2022 IEEE/ACM 44th International Conference on Software Engineering (ICSE)

Social Science Theories in Software Engineering Research

Tobias Lorey Paul Ralph Michael Felderer
Department of Computer Science, Faculty of Computer Science, Department of Computer Science,
University of Innsbruck Dalhousie University University of Innsbruck
Innsbruck, Austria Halifax, Canada Innsbruck, Austria
tobias.lorey@student.uibk.ac.at paulralph@dal.ca michael.felderer@uibk.ac.at

ABSTRACT incorporating theories and research methods from social science—

As software engineering research becomes more concerned with including psychology, sociology and management (especially infor-
the psychological, sociological and managerial aspects of software mation systems)—which study individuals, groups and organiza-
development, relevant theories from reference disciplines are in- tions.
creasingly important for understanding the field’s core phenomena On one hand, methods from social sciences (e.g. case studies,
of interest. However, the degree to which software engineering experiments, questionnaires), which investigate individuals, groups
research draws on relevant social sciences remains unclear. This or organizations, are now widely used in software engineering [82].
study therefore investigates the use of social science theories in five On the other hand, theories from social sciences—systems of ideas
influential software engineering journals over 13 years. It analyzes for explaining, describing, predicting or analyzing human phenom-
not only the extent of theory use but also what, how and where ena [26]—do not seem widely applied in SE research despite being
these theories are used. While 87 different theories are used, less critical to advance SE as a scientific discipline. One reason could be
than two percent of papers use a social science theory, most theories SE’s formal and technical roots [72], while social phenomena were
are used in only one paper, most social sciences are ignored, and the historically marginalized [30].
theories are rarely tested for applicability to software engineering However, researchers are increasingly recognizing that stronger
contexts. Ignoring relevant social science theories may (1) under- theoretical foundations are needed for SE to mature as an applied
mine the community’s ability to generate, elaborate and maintain a science [70]. Several initiatives for increasing attention to theory
cumulative body of knowledge; and (2) lead to oversimplified mod- in SE have been proposed. Sjoberg et al. [68] devised a framework
els of software engineering phenomena. More attention to theory is for describing theories and guidelines for proposing, employing,
needed for software engineering to mature as a scientific discipline. testing and modifying them. Hannay et al. [27] conducted a sys-
tematic literature review of theories used in software engineering
CCS CONCEPTS experiments. SE-specific guidelines have been created for theory
development [58, 73]. General theories of software engineering
• General and reference → Surveys and overviews; • Software
have been proposed [31, 56]. Stol & Fitzgerald [70] argued for train-
and its engineering;
ing PhD students in theorizing. Finally, a series of international
KEYWORDS workshops explored possibilities for general and core theories in
SE [21, 33, 60–62].
software engineering, theory, social science This paper extends these initiatives with a focus on theories
ACM Reference Format: from social sciences. It aims to increase attention to social science
Tobias Lorey, Paul Ralph, and Michael Felderer. 2022. Social Science Theo- theories and help SE researchers use social science theories more
ries in Software Engineering Research. In 44th International Conference on effectively. Increasing attention to social science theories will help
Software Engineering (ICSE ’22), May 21–29, 2022, Pittsburgh, PA, USA. ACM,
researchers to:
New York, NY, USA, 12 pages. https://doi.org/10.1145/3510003.3510076
(1) create, accumulate and preserve knowledge [62, 68, 70];
1 INTRODUCTION (2) cope with the multi-disciplinarity of SE research [71];
(3) improve the quality of their research [70]; and
The software engineering (SE) research community increasingly
(4) reduce the need for trial and error [32, 62].
accepts that software projects have important psychological, socio-
logical and managerial properties [59]. SE research has therefore More generally, “theory provides explanations and understand-
increasingly considered software development’s human and orga- ing in terms of basic concepts and underlying mechanisms, which
nizational aspects alongside its formal and technical aspects [31]. constitute an important counterpart to knowledge of passing trends”
Embracing the social aspects of software development involves [27]. Social science theories help us understand core SE phenomena
(e.g. programming) and their impacts on individuals, teams, projects
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed and organizations [68].
for profit or commercial advantage and that copies bear this notice and the full citation This paper therefore analyzes which, how, where, and to what
on the first page. Copyrights for components of this work owned by others than the extent social science theories are used in SE research. It provides a
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission complete list of the theories encountered, elaborating on the most
and/or a fee. Request permissions from permissions@acm.org. common. It explains how inattention to social science leads SE
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA researchers to oversimplify and over-rationalize core phenomena.
© 2022 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 978-1-4503-9221-1/22/05. . . $15.00 Finally, it highlights areas in which social science theories may be
https://doi.org/10.1145/3510003.3510076 especially relevant, but are presently neglected.

1994
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

This paper is structured as follows: Section 2 presents relevant include the technology acceptance model [17], cognitive fit
background and related work on theories in social sciences and SE, theory [81], and the theory of task-technology fit [25].
including an overview of the classification criteria employed by this (5) Theories for design and action provide detailed instructions on
paper. Sections 3 and 4 describe the research method and present how to do something, like recipes. These are less common in
the results. Section 5 elaborates on the theories used most often in the social sciences than in software engineering. Multiview,
SE research. Section 6 discusses the study’s implications, limitations the methodology and contingency framework, is a good
and avenues for future research. Finally, Section 7 concludes the example [3].
paper with a summary of its contributions.
Theories also exist on a kind of generalizeability spectrum from
2 BACKGROUND observations and empirical generalizations to middle-range theo-
This section explains our operationalization of social science and ries, general theories, grand theories, and theories of everything [61,
the classification schemes we later employ. 72]. SE researchers are mainly concerned with empirical generaliza-
We consider a theory a social science theory if it originates in one tions, middle-range theories and general theories [72]. An empirical
of the disciplines listed by The Social Science Encyclopedia [36]: an- generalization is “an isolated proposition summarizing observed
thropology, cognitive science, criminology and law, cultural studies, uniformities of relationships between two or more variables” [44];
demography, economics, education, evolution, gender, geography, for instance, Moore’s law. Middle-range theories (e.g. COCOMO [8];
health and medicine, history, industrial relations and management, the software engineering principles of Davis [16]) “lie between the
language, linguistics and semiotics, mental health, methods of so- minor but necessary working hypothesis that evolve in abundance
cial research, philosophy, political theory, politics and government, during day-to-day research and the all-inclusive systematic efforts
psychology, social problems and social welfare, and sociology. Some to develop a unified theory that will explain all the observed unifor-
notes on this list are warranted: mities” [44]. A general theory, meanwhile, “applies across [a] field
and unifies existing empirical and theoretical work” [33]. No gen-
• Some of these fields (e.g. geography) mix social, natural or
eral theory of software engineering is widely accepted, but a few
applied science.
have been proposed (e.g. the tarpit [31]; sensemaking-coevolution-
• Some fields intersect several broader social sciences (e.g.
implementation theory [55]).
communication studies intersects cultural studies, linguistics,
Researchers also distinguish between variance theories, process
management, and sociology).
theories, and taxonomies. Variance theories explain and predict
• Each field has many subfields (e.g. information systems,
the variance in one or more dependent variables using indepen-
which shares many similarities with SE, is a subfield of man-
dent, moderating and mediating variables [14]. Process theories
agement).
explain how an entity changes and develops [78]. Taxonomies (also
• We are not sure that philosophy is a social science, per se,
called theories for understanding, descriptive theories, frameworks
but tentatively include it anyway.
and typologies) organize instances (individual things) into classes
Theories can be classified on numerous dimensions including (abstract descriptions) [58].
what they are for, their scope and their structure, as well as the way Variance theories typically posit causal relationships among con-
they are used and where in an article they are used. structs [63]; process theories and taxonomies typically do not [58].
Gregor [26] proposed a taxonomy for classifying information Process theories usually focus on non-causal relationships between
systems theories according to their purpose. She argued that theo- entities (e.g. actors, objects) while taxonomies focus on similari-
ries can serve four purposes (analysis and description, explanation, ties between entities. Variance theories are often associated with
prediction, and prescription), leading to five types: positivist, quantitative, statistical research while process theories
(1) Theories for analyzing, also called frameworks or taxonomies, and taxonomies are often associated with interpretivist, qualitative,
are relatively simple, descriptive theories. A good example non-statistical research [45]. In many scientific disciplines, includ-
is Iivari et al.’s dynamic four-tiered framework for classify ing information systems, variance theories are more common than
information systems development methodologies and ap- process theories. However, process theories better explain the re-
proaches [28]. lationships between inputs and outcomes [14]. Variance theories
(2) Theories for explaining, also called theories for understanding, and process theories can sometimes be combined [48].
attempt to explain what, how, when, why or where phenom- Meanhile, theories can serve different purposes within a research
ena occur. They do not include causal hypotheses or make article [27]:
numerical predictions. Darwinian evolution is a theory for
explaining, as is dual coding theory [49]. • Design: The theory influences a study’s research method
(3) Theories for predicting predict future events without speci- including research questions, hypotheses, tasks, materials,
fying causal relationships (e.g. weather forecasting models). simulation parameters, or research models and frameworks.
These theories have testable propositions but no prescrip- • Explanation: The theory explains a study’s results. There
tions. may be a fine line between a theory that explains results and
(4) Theories for explaining and predicting posit causal explana- a theory that analyzes results.
tions and testable propositions (e.g. quantitative hypotheses). • Applied: The theory is used for its intended purpose (e.g.
Prescriptions may be included but are not the focus. This predicting technology adoption) within a study, or to analyze
type is common in information systems, where examples study results.

