3. Research approach: inductive from case studies

3.1.1 Process data: Over a decade, ways that open sourcing does and doesn't work with private sourcing were discovered

Insight into evolving phenomena such as organizational learning, innovation and change can be gained through studying data longitudinal over time, rather than as cross-sectional views at a single point in time. When a body of theory is limited and fine-grained qualitative data is available, new theories can be built and later tested with time series and event history methods.

Process data are series of events, activities and choices ordered over time. For researchers, four characteristics that makes them difficult to analyze and manipulate.

First, they deal mainly with sequences of "events": conceptual entities that researchers are less familiar with. Second, they often involve multiple levels and units of analysis whose boundaries are ambiguous. Third, their temporal embeddedness often varies in terms of precision, duration, and relevance. Finally, despite the primary focus on events, process data tend to be eclectic, drawing in phenomena such as changing relationships, thoughts, feelings, and interpretations (Langley, 1999, p. 692).

The data on OSwPS in Appendices A and B follows that list of four characteristics.

On (i) “data composed of events”, the most interesting aspects are “stories about what happened and who did what when”. Actions and decision made under uncertainty may provide more insight than directions espoused a priori.

On (ii) “data on multiple units and levels of analysis with ambiguous boundaries”, some actions were taken by individuals formally charged with responsibilities and resources, whereas others have relied on guidelines and volunteerism, within and beyond organizational boundaries.

On (iii) “data of variable temporal embeddedness”, the sequences of events occurs at irregular periods, and histories recalled retrospectively tend to focus on significant changes. Individuals and groups learned about OSwPS not only within the confines of a single case, but across the organization in complementary or peripheral projects and activities.

On (iv) “data that are eclectic”, variables and events are intertwined. In the cases that have spanned years, the players – both individuals and organizations – have changed. The cases may include unclear and evolving purposes altered from hands-on learning, factors in the larger organization, or unforeseen external pressures.

A study based on longitudinal data in the real world does not have the benefit of data on “the road not taken”. There is, however, a reality of action from conscious choices made by actors immersed in the situation.

3.1.2 Multilevel data: open sourcing while private sourcing coevolved for individuals, teams, corporations and non-profits

The origins for seven case histories on OSwPS vary: some started as formally funded corporate initiatives, and others started as voluntary contributions by individuals working on their own time. In an open sourcing style, relatively few people are committed full-time to an exploratory project, yet success is later measured through the widespread adoption of an innovation.

In a multilevel approach, individuals, groups and organizations are embraced simultaneously, with two advantages. Firstly, researchers can avoid significant fallacies associated with single-level empirical research.¹²⁵ Secondly, a multilevel perspective opens up new opportunities for theory understanding linkages between levels.

In a business working in OSwPS at an organizational level, there will be also be individuals and groups concurrently working in open sourcing only (OSo) or in private sourcing only (PSo). Further, employees can be active with other parties outside of the company participating in OSo and PSo communities. An approach that includes data from multiple levels encourages an appreciation of the interdependencies that are simultaneously in play in the real world.

With multilevel process data as the source for study, the next section describes the approach for analysis.

3.2.1 Sequencing actions and circumstances aims to explain outcomes in the stream of changes within a domain

The states of open sourcing and private sourcing for an initiative or offering are dynamic. The time-based nature leads to an analysis of process. In the context of organization science, a working definition for process is “a sequence of individual and collective events, actions and activities unfolding over time in context” (Pettigrew, 1997, p. 338). Process thinking “may involve consideration of how and why things – people, organizations, strategies, environments – change, act and evolve over time ... or ... how such ‘things’ come to be constituted, reproduced, adapted and defined through ongoing processes (Langley, 2007, p. 271). The role of time is more prominent than in cross-sectional models that may assume an equilibrium state, and variance theories that do not take advantage of temporally embedded accounts.

Processual analysis involves more than just the telling of stories in a case history. “The irreducible purpose of a processual analysis remains to account for and explain the why, why and how of the links between context, processes and outcome” (Pettigrew, 1997, p. 340). With the events and chronologies as building blocks, processual analysis includes three factors: (i) “a search for patterns in the process and presumably some attempt to compare the shape, character and incidence of this pattern in case A compared with case B”; (ii) “a quest to find the underlying mechanisms which shape any patterning in the observed processes”, and (iii) inductive pattern recognition hand in hand with a deductive component, particularly in the structuring of data sets (Pettigrew, 1997, p. 339). This processual analysis shows up in the sequences of events of open sourcing, and of private sourcing, described in detail in Appendix 1, and summarized in Chapter 4. Within each cases, IBM guided its private sourcing business directions learning from the activities within the open sourcing community; and simultaneously, the open sourcing community was learning from the private sourcing activities of IBM.

