducing a taxonomy for classifying qualitative spreadsheet errors. Journal of Organizational and End User Computing, 27(1), 33–56. https://doi.org/10.4018/joeuc.2015010102

Application domain: medical informatics (L5b)

  • E129. Koppel, R., Wetterneck, T., Telles, J.L. & Karsh, B.-T. (2008). Workarounds to barcode medication administration systems: Their occurrences, causes, and threats to patient safety. Journal of the American Medical Informatics Association, 15(4), 408–423. https://doi.org/10.1197/jamia.m2616
  • E130. Harris, P.A., Taylor, R., Thielke, R., Payne, J., Gonzalez, N. & Conde, J.G. (2009). Research electronic data capture (REDCap): A metadata-driven methodology and workflow process. Journal of Biomedical Informatics, 42(2), 377–381. https://doi.org/10.1016/j.jbi.2008.08.010
  • E131. Balka, E. (2010). Ghost charts and shadow records: Implication for system design. Studies in Health Technology and Informatics, 160(1), 686–690. https://doi.org/10.3233/978-1-60750-588-4-686
  • E132. Saleem, J.J., Flanagan, M., Militello, L.G., Arbuckle, N., Russ, A.L., Burgo-Black, A.L. & Doebbeling, B.N. (2011). Paper persistence and computer-based workarounds with the electronic health record in primary care. Proc. HFES Annual Meeting, 55(1). https://doi.org/10.1177/1071181311551136 [unverified] (number of sites and study period)
  • E133. Mörike, F., Spiehl, H.L. & Feufel, M.A. (2022/2024). Workarounds in the shadow system: An ethnographic study of requirements for documentation and cooperation in a clinical advisory center. Human Factors, 66(3), 636–646. https://doi.org/10.1177/00187208221087013
  • E134. Prakash, M.P. & Thiagalingam, A. (2024). The role of clinician-developed applications in promoting adherence to evidence-based guidelines: Pilot study. JMIR Cardio, 8, e55958. https://doi.org/10.2196/55958

Application domain: scientific computing and research software (L5c)

  • E135. Hannay, J.E., MacLeod, C., Singer, J., Langtangen, H.P., Pfahl, D. & Wilson, G. (2009). How do scientists develop and use scientific software? Proc. ICSE Workshop SECSE 2009. https://doi.org/10.1109/secse.2009.5069155 [unverified] (number of respondents)
  • E136. Prabhu, P. et al. (2011). A survey of the practice of computational science. Proc. SC11 (State of the Practice). https://doi.org/10.1145/2063348.2063374 [unverified] (methodological detail)
  • E137. Storer, T. (2017/2018). Bridging the chasm: A survey of software engineering practice in scientific programming. ACM Computing Surveys, 50(4), 1–32. https://doi.org/10.1145/3084225
  • E138. Pinto, G., Wiese, I. & Dias, L.F. (2018). How do scientists develop scientific software? An external replication. Proc. IEEE SANER 2018. https://doi.org/10.1109/saner.2018.8330263
  • E139. Wang, J., Kuo, T., Li, L. & Zeller, A. (2020). Assessing and restoring reproducibility of Jupyter notebooks. Proc. ASE 2020. https://doi.org/10.1145/3324884.3416585
  • E140. Pimentel, J.F., Murta, L., Braganholo, V. & Freire, J. (2021). Understanding and improving the quality and reproducibility of Jupyter notebooks. Empirical Software Engineering, 26, art.65. https://doi.org/10.1007/s10664-021-09961-9
  • E141. Cosden, I.A., McHenry, K. & Katz, D.S. (2022/2023). Research software engineers: Career entry points and training gaps. Computing in Science & Engineering, 24(6), 14–21. https://doi.org/10.1109/mcse.2023.3258630 [unverified] (methodological detail)
  • E142. Samuel, S. & Mietchen, D. (2024). Computational reproducibility of Jupyter notebooks from biomedical publications. GigaScience, 13, art.giad113. https://doi.org/10.1093/gigascience/giad113

(All access dates are 2026-08-13. Details of the reachability checks, and the list of URLs that could not be reached, are recorded in the Provenance section of the internal ledger source/review/end-user-development/papers.md.)

← All Notes · Home