1995
Social Science Theories in Software Engineering Research ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA

• Motivation: The theory inspires the article’s topic (but its con- Methodology (TOSEM), Empirical Software Engineering (EMSE), the
cepts are not used explicitly to formulate the study design; Journal of Systems and Software (JSS), and Information and Soft-
that would be Design). ware Technology (IST). We chose these journals because they are
• Tested: The article evaluates the theory using primary data widely considered to be among the most prestigious, influential
(e.g. an experiment) or secondary data (e.g. a systematic and academically rigorous journals that sit squarely in SE.
literature review). To keep the review recent, tractable and reproducible, we limit
• Modified: The article incrementally modifies, adapts or ex- our search to articles published between 1 January 2007 and 31
tends the theory. December 2019. Within this thirteen-year period, the five selected
• Basis: The article proposes a new theory, with its own char- journals published more than 5,900 articles. Of these, we manually
acter and concepts, partly based on an existing theory. excluded editorials, introductions, book reviews, and other non-
Hannay et al. [27] include an eighth role—proposed—however, research articles.
theories proposed by SE articles are beyond the scope of this review We scanned the remaining research articles for theory use ac-
because they do not originate in a social science. We added the role cording to the following process:
applied to accommodate theories that are used for their intended (1) We performed an electronic search for the term “theory”
purpose. Furthermore, we follow Hannay et al. [27] in excluding within each entire article, which can serve as an initial indi-
tangential references to theories and descriptions of theories, as in cator for theory use.
a related work section, and focus on substantial uses of theories. (2) An electronic search for the names of the most commonly
Finally, we can classify theories according to where in an article used theories in information systems according to the Theo-
they are used. The IMRaD pattern—“Introduction, Method, Result, ries Used in IS Research Wiki1 or Dwivedi et al. [19]. This
and Discussion”—is the de-facto standard for scientific writing step was included to avoid missing theories that do not have
in many fields [6]. (In IMRaD, introduction includes related work the word “theory” in the title (e.g. Task Technology Fit) and
while discussion includes implications, limitations, future work and potentially explore similarities between SE and IS.
conclusions.) While each article is unique, most can be mapped to
the IMRaD structure. We can ignore front and back matter such as Each paper returned by these searches was then manually checked
the title, abstract, keywords, references, acknowledgements, and against the following inclusion criteria:
appendices as these sections rarely contain substantive uses of (1) the “theory” meets our definition of a collection of ideas for
theories. understanding, analyzing, explaining or predicting;
(2) the theory takes on one of Hannay et al.’s roles as adapted
3 METHODOLOGY for our study (Section 2);
We conducted a critical review. Critical reviews [e.g. 4, 73, 83] are (3) the theory originates in one of the social science subfields
similar in execution to systematic reviews [cf. 34], except that criti- listed in the Social Science Encyclopedia [36]. (If not obvious,
cal reviews analyze a sample of papers to make a point (often about we examined the reference list or searched for the theory in
theory or methods), whereas systematic reviews aim to synthesize relevant scientific databases).
all relevant evidence. We exclude theories that originated in software engineering,
Briefly, we identified clear and upfront analysis questions, se- mere observations, grand theories, theories of everything, and the-
lected a search strategy including explicit inclusion and exclusion ories that were mentioned but not substantively used (Fig. 1).
criteria, retrieved a sample of articles from prominent SE journals,
extracting instances of theory use from the articles, and analyzing
the theories and their uses by applying a prior coding scheme.
3.3 Data Extraction and Analysis
Next, we extracted relevant data from the articles. We extracted
3.1 Objective and Questions metadata including title, authors, journal, volume, issue (if applica-
ble) and publication. In addition to metadata, we extracted data on
This study aims to provide a comprehensive summary of the state
the theories and articles themselves. For theories, we extracted the
of social science theory use in software engineering research. We
originating discipline, theory role, theory usage, theory nature, and
refine this goal into five specific analysis questions (AQs) all of
Gregor’s theory type. For articles, we extracted the corresponding
which end with an implied ‘in SE research’:
SWEBOK classification.
(AQ1) To what extent are social science theories used? To gather this data, we analyzed each theory to determine its orig-
(AQ2) What social science theories are used? inating discipline, type (according to Gregor [26]), scope (middle-
(AQ3) Which social sciences provide the theories? range, etc.) and nature (process, variance or taxonomy). Then we
(AQ4) Which types of theories are used? analyzed each theory usage to determine the theory’s role and where
(AQ5) For which purposes are theories used? it was used (introduction, methods, results or discussion).
(AQ6) Where in the papers are theories used? Finally, we classified each article according to the fourteen knowl-
(AQ7) In what knowledge areas are theories used? edge areas defined by the software engineering body of knowledge
(SWEBOK) [10]—design, testing, etc.
3.2 Journal, Article and Theory Selection
We selected five SE journals to study: IEEE Transactions on Software
Engineering (TSE), ACM Transactions on Software Engineering and 1 https://is.theorizeit.org/

1996
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

3.4 Validation
We validated our analysis in two ways. First, the second and third
TSE TOSEM EMSE JSS IST authors performed a detailed audit of the first author’s classifica-
(839) (299) (710) (2,567) (1,515) tions. This review produced the edge case decisions described in
Section 3.3.1 and many improvements to the classification of theo-
ries. Next, we extracted the authors’ contact information from each
article that applied one or more theories and sent the authors a ques-
Total retrieved: 5,930 tionnaire regarding their article, the theory (or theories) they used
and their experience with social science theories. Questionnaire
results are reported in Section 4.8