This process thinking is compatible with the strategy-as-practice perspective “with its focus on strategy as ‘something that people do’ on micro-level activities and practices. In general, though, process “does not necessarily demand a micro focus and can be applied to temporally evolving phenomena at a variety of different levels (individual, organizational, sector, field), including phenomena related to strategy content issues” (Langley, 2007, p. 272).

3.2.2 Replicating theoretically across multiple case studies infers a business context changing systemically, rather than just situationally

This study on OSwPS takes an approach of multiple case studies in theoretical replication, which should lead to contrasting results for predictable reasons (Yin, 2003, p. 47). While these cases are centered on a single company, their contexts vary. Some cases involve open sourcing communities external to the company; some emphasize internal employee participation; some cross over from internal to external. Some cases center more on formal roles in developing and delivering product, others engage volunteers. This applies an integrative form of "longitudinal replication" where "temporal brackets … are constructed in progressions of events and activities separated by identifiable discontinuities in the temporary flow" (Langley, Smallman, Tsoukas, & Van de Ven, 2013, p. 7).

The alternative of multiple case studies for literal replication does not lend itself to the time-oriented nature of situations and choices in those contexts. The aim is for “more robust, generalizable and testable theory (Eisenhardt & Graeber, 2007, p. 27). The learning within the ten years between 2001 and 2011 sees a ramp from OSwPS as unconventional way of doing business, to practices acknowledged and adopted in a variety of industries.

3.2.3 Appreciating contexts changing sees individual/workgroup dynamics coevolving with organizational/institutional redefinitions

While the focus of this study on OSwPS has been centered on seven case studies, the processes are both shaped and shaped by context.¹²⁶

In a multilevel study the boundaries between inner context (i.e. including individuals and groups within an organization) and outer context (e.g. an organization in its environment) dissolve.¹²⁷ Participation in open sourcing communities, and privileged interorganizational sharing of private sourcing content presents contexts that transcends boundaries.¹²⁸ Each case has its own context, yet those contexts overlap as learning from one situation in OSwPS informs another. The inner context is addressed more in Chapter 4; the outer context follows in Chapter 5.

3.3.1 Abstracting towards theory draws from concrete case studies supplemented by descriptions of contemporaneous contexts

In this study on OSwPS, inductive reasoning¹²⁹ is applied: cases are used to build theory (i.e. rules as beliefs about the way the world is structured).

“The building of theory occurs in two major stages – the descriptive stage and the normative stage” (Carlile & Christensen, 2005, p. 2). Chapters 6, 7 and 8 center on the former, the descriptive stage. Chapter 9 centers on the latter, the normative stage. The building of descriptive theory involves (i) observing, describing and measuring the phenomena; (ii) categorizing by the attributes of the phenomena; to induce (iii) preliminary statements of correlation.¹³⁰ The building of normative theory involves (i) observing, describing and measuring the phenomena (as with descriptive theory building); (ii) categorization of the circumstances in which we might find ourselves; to induce (iii) statement(s) of causality.

Observation of the phenomenon, as outlined in section 3.1, follows the processual history in seven cases and their containing contexts into descriptions of constructs. Categorization defines the attributes of the phenomenon, simplifying and organizing the observations into frameworks and typologies. Association defines the relationship between the categories defined, and the outcomes observed, resulting in preliminary models.

The result of these three steps leads to assertions – descriptive theory – across the scope of the data collected. Improving the theories from this study of OSwPS can then be conducted in subsequent deductive processes, where the models are tested with different sets of data.

The goal of this theory-building – both descriptive and normative – is to primarily to provide knowledge useful to practitioners in business. It aims to improve practices in management, with subsequent deductive work later improving the validity into a practical science.¹³¹

The applicability of findings beyond the domain in which the data have been collected present opportunities for abduction in putting normative theory-building into practice. Entrepreneurial business people, as well as researchers, may modify and/or extend the models, frameworks and typologies from the pioneering cases in the software business to other industries.¹³²

Within the narrow scope of software development, practices in open sourcing and private sourcing are rather well-known and in common use. In the slightly broader scope of technology businesses, the definitions of open sourcing and private sourcing provide within this study will be somewhat unfamiliar in application. In an unbounded scope across the world of business, the interest and applicability in these models has yet to be asked.