Non-research articles excluded: 350 4 RESULTS

This section presents our results, organized by analysis question.
Below, we reference primary studies as [P...]. A comprehensive
replication package, including the complete list of primary studies,
Studies assessed for social science is available (see Data Availability).
theory-use: 5,580
4.1 To what extent are social science theories
used in SE research?
Studies excluded: 5,478 Only 102 (1.8%) of the 5580 included articles used a social science
theory in some substantial way. Only 24 used more than one theory.
For example, Moody [P1] justifies scientific foundations for creat-
ing visual notations using six theories: the Shannon-Weaver model,
cognitive fit, cognitive load, dual coding, feature integration theory,
Total primary studies included: 102 and the theory of symbols. Meanwhile, Prechelt and Oezbek [P2]
used the garbage can model, actor-network theory, structuration
theory, and path dependence to study open source software process
Figure 1: PRISMA Diagram
innovation. More recently, Menolli et al. [P3] employed concepts
from organizational learning, the theory of communities of practice,
and Nonaka’s dynamic theory of organizational knowledge cre-
3.3.1 Edge Cases. This paper focuses on theories imported from ation to knowledge sharing and learning in software development
social sciences. We therefore exclude theories from mathematics companies.
(e.g. chaos theory), natural sciences, (e.g. the theory of evolution)
and theories homegrown in SE (e.g. theory-w [9], sensemaking-
coevolution-implementation theory [55], the theory of distances in 4.2 What social science theories are used in SE
SE [7]). Of course these other theories are also important—they are research?
just not the focus of this study. The primary studies substantially used 87 unique theories. Of these,
We also excluded SERVQUAL [50] because although it is of- 27 (31%) appear in two or more articles (Table 1) while 60 are
ten presented as a theory, it is clearly an instrument [75]. We do, used in a single article (Table 2). The most commonly used theory
however, include expectation disconfirmation theory, from which is the technology acceptance model (TAM) [17]. Tied for second
SERVQUAL is derived.2 place, with five uses each, are dual coding theory and diffusion of
Three edge cases we did include are coordination theory, real innovations.
options theory and argumentation theory. Coordination theory
and real options theory could be seen as a bodies of research (like
game theory). However, we included coordination theory because, 4.3 Which social sciences provide the theories
in some usages, it clearly adapts sociological concepts to software used in SE research?
engineering. Meanwhile, we conceptualize real options theory as a The 87 theories used originate from eight social sciences: psychol-
theory for analyzing (Gregor’s Type I) because it is used to analyze ogy (34), industrial relations / management including information
investment decisions. Toulmin’s [77] argumentation theory, in con- systems and organizational studies (23), economics (10), communi-
trast, is sometimes seen as a method; however, we again consider cation studies (8), sociology (8), education (2), linguistics / semiotics
it a Type I theory because it facilitates analyzing the structure of (1), and philosophy (1). (Communication studies was not included
an argument. in the Social Science Encyclopedia because it is an “intersecting
discipline”—see Section 2— but the eight theories in question are
2 SERVQUAL is typically capitalized despite being a portmanteau of ‘service’ and best understood as originating in the interdisciplinary communica-
‘quality’, not an acronym tion studies community.)

1997
Social Science Theories in Software Engineering Research ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA

Table 1: Overview of theories used in more than one article

Theory (T.) Ref Description From1 n Articles

Technology Acceptance [17] posits that usefulness and ease-of-use cause intention to adopt MIS 13 [P4-P15, P20]
Model
Diffusion of Innovations T. [65] explains how ideas and products (innovations) spread over time Comm. 5 [P16-P20]
Dual Coding T. [49] posits that verbal and nonverbal stimuli are processed by different Psyc. 5 [P1, P21-P24]
cognitive subsystems
Coordination T. [41] a body of principles about how the activities of separate actors can Org. 4 [P25-28]
be coordinated
T. of Reasoned Action [23] posits that attitude and subjective norm lead to behavioral intention Psyc. 4 [P20, P39-P41]
and actual behavior
Cognitive Fit T. [81] posits that problem solving performance depends on the alignment MIS 3 [P1, P23, P29]
between problem representation and tasks
Cognitive T. of [43] explains how multimedia representations are processed, remembered Psyc. 3 [P22, P30, P31]
Multimedia Learning and recalled
Organizational Learning T. [2] explains the processes by which organizations create knowledge, Mgmt 3 [P3, P27, P32]
including single- and double-loop learning
Real Options T. [46] analyzes the value of potential investments by modeling them as Econ. 3 [P33-35]
options (like stock options)
Media-Richness T. [15] posits that richer communication mediums improve communication Comm. 3 [P36-P38]
effectiveness
Cognitive Load T. [74] explains how cognitive load affects learning Psyc. 3 [P1, P42, P43]
Task-Technology Fit [25] posits that the alignment between technologies and tasks improves MIS 3 [P7, P37, P49]
performance and utilization
Organizational Knowledge [47] explains how an organization’s knowledge is created by a dialogue Org. 2 [P3, P44]
Creation between tacit and explicit knowledge
Structuration T. [24] understands human behavior in terms of interactions between struc- Soc. 2 [P2, P45]
ture and agents
Control T. [35] explains how control is carried out in organizations to make employ- Org. 2 [P46, P47]
ees adhere to certain standards
Agency T. [29] explains the relationship between agents and principals where agents Org. 2 [P46, P48]
make decisions on behalf on principals
Information Foraging T. [52] explains human information search mechanisms and posits that Psyc. 2 [P50, P51]
people have built-in search mechanisms
Goal-Setting T. [39] explains how goals affect task performance; posits that motivating Psyc. 2 [P52, P53]
goals improve performance
Contingency T. [22] posits that organizational outcomes depend on the alignment be- Org. 2 [P54, P55]
tween the leadership style and the situation.
T. of Argumentation [77] describes how humans argue and how arguments are structured Comm. 2 [P56, P57]
Activity T. [20] describes activities of subjects in relation to the community and Psyc. 2 [P58, P59]
other constructs
Transaction Cost Econ. [13] explains how costs occur when transactions are performed within a Econ. 2 [P60, P61]
market
UTAUT [79] explains why organizations adopt technologies MIS 2 [P62, P63]
Human Error T. [64] describes a four stage classification of how humans produce errors Psyc. 2 [P64, P65]
Contact Hypothesis [1] explains mechanisms of group contact that reduce prejudice Psyc. 2 [P66, P67]
Resource Dependency T. [51] studies the relationship between an organization and its resources Soc. 2 [P61, P53]
Media Synchronicity T. [18] examines media in the context of synchronicity MIS 2 [P37, P72]

1 Comm.=Communication Studies; Econ.=Economics; MIS=Management Information Systems; Mgmt=Management; Org.=Organizational

Studies; Psyc.=Psychology; Soc.=Sociology

1998
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

Table 2: List of theories used in one article

Activity t. Cultural lag t. Personal construct t. Subjective t. of value

Actor-network t. Cumulative prospect t. Power dependency t. Task-media fit
Adaptive structuration t. Dickinson and Mcintyre’s Prejudice t. The integrated model of group development
Bases of power teamwork model Rhetorical t. of diffusion T. of affordances
Bounded rationality Dual-process t. of cognition Shannon-Weaver model T. of argumentation
Cognitive apprenticeship t. Dynamic capability t. Shared mental models t. T. of group productivity
Cognitive-based view Expectancy t. Situated cognition t. T. of intellectual capital
Cognitive model of media Expectation disconfirmation t. Social capital t. T. of planned behavior
choice Experiential learning t. Social cognitive t. T. of power and conflict
Common ground t. Feature integration t. Social education t. T. of symbols
Common teamwork t. Garbage can model Social identity t. T. of technology diffusion and org. learning
Communities of practice t. Herzberg’s two factor t. Social interdependence t. T. of work performance
Competitive strategy Institutional t. Social motivation t. Time, interaction, and performance t.
Construal level t. IS success model Social presence t. Trust t.
Constructive controversy t. Knowledge-based t. of the firm Socio-technical model Value network t.
Contextual integrity t. Path dependence Speech-act t. Whistleblowing t.

4.4 What types of theories are used?

Of the 87 theories, 75 (87%) are middle-range theories (e.g. TAM).
The remaining twelve are general theories (e.g. structuration the-
ory). Process theories (39) were more common than variance the-
ories (36) or taxonomies (12). We did not identify any empirical
generalizations (see Section 6.2: Limitations).
In terms of Gregor’s theory types, the most common type was
Type II: theories for explaining (38), followed by Type IV: theories
for explaining and predicting (35), and Type I: theories for ana-
lyzing (13). These numbers track well with those in the previous
paragraph because most process theories explain, most variance
theories explain and predict, and most taxonomies analyze. Again,
this shows that theories for explaining, but not predicting, are more
common in SE than generally assumed [cf. 58].
One Type V (for design and action) theory and no Type III (for Figure 2: Distribution of theory roles
predicting) were identified. This is not surprising since Types III
and V appear less common in social science [26].
Theories are less often tested, or used to analyze results, which
4.5 For which purposes are theories used? happens more in the results and discussion sections.
One might expect articles that test a social science theory to use
We identified 198 individual theory uses (Fig. 2). Some articles use
that theory throughout, or at least in multiple sections. Rather, we
multiple theories or use one theory in multiple ways. The most
found that the articles that tested a theory did so tangentially to
common roles are design (53%) and explanation (16%). That is, SE
their core purpose, such that the tested theory was mainly used in
research tends to use theories to inform the methodological design
one section of the paper.
of a study or explain and analyze the results. Theories are used to
design research protocols and task materials; to formulate hypothe-
ses, research questions and design frameworks or models; and to 4.7 In what knowledge areas are theories used?
develop questionnaires and other instruments. We attempted to map the topic of each article into the SWEBOK
knowledge areas. The most common knowledge areas are profes-
sional practice (24) and models and methods (13). For example,
4.6 Where in the papers are theories used? Babar et al.’s [P68] study of software architecture evaluation in
Theories are mostly used in method (40%) and introduction (26%) distributed versus co-located meetings exemplifies the professional
sections; less often in results (19%) or discussion (15%) sections. practice area. Meanwhile, Yu and Petter’s [P69] use of shared mental
This distribution roughly corresponds to the purposes for which models theory to understand agile software development processes
the theories are used. That is, since theories are often used to design is a good example of the models and methods area. Many theories
a study, motivate the research, or inform further theorizing, they were also used in the context of software requirements (12), eco-
tend to be used in the introduction and method parts of a paper. nomics (9), and SE management (7). Theories were less used in the

1999
Social Science Theories in Software Engineering Research ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA

context of construction (4), SE process (4), testing (3), design (3), • Design/method (29%) and design/introduction (23%) are the
maintenance (3), and quality (1). No articles used theories in the most common combinations. That is, the theory informs
context of other knowledge areas. the design of the study, and this is explained in either the
However, 19 articles did not map well into the SWEBOK knowl- introduction or the method section. For example, Babar et
edge areas. Of these, twelve can be best categorized as technology al. [P68] formulate their hypothesis based on task-media fit
adoption (e.g. investigating technology adoption and acceptance in the methodology section. Marsan et al. extend concepts
in Botswana [P16]). Seven are about research methods (e.g. us- of institutional theory and rhetorical theory to propose a
ing the garbage can model to find a research method for studying research question in their introduction [100].
open-source software [P2]). • Explanation/discussion is also common (11%). For example,
Naturally, social science theories are used more in the primarily Hyrynsalmi et al. [102] use cultural lag theory to explain
social knowledge areas than in the primarily technical knowledge value creation mechanisms for mobile applications in their
areas. The more technical knowledge areas may use more mathe- discussion section.
matical and natural-science foundations, or may be less theoretical • Motivation/introduction, applied/method, and basis/method
than the more social knowledge areas. That social theories are used each make up another approximately 3%. The first is self
so rarely in software design is puzzling. explanatory. As an example of the second, Yu and Petter
Perhaps more concerning is that the SWEBOK knowledge area [P69] used shared-mental models theory to analyze different
taxonomy appears deficient for classifying either SE research or SE kinds of agile practices in their methods section. Moody [P1]
practice. Much of the research we reviewed does not fit in any of the introduces a design theory on the basis of various existing
categories, and many common practices resist classification. How theories, as an example for the third.
would one classify, for example, research on continuous integration The number of articles contributed by each journal varies greatly.
practices? TOSEM and ESE contributed only 274 and 642 papers respectively,
while IST and JSS contributed 1419 and 2453. TSE was in the middle
4.8 Member checking results with 792. This basically reflects publication numbers—JSS and IST
Approximately 20% of the authors completed our questionnaire, publish a lot more articles.
leading to minor changes to the classification of some theories, However, even after accounting for their relative contribution
mainly in how theories are used within the articles. A sample issue to the sample, journals vary significantly in their use of theories.
that was discussed upon analyzing the questionnaire was the origin Less than 1% of the TOSEM papers used theories, followed by TSE
of Adaptive Structuration Theory (MIS) compared to Structuration with 1.51%, JSS with 1.59%, ESE with 1.87%, and 2.60% IST papers.
Theory (sociology). Benefits of using social science theories in SE This suggests that either IST is most sympathetic to research using
research mentioned by respondents include: social science theories, or that researchers who use social science
• improving understanding of phenomena occurring in SE; theories prefer submitting to IST.
• theoretical grounding of their own research; and We might expect more articles using social science theories in
• connecting technical and human aspects such as the way of recent years, as SE has become more concerned with the human
thinking, teamwork, and failure. and social aspects of development. However, no particular trends
are evident in the data overall, or within any specific journal.
Challenges mentioned include:
• understanding the theoretical background as social science 5 MOST USED THEORIES
theories are often more abstract; This section describes in more detail some of the most used theories
• finding corresponding theories and mapping them to SE; in our sample. While space limitations prevent comprehensive
• theories that seem useful but lack necessary precision. descriptions of all 87 theories, elaborating a few prevalent ones
Some authors are sympathetic to theories, but find difficulty should give the reader a better sense of the sort of social science
in publishing due to reviewers not understanding social science theories being used in SE research.
theories. One response mentioned the psychological uncertainty
when using qualitative constructs and the challenge of accepting 5.1 The Technology Acceptance Model
that “everyone does it differently”. The most used theory in this review is the technology acceptance
model (TAM). TAM is a middle-range variance theory for explaining
4.9 Exploratory findings and predicting organizational adoption of new technologies [17].
Several findings beyond the original analysis questions emerged. TAM posits that an organization’s intention to adopt a technology
This section briefly reports the more interesting incidental findings. depends on perceived usefulness and perceived ease of use. It is
Software engineering is closely related to information systems, based on Fishbein and Ajzen’s theory of reasoned action [23].
so we might expect a large overlap in theory use. However, only 28 TAM is arguably the most widely used theory in information
(32%) of the theories used in our sample are listed in the Theories system research. It has been extensively tested, refined and ex-
Used in IS Research Wiki or Dwivedi et al.’s [19] survey of theories tended, most notably by the Unified Theory of Acceptance and Use
commonly used in IS research. of Technology (UTAUT; Fig. 3) [80].
Meanwhile, cross-referencing where theories are used with the However, TAM has also been widely criticized [12]. For example,
role the theory plays highlights a few combinations: Benbasat and Barki [5] argue that TAM “has created an illusion