3.3.2 Generating pattern language with a paradigm grounds a theory emerging with hypothesizing

In chapters 6, 7 and 8, analysis of the cases is presented in a repeating sequence of (i) paradigm description, (ii) theory emerging with the paradigm, (iii) generative pattern language entailing theory, and (iv) hypothesizing for a theory within the paradigm. This sequence reflects refinement of the data from constructs, through frameworks and typologies, towards models.

From the cases, theories have emerged inductively: quality-generating sequencing, in section 6.2; affordances wayfaring in section 7.2; and anticipatory appreciating in section 8.2. Simultaneously, three paradigms in which the theories are further developed have emerged inductively. This follows from a multiparadigm approach to theory-building that may surface inherent and irreconcilable theoretical differences and lead for more comprehensive interplay.¹³³

A paradigm is a world view underlying the theories and methodology of a particular scientific subject.¹³⁴ Outlining key concepts in a paradigm clarifies the use of words in specific meanings. The paradigm discussed in section 6.1 is architectural problem-seeking; in section 7.1, it is inhabiting disclosive spaces; and in section 8.1, it is governing subworlds.

Within each of the three chapters that follow, a generative pattern language format is used to establish some consistency in the structure of descriptions around theories and paradigms. The pattern language approach has been adapted from its 1960s roots in built environments to today’s world of service systems.

In the domain of architecting built environments, a pattern was originally defined as an abstract relation (e.g. between solution and problem) that occurs within some conditions (e.g. a range of contexts).¹³⁵ Architecture of "unselfconsciously" constructed artifacts of tradition were observed to not suffer from the adaptation, quality and usability failures criticized in the products of rational professionalized design practices.¹³⁶ Building on experiences evolving in architectural practice since the 1960s, formalization of a scientific pattern method has been proposed some 50 years later.¹³⁷

The origins of pattern language are rooted in a mathematical sense of formal language (as can be compared to natural language).¹³⁸ With formal languages, inductive inference that includes both positive data and negative data is more powerful than inference only from positive data.¹³⁹ In specifying a constructive inference method for a formal language, the grammar should have characterizable results (i.e. correct identification) and efficiency (in a relatively small language that could be produced in polynomial time).¹⁴⁰

Pattern languages have been proposed as lingua francas (i.e. common languages – as multiple lingua franca) accessible to all stakeholders in a design process.¹⁴¹ In a design process with high diversity (e.g. interactive systems involve visual designers, social scientists and technologists), a pattern language that includes both vocabulary and conceptual frameworks brought by disciplines and professionals, can serve as a lingua franca for communicating in a more egalitarian way.

In developing a pattern language, induction precedes deduction. Pattern language can be used on empirical studies as representations from which design work is derived.¹⁴² From qualitative data, (i) inductive analysis methods are used to identify patterns, (ii) deductive methods are used to articulate (i.e. structure and format) the patterns into a language; and (iii) comparative methods can be used to extend the contexts in which the pattern language has validity.¹⁴³ Pattern language can be generated through collaborations amongst people who have tacit knowledge within a target domain.¹⁴⁴ The validity of a pattern language is not testable as the hypothesis of a single pattern, but instead in the network of hypotheses and their interactions as a complex system.¹⁴⁵

A feature of pattern language that is often under-emphasized is the associative network that selectively links patterns to others at (a) larger scale(s) and/or (a) smaller scale(s). Recasting pattern language from its structuralist roots to a more interactionalist philosophy, an alternative format makes containing and contained systems more explicit (Ing, 2016, p. 11). The pattern form proposed for framing service systems thinking is outlined in Table 3.1.

This adaptation sees OSwPS (and infers PSo and OSo) with parties engaged in service systems. The following definitions guide authoring of the pattern.

(i) Pattern label: A service system involves an interaction – minimally between a beneficiary and a provider – and thus can be expressed as a present participle. Attaching an "-ing" suffix to a verb in English changes it to a continuous form. For conciseness, a participial phrase with a noun suffices (and a longer expression draws risks of a dangling participle).