2000
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

Performance
Expectancy

E
ort Behavioural
Expectancy Intention
Use
Behavior
Social
Inuence

Facilitating
Conditions

Voluntariness
Gender Age Experience
of Use

Figure 3: Unified theory of acceptance and use of technology

[80]

Figure 4: Diffusion of innovations (adapted from [65])

of progress in knowledge accumulation” and attempts to extend
it have created “a state of theoretical chaos”. More generally, no
one ever argued that software should be useless and difficult to
use. Pointing out that usefulness and usability are important is not (5) Confirmation. The individual decides to retain or abandon
terribly helpful without some theory of what makes a software the innovation. This stage can last indefinitely.
product useful and usable. Diffusion of innovations further posits five types of adopters.
In twelve out of thirteen articles, TAM is used in the design role. From quickest to slowest to adopt, they are: innovators, early
For instance, Mohagheghi et al. [P4] employ TAM to construct an adopters, the early majority, the late majority, and laggards (Fig. 4).
interview and survey about model-driven engineering. Diffusion of innovations is not without its critics. For example,
Lyytinen and Damsgaard [40] present six criticisms:
5.2 Diffusion of Innovations
(1) “Technologies are not discrete packages.”
Diffusion of innovations is a middle-range process theory for ex- (2) “Technologies do not diffuse in a homogenous and fixed
plaining how innovations spread through communication channels social ether.”
over time within a social system [65]. It models diffusion as a kind (3) Push and pull forces do not fully explain diffusion rates.
of communication in which a new idea or product is shared among (4) The theory does not include all of the parameters that influ-
individuals in a social system. ence adoption decisions.
The main elements of this theory are innovation, communication (5) “Diffusion does not necessarily traverse through distinct
channels, time, and a social system. An innovation is simply an stages, which exhibit no feedback.”
idea or product that is perceived as new by potential adopters. (6) “Time scales are not necessarily short and the history of
Communication is the process in which the participants in the social decisions is not unimportant.”
system share information to create consensus. Communication
channels (including mass media and interpersonal communication) Diffusion of innovations is used in five papers and is therefore
are the means by which an innovation is communicated among tied (with dual coding theory; next) for second most common the-
members of a social system. Meanwhile, “a social system is defined ory.
as a set of interrelated units that are engaged in joint problem
solving to accomplish a common goal. The members or units of a 5.3 Dual Coding Theory
social system may be individuals, informal groups, organizations, Dual Coding Theory is an influential general theory for explaining
and/or subsystems” [65]. memory and cognition [49]. It posits that people have two cognitive
The theory posits that innovations diffuse in five stages [65]: subsystems: one for words and one for images. Our minds therefore
(1) Knowledge. An individual learns about an innovation. use two main kinds of representations or “codes:” analogue codes
(2) Persuasion. The individual forms an opinion (positive or (images) and symbolic codes (words). The theory aims to explain
negative) about the innovation. how these two subsystems interact and affect cognition. Therefore,
(3) Decision. The individual decides whether to adopt or reject presenting the same idea using both words and images should help
the innovation. This may include trying a new product or people remember, because the idea is encoded in both cognitive
attending demonstrations. subsystems.
(4) Implementation. The individual uses the innovation and However, dual coding theory remains controversial. Many psy-
judges its usefulness. If an individual decides to adopt an chologists prefer the common coding theory of memory, which
innovation, it is put to use during the implementation stage. posits a single kind of representation and cognitive system [76].

2001
Social Science Theories in Software Engineering Research ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA

June 2021 issue of MIS Quarterly appear to use one or more social
Internal Mental
Problem Solving theories. At least three possible explanations are evident:
Representation of Representation
Performance
Problem Domain for Task Solution (1) SE research primarily uses homegrown theories;
(2) SE research primarily uses natural science theories;
(3) SE research is primarily atheoretical.
External Problem Problem-Solving Distinguishing between these three explanations would require a
Representation Task different kind of study, with a different sampling strategy. However,
based on our experience with initiatives aimed at encouraging
theory development and use, we suspect SE research is largely less
Figure 5: Cognitive fit theory (adapted from [67]) concerned with theory (more atheoretical) than other similar fields.
Compared to management, for example, SE researchers appear
more concerned with making tools than theories.
Additionally, some of the theories that are being used are prob-
5.4 Coordination Theory lematic. TAM, for example, is outdated [80], does not explain a core
Coordination theory is a general theory for explaining “how the SE phenomenon, and has been widely criticized [12].
activities of separate actors can be coordinated” [41]. It includes a Broadly considered, the results of this study suggest that SE
framework for analyzing coordination in complex processes, as well research is not making good use of social science theory. Failing to
as a typology of dependencies and coordination mechanisms [42]. incorporate social theories is problematic in several ways.
It posits that in an organization, some tasks depend on other tasks, Theories underpin almost all natural, social and applied sciences.
which creates coordination problems. Actors solve these problems Sound theoretical foundations are essential for maturing into a
using coordination mechanisms. Coordination theory posits different fully developed research discipline [70]. Without a collection of
kinds of coordination problems and mechanisms. core theories, a discipline struggles to generate, accumulate and
Coordination theory is used to design not only human organiza- preserve knowledge [62, 68, 70].
tions but also technologies (e.g. tools for facilitating coordination, Second, SE is intrinsically interdisciplinary—it intersects com-
distributed computer systems) [41]. puter science, economics, management, mathematics, philosophy,
psychology, and sociology. The social and technical aspects of cre-
5.5 Cognitive Fit Theory ating software are inextricably entwined; they cannot be cleaved
Cognitive fit is a middle range, variance theory for explaining and apart and understood independently. More than two percent of SE
predicting task performance. It “describes the relationships between research needs to draw on social science theories to have any hope
graphical and tabular representations and the types of tasks they of understanding the full scope of software development.
support” [81]. It posits that matching information representations In our experience, this inattention to social theory already harms
to problem solving tasks will increase task performance (Fig. 5). SE research, practice, and education. Some SE research tends to
For example, to navigate, we might want a map, but to calculate oversimplify and over-rationalize social phenomena. For exam-
distance traveled, we might want a table showing distances between ple, where an SE researcher asks “what is the problem the system
destinations. should solve?” a sociologist sees multiple stakeholders, who do
Cognitive fit theory has evolved over time. For example, Samuel not agree on the nature of the problem or the aims of the system
et al. [66] extend cognitive fit theory to introduce the dual-domain [11], jockeying for control of a project’s agenda—there is no “the
problem solving framework. While uncontroversial, cognitive fit is problem”. Where a software professional sees “requirements elicita-
only one of multitudinous antecedents of task performance, and tion”, a psychologist sees a prospective user and an analyst analyst
may be overwhelmed by other, more important factors. co-constructing preferences specific to that moment and context
[37]—the outcomes are neither “requirements” nor “ellicited” [54].
6 DISCUSSION Moreover, while students of other applied sciences learn to design
innovative systems based on ill-defined opportunities, SE students
6.1 Implications primarily learn to construct routine, fully-specified systems, as if
Our analysis produced several surprising findings: figuring out what to build is someone else’s job [cf. 53].
(1) Less than 2% of the articles use a social science theory in a Furthermore, infrequent use of social theory is surprising given
substantive way. two trends in SE research:
(2) SE research rarely tests a social science theory, for example, (1) The SE academic community is midway through a shift from
to see if it actually applies to SE. rationalism (it works because it intuitively makes sense) to
(3) Many relevant social sciences are not referenced at all (e.g. empiricism (it works because the results of this empirical
criminology). study demonstrate that it works) [57].
(4) Process theories are used more often than variance theories, (2) The SE community appears increasingly receptive to soci-
which shows that process theories are much more common ological research methods including grounded theory [73]
in SE than generally assumed [cf. 58]. and ethnography [83].
Intuitively, less than two percent of papers using social theories As the community demands more rigorous research, demands
seems low. For comparison, at least two thirds of the papers in the more empirical studies, and attempts more qualitative and social

2002
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

research, we expect researchers increasingly to draw on reference Additionally, SE researchers lean toward theories from psychol-
disciplines for both supporting foundations, evidence and method- ogy (to understand professionals) and management (to understand
ological guidance. teams and organizations). We rarely draw on education (to un-
derstand professional development), criminology (to understand
deviant behavior), health (to understand professionals’ wellbeing),
6.2 Limitations history (to understand the evolution of SE), or philosophy (to un-
The results of this study should be weighed against several limita- derstand a great many things). Exploiting the vast array of relevant
tions. It is very likely that some articles in the sample were missed knowledge from myriad disciplines may help launch many fruitful
because they used a theory, but not the word “theory.” We mitigated research programs.
this threat by searching for a list of specific theories by name, but More specifically, this study investigates use of social science
that list too is incomplete. theories. Similar reviews of homegrown SE theories and natural
Similarly, our search process may not have been sensitive to em- science theories used in SE are needed.
pirical generalizations. These generalizations would not be listed in
the IS Theory Wiki or or Dwivedi et al.’s [19] review, and would not 6.4 Comparison to previous reviews
normally have “theory” in their names, if they even have names. Several studies have examined the status of theory use in SE re-
Therefore, the fact that we did not uncover any empirical general- search. Hannay et al. [27] present a systematic literature review of
izations from social science does not mean none were used. theories used in software engineering experiments. Stol & Fitzger-
The five journals and 13-year period we studied are not rep- ald [70] argue that SE researchers often use “theory fragments” for
resentative of all SE outlets across all time. We used purposive, analysis. Lim et al. [38] examine the theoretical basis of information
search-based sampling [4] to select good, influential journals rather systems research. Ralph et al. [59] describe seven theories that, they
than random journals and focused on more recent years because argue, ought to be core to SE research.
we want to understand the trajectory of the field, which is mostly Like Hannay et al., we find that theories are mostly used to moti-
set by recent papers in top outlets. vate and design studies; fewer articles test or modify theories. Like
Meanwhile, researchers do not agree on the definition of theory. Lim et al., we find that TAM is widely-used, and that psychology
Different researchers might have included or excluded different and economics are popular reference disciplines. Like Ralph et al.,
theories. We mitigate this threat by providing a complete list of the we argue that SE researchers should use more social theory.
included theories (Section 4.2) and discussing edge cases (Section However, this review extends previous work in several ways.
3.3.1). We also tried searching for “model”, but this produced too Stol & Fitzgerald [70] is more of a position paper with a limited
many false positives. review. Hannay et al. [27] only consider experiments, and their
We validated our classifications through audits (i.e. the second review is getting dated. Lim et al. [38] looked at a different field.
and third authors extensively reviewing the classifications) and Ralph et al. [59] recommend social theories rather than studying
member checking (i.e., sending a questionnaire to primary study which are currently used. In summary, we extend these reviews by
authors). This led to significant recategorization of Gregor’s theory studying a larger, broader, more recent sample of primary studies,
types and a handful of minor changes. However, different analysts leading to somewhat different recommendations.
may classify theories, usages and topic areas differently.
7 CONCLUSION
6.3 Areas for future research The presented study examined substantive uses of social science
SE researchers should consider using more social theories from theories in more than 5500 articles published in five leading soft-
reference disciplines. Using social theories more often, and more ef- ware engineering journals over a thirteen-year period. Despite
fectively, will help SE researchers produce more rigorous, nuanced, several calls for more attention to theory in SE research, increasing
insightful research. Furthermore, SE researchers should move be- attention to the social aspects of development, and increasing use
yond using theories to motivate their papers or specify the research of behavioral and qualitative research, only about two percent of the
methodology. We need to test social theories in SE contexts and reviewed articles use social science theories in any substantial way.
theorize about SE phenomena using concepts from the social sci- In other words, SE research tends to ignore relevant theories from
ences. For example, iteratively evaluating and adapting the theory social sciences.
of boundary objects [69] could help us better understand product When social science theories are used, they are most commonly:
backlogs, user stories, and other common non-code artifacts. • middle-range theories;
Similarly, social theories are often used to examine project man- • process theories;
agement, requirements engineering, modeling, software develop- • theories for explaining (but not predicting);
ment methods and technology adoption. SE researchers tend not • drawn from psychology or management;
to use them to examine designing, constructing, testing, evaluating • used in some way related to designing the study;
and configuring software. However, the latter are socio-technical • used in the introduction or methods section of the paper;
processes in which technical phenomena are entangled with social • used in research concerning professional practice, technol-
phenomena. Again, social theories should help. For example, nu- ogy adoption or models and methods of software develop-
merous theories concerning power and influence could help explain ment; and
interpersonal conflict during peer programming. • published in Information and Software Technology.