(ii) Voices on issues: A service system introduces human perspectives into a pattern. Voices are heard (or not heard) from archetypal roles of stakeholders. Stakeholders of a service system could include not only beneficiaries, sponsors and funders, but also neighbours and regulatory bodies seeking constraints on impacts (sometimes called externalities). Expressing issues as questions, rather than problem as statements, can serve as checklist for determining whether an issue has been dealt with, or at least acknowledged. An issue for a service system could be expressed as a one-to-many relation.

(iii) Affording value(s): Affording an action possibility with value for a service system under specified conditions is neither necessary nor sufficient to resolve an issue. With an affordance, an object or event becomes available for an individual or a group to use in practice.

(iv) Spatio-temporal frames: The when and where of affording value(s) is socially negotiated between a client and a service provider. As an example, a service system that electronically provides an affordance for routing around traffic congestion, consider the when and where of (i) a dedicated GPS navigation device and (ii) a mapping app on a smartphone. For the features of aided routing for either a driver or a pedestrian, the service provider must have already mapped that territory before the client encounters that place. For the driver or pedestrian, a smartphone app normally requires an active wireless Internet service, while the GPS doesn’t. A service system providing an "affordance" would presume that a person has eyesight and would not autonomically navigate himself or herself into a ditch.

(v) Containing systems: Authentic systems thinking starts from containing whole.¹⁴⁶ A client of a service system is typically defined as being outside its systems boundaries, and therefore part of the containing whole. In situations where a client is not merely a consumer, but instead a coproducer, the definition of the system changes. "Value co-produced by two or more actors, with and for each other, with and for yet other actors, invites us to rethink organizational structures and managerial arrangements for value creation inherited from the industrial era" (Ramírez, 1999, p. 49).

(vi) Contained systems: From panarchy theory, smaller and faster systems can "revolt" to change the system of interest (Gunderson & Holling, 2002). In service systems, this is typically demonstrated by a segment of clients voicing shared concerns resulting in a response by the provider, or exit from that relationship to other alternatives.

As the data for the research is focused on OSwPS as a concern, induction has led to three patterns in sections 6.3, 7.3 and 8.3. A concern is a matter of interest originating from one or more stakeholders in a system.¹⁴⁷ Concerns can be typed in many ways, e.g. logical concerns vs. physical concerns; simple concerns vs. composite concerns.¹⁴⁸ Concerns are defined to exist in multiplicity, concurrently and in overlapping classifications or dimensions.¹⁴⁹ The separation of concerns challenges decomposing matters of interest while composing systems with reduced complexity that facilitate evolution.¹⁵⁰ Within a paradigm, OSwPS is regarded as a concern separate from, although related to OSo and PSo. To gain an appreciation of contrasts with OSo and PSo, partial patterns have inferred for each of those concerns, without collecting additional data to validate them. This approach leads to 3 full patterns and 6 partial patterns in each of sections 6.3, 7.3 and 8.3. The motivation for this extension is to strengthen the inductive approach of theory building, relying first on the data collected specifically for this study. The subsequent inclusion of complementary data is a process of enfolding that “involves asking what is this similar to, what does it contradict, and why” (Eisenhardt, 1989, p. 544).

In sections 6.4, 7.4 and 8.4, hypotheses are stated formally as emerging propositions that may be deductively tested with datasets in similar or different businesses.

3.4.1 Corollaries learning across plural paradigms reflection on the design of underlying inquiring systems

Inquiring systems can be categorized into five ways of knowing: (i) inductive-consensual, (ii) analytic deductive, (iii) multiple realities, (iv) dialectic, and (v) unbounded systems thinking (Mitroff & Linstone, 1993). These correspond to the philosophies of (i) Locke, (ii) Leibniz, (iii) Kant, (iv) Hegel, and (v) Singer (Churchman, 1971). This study on OSwPS recognizes multiple realities, and the potential for dialectic. The presentation building descriptive theories each associated with a distinctive paradigm does not exclude the potential for additional worldviews. This is consistent with the principle of “sweeping in” more knowledge towards unbounded systems thinking.

In prior organizational research, pluralistic contexts have been acknowledged. They may be characterized by features such as multiple objectives, diffuse power and knowledge-based work processes, with some organizations “more pluralistic than others”.¹⁵¹ The decentralization of authority around clusters of customer and assets results in a work life where the combined resources of empowered