2003
Social Science Theories in Software Engineering Research ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA

In contrast, SE research tends not to: [13] Ronald H Coase. 1937. The nature of the firm. Economica 4 (1937).
[14] Kevin Crowston. 2000. Process as theory in information systems research. In
• test theories being used; Organizational and Social Perspectives on Information Technology. Springer, 149–
• use theories to analyze data; 164.
• apply theories to software design, testing, evaluation or con- [15] Richard L Daft and Robert H Lengel. 1983. Information richness. A new approach
to managerial behavior and organization design. Technical Report. Texas A and
figuration; M Univ College Station Coll of Business Administration.
• draw theories from relevant reference disciplines including [16] Alan M Davis. 1995. 201 Principles of Software Development. McGraw-Hill, Inc.
[17] Fred D Davis. 1989. Perceived usefulness, perceived ease of use, and user accep-
health, criminology or history. tance of information technology. MIS Quarterly (1989), 319–340.
Applied sciences benefit from balancing more concrete tech- [18] Alan R Dennis, Robert M Fuller, and Joseph S Valacich. 2008. Media, tasks, and
communication processes: A theory of media synchronicity. MIS Quarterly 32, 3
niques and more abstract theories. Techniques encapsulate practical, (2008), 575–600.
specific, contextualized, often short-term knowledge that practi- [19] Yogesh K Dwivedi, Michael R Wade, and Scott L Schneberger. 2011. Information
tioners can readily apply. Theories encapsulate more abstract, gen- systems theory: Explaining and predicting our digital society. Vol. 1. Springer.
[20] Yrjö Engeström et al. 1999. Activity theory and individual and social transforma-
eralizable, long-term knowledge. They insulate a field against fads, tion. Perspectives on Activity Theory (1999), 19–38.
guide research programs and help us understand novel situations. [21] Iaakov Exman, Dewayne E. Perry, Balbir Barn, and Paul Ralph. 2016. Separability
For SE research to mature and improve, researchers therefore principles for a general theory of software engineering: Report on the GTSE
2015 workshop. SIGSOFT Software Engineering Notes 41, 1 (Feb. 2016), 25–27.
need to make better use of theory: social science theories to un- https://doi.org/10.1145/2853073.2853093
derstand social phenomena, natural science theories to understand [22] Fred E Fiedler. 1964. A contingency model of leadership effectiveness1. In
Advances in Experimental Social Psychology. Vol. 1. Elsevier, 149–190.
physical and mathematical phenomena, and more homegrown the- [23] M. Fishbein and I. Ajzen. 1975. Belief, attitude, intention, and behavior: an intro-
ories to understand SE-specific phenomena. duction to theory and research. Addison-Wesley Pub. Co., Reading, MA, USA.
No one is saying that SE should abandon developing and evalu- [24] Anthony Giddens. 1986. The Constitution of Society: Outline of the Theory of
Structuration. Vol. 349. Univ of California Press.
ating new technologies or techniques. Rather, SE has twin aims of [25] Dale L Goodhue and Ronald L Thompson. 1995. Task-technology fit and individ-
understanding software development and creating tools and tech- ual performance. MIS Quarterly 19, 2 (1995), 213–236.
niques for developing software better. Our abilities to build better [26] Shirley Gregor. 2006. The nature of theory in information systems. MIS Quarterly
30, 3 (2006), 611–642.
systems and to explain software development phenomena necessar- [27] Jo Hannay, Dag Sjøberg, and Tore Dyba. 2007. A systematic review of theory use
ily co-evolve. Many of our hardest problems lie at the intersection in software engineering experiments. IEEE Transactions on Software Engineering
33, 2 (Feb. 2007), 87–107. https://doi.org/10.1109/TSE.2007.12
of social phenomena, technical phenomena, understanding, and cre- [28] Juhani Iivari, Rudy Hirschheim, and Heinz K Klein. 2000. A dynamic framework
ating. Only by accepting the fundamental interconnections among for classifying information systems development methodologies and approaches.
these four dimensions can we address our greatest challenges. Journal of Management Information Systems 17, 3 (2000), 179–218.
[29] Michael C Jensen and William H Meckling. 1976. Theory of the firm: Managerial
behavior, agency costs and ownership structure. Journal of Financial Economics
DATA AVAILABILITY 3, 4 (1976), 305–360.
[30] Michael John, Frank Maurer, and Bjørnar Tessem. 2005. Human and social factors
The anonymized replication package includes the questionnaire of software engineering: workshop summary. ACM SIGSOFT Software Engineering
template, the list of primary studies, and a spreadsheet containing Notes 30, 4 (2005), 1–6.
all extracted data on articles and theories being used, as well as [31] Pontus Johnson and Mathias Ekstedt. 2016. The Tarpit–A general theory of
software engineering. Information and Software Technology 70 (2016), 181–203.
decision rules for inclusion and exclusion criteria. It is available at [32] Pontus Johnson, Mathias Ekstedt, and Ivar Jacobson. 2012. Where’s the theory
https://doi.org/10.5281/zenodo.6036076. for software engineering? IEEE Software 29, 5 (2012), 94–96. https://doi.org/10.
1109/MS.2012.127
[33] Pontus Johnson, Paul Ralph, Michael Goedicke, Pan-Wei Ng, Klaas-Jan Stol, K.
REFERENCES Smolander, Iaakov Exman, and Dewayne E. Perry. 2013. Report on the second SE-
[1] Gordon W Allport. 1954. The Nature of Prejudice. Addison-Wesley. MAT workshop on general theory of software engineering (GTSE 2013). SIGSOFT
[2] Chris Argyris. 1995. Action science and organizational learning. Journal of Software Engineering Notes 38, 5 (2013), 47–50.
Managerial Psychology 10, 6 (1995), 20–26. [34] Staffs Keele et al. 2007. Guidelines for performing systematic literature reviews in
[3] D E Avison and A T Wood-Harper. 1986. Multiview&Mdash;an Exploration in software engineering. Technical Report. Citeseer.
Information Systems Development. Aust. Comput. J. 18, 4 (Nov. 1986), 174–179. [35] Laurie J Kirsch. 1996. The management of complex tasks in organizations:
[4] Sebastian Baltes and Paul Ralph. in press. Sampling in software engineering Controlling the systems development process. Organization Science 7, 1 (1996),
research: A critical review and guidelines. Empirical Software Engineering (in 1–21.
press). [36] Adam Kuper and Jessica Kuper (Eds.). 2004. The Social Science Encyclopedia (3rd
[5] Izak Benbasat and Henri Barki. 2007. Quo vadis TAM? Journal of the Association ed.). Routledge, Chippenham, UK.
for Information Systems 8, 4 (2007), 7. [37] Sarah Lichtenstein and Paul Slovic. 2006. The Construction of Preference. Cam-
[6] Marc Bertin and Iana Atanassova. 2014. A study of lexical distribution in citation bridge University Press, Cambridge, UK.
contexts through the IMRaD standard. PloS Negl. Trop. Dis 1, 200,920 (2014), [38] Sanghee Lim, Terence JV Saldanha, Suresh Malladi, and Nigel P Melville. 2013.
83–402. Theories used in information systems research: Insights from complex network
[7] Elizabeth Bjarnason, Kari Smolander, Emelie Engström, and Per Runeson. 2016. A analysis. JITTA: Journal of Information Technology Theory and Application 14, 2
theory of distances in software engineering. Information and Software Technology (2013), 5.
70 (2016), 204–219. [39] Edwin A Locke and Gary P Latham. 1990. A Theory of Goal Setting & Task
[8] Barry W Boehm, Chris Abts, A Winsor Brown, Sunita Chulani, Bradford K Clark, Performance. Prentice-Hall, Inc.
Ellis Horowitz, Ray Madachy, Donald J Reifer, and Bert Steece. 2009. Software [40] Kalle Lyytinen and Jan Damsgaard. 2001. What’s wrong with the diffusion of
Cost Estimation with COCOMO II. Prentice Hall PTR. innovation theory?. In Working Conference on Diffusing Software Product and
[9] Barry W. Boehm and Rony Ross. 1989. Theory-W software project management Process Innovations. Springer, 173–190.
principles and examples. IEEE Transactions on Software Engineering 15, 7 (1989), [41] Thomas W Malone et al. 1988. What is coordination theory? (1988).
902–916. [42] Thomas W Malone and Kevin Crowston. 1994. The interdisciplinary study of
[10] Pierre Bourque and Richard E. Fairley. 2014. Guide to the Software Engineering coordination. ACM Computing Surveys (CSUR) 26, 1 (1994), 87–119.
Body of Knowledge (SWEBOK (R)): Version 3.0. IEEE Computer Society Press. [43] RE Mayer. 2001. Multimedia Learning. 2001.
[11] P. Checkland. 1999. Systems Thinking, Systems Practice. Wiley, Chichester. [44] Robert King Merton and Robert K Merton. 1968. Social Theory and Social Structure.
[12] Mohammad Y Chuttur. 2009. Overview of the technology acceptance model: Ori- Simon and Schuster.
gins, developments and future directions. Sprouts: Working Papers on Information [45] Lawrence B Mohr. 1982. Explaining Organizational Behavior. Jossey-Bass.
Systems 9, 37 (2009).

2004
ICSE 2022, May 21–29, 2022, Pittsburgh, PA, USA Lorey et al.

[46] Stewart C Myers. 1977. Determinants of corporate borrowing. Journal of Financial [65] Everett M Rogers. 1983. Diffusion of Innovations. The Free Press.
Economics 5, 2 (1977), 147–175. [66] Binny M Samuel, Linwood Watkins, Andrew Ehle, and Vijay Khatri. 2015. Cus-
[47] Ikujiro Nonaka. 1994. A dynamic theory of organizational knowledge creation. tomizing the representation capabilities of process models: understanding the
Organization Science 5, 1 (1994), 14–37. effects of perceived modeling impediments. IEEE Transactions on Software Engi-
[48] Ana Ortiz de Guinea and Jane Webster. 2014. Overcoming variance and process neering 41, 1 (2015), 19–39.
distinctions in information systems research. (2014). [67] Teresa M Shaft and Iris Vessey. 2006. The role of cognitive fit in the relationship
[49] Allan Paivio. 1990. Mental Representations: A Dual Coding Approach. Oxford between software comprehension and modification. MIS Quarterly 30, 1 (2006),
University Press. 29–55.
[50] Ananthanarayanan Parasuraman, Valarie A Zeithaml, and Leonard L Berry. 1988. [68] Dag IK Sjoberg, Tore Dyba, and Magne Jorgensen. 2007. The future of empirical
Servqual: A multiple-item scale for measuring consumer perc. Journal of Retailing methods in software engineering research. In 2007 Future of Software Engineering.
64, 1 (1988), 12. IEEE Computer Society, 358–378.
[51] Jeffrey Pfeffer and Gerald R Salancik. 2003. The External Control of Organizations: [69] Susan Leigh Star and James R. Griesemer. 1989. Institutional ecology, ’translations’
A Resource Dependence Perspective. Stanford University Press. and boundary objects: Amateurs and professionals in Berkeley’s museum of
[52] Peter Pirolli and Stuart Card. 1995. Information foraging in information access vertebrate zoology, 1907-39. Social Studies of Science 19, 3 (1989), 387–420.
environments. In Proceedings of the SIGCHI Conference on Human Factors in [70] Klaas-Jan Stol and Brian Fitzgerald. 2013. Uncovering theories in software
Computing Systems. ACM Press/Addison-Wesley Publishing Co., 51–58. engineering. In 2013 2nd SEMAT Workshop on a General Theory of Software
[53] Paul Ralph. 2012. Improving coverage of design in information systems education. Engineering (GTSE). IEEE, 5–14.
In Proceedings of the International Conference on Software Engineering. IEEE, [71] Klaas-Jan Stol and Brian Fitzgerald. 2015. Theory-oriented software engineering.
Orlando, Florida, USA. Science of Computer Programming 101 (2015), 79–98.
[54] Paul Ralph. 2013. The illusion of requirements in software development. Require- [72] Klaas-Jan Stol, Michael Goedicke, and Ivar Jacobson. 2016. Introduction to the
ments Engineering 18, 3 (2013), 293–296. special section—general theories of software engineering: New advances and
[55] Paul Ralph. 2015. The sensemaking-coevolution-implementation theory of implications for research. Information and Software Technology 70 (2016), 176–
software design. Science of Computer Programming 101 (2015), 21–41. https: 180.
//doi.org/10.1016/j.scico.2014.11.007 [73] Klaas-Jan Stol, Paul Ralph, and Brian Fitzgerald. 2016. Grounded theory in
[56] Paul Ralph. 2016. Software engineering process theory: A multi-method com- software engineering research: A critical review and guidelines. In Proceedings
parison of Sensemaking-Coevolution-Implementation Theory and Function- of the International Conference on Software Engineering. IEEE, Austin, TX, USA,
Behavior-Structure Theory . Information and Software Technology 70 (2016), 120–131.
232–250. https://doi.org/10.1016/j.infsof.2015.06.010 [74] John Sweller. 1988. Cognitive load during problem solving: Effects on learning.
[57] Paul Ralph. 2018. The two paradigms of software development research. Science Cognitive Science 12, 2 (1988), 257–285.
of Computer Programming 156 (2018), 68–89. [75] Mary Tate, Joerg Evermann, and Guy Gable. 2015. An integrated framework for
[58] Paul Ralph. 2018. Toward methodological guidelines for process theories and theories of individual attitudes toward technology. Information & Management
taxonomies in software engineering. IEEE Transactions on Software Engineering 52, 6 (2015), 710–727.
45, 7 (2018), 712–735. [76] Nigel Thomas. 2014. Dual coding and common coding theories of memory. In
[59] Paul Ralph, Mike Chiasson, and Helen Kelley. 2016. Social theory for software Stanford Encyclopedia of Philosophy. Stanford University Press. https://plato.
engineering research. In Proceedings of the International Conference on Evaluation stanford.edu/entries/mental-imagery/theories-memory.html
and Assessment in Software Engineering. ACM, Limerick, Ireland. https://doi.org/ [77] Stephen E Toulmin. 2003. The Uses of Argument. Cambridge University Press.
10.1145/2915970.2915998 [78] Andrew H. Van de Ven. 2007. Engaged Scholarship: A Guide for Organizational
[60] Paul Ralph, Gregor Engels, Ivar Jacobson, and Michael Goedicke. 2015. 4th and Social Research. Oxford University Press, Oxford, UK.
SEMAT workshop on general theory of software engineering (GTSE 2015). (May [79] Viswanath Venkatesh, Michael G Morris, Gordon B Davis, and Fred D Davis.
2015), 983–984. 2003. User acceptance of information technology: Toward a unified view. MIS
[61] Paul Ralph, Iaakov Exman, Pan-Wei Ng, Pontus Johnson, Michael Goedicke, Quarterly 27, 3 (2003), 425–478.
Alper Tolga Kocata, and Kate Liu Yan. 2014. How to develop a general theory of [80] Viswanath Venkatesh, Michael G. Morris, Gordon B. Davis, and Fred D. Davis.
software engineering: Report on the GTSE 2014 workshop. SIGSOFT Software 2003. User acceptance of information technology: Toward a unified view. MIS
Engineering Notes 39, 6 (Dec. 2014), 23–25. https://doi.org/10.1145/2674632. Quarterly 27, 3 (2003), 425–478.
2674647 [81] Iris Vessey. 1991. Cognitive fit: A theory-based analysis of the graphs versus
[62] Paul Ralph, Pontus Johnson, and Howell Jordan. 2013. Report on the first SEMAT tables literature. Decision Sciences 22, 2 (1991), 219–240.
workshop on a general theory of software engineering (GTSE 2012). SIGSOFT [82] Claes Wohlin, Per Runeson, Martin Höst, Magnus C Ohlsson, Björn Regnell, and
Software Engineering Notes 38, 2 (2013), 26–28. Anders Wesslén. 2012. Experimentation in Software Engineering. Springer Science
[63] Paul Ralph and Ewan Tempero. 2018. Construct validity in software engineering & Business Media.
research and software metrics. In Proceedings of the 22nd International Conference [83] He Zhang, Xin Huang, Xin Zhou, Huang Huang, and Muhammad Ali Babar. 2019.
on Evaluation and Assessment in Software Engineering. ACM Press, New York, Ethnographic research in software engineering: a critical review and checklist. In
New York, USA, 13–23. https://doi.org/10.1145/3210459.3210461 Proceedings of the 2019 27th ACM Joint Meeting on European Software Engineering
[64] James Reason. 1990. Human Error. Cambridge University Press. Conference and Symposium on the Foundations of Software Engineering. 659–670.

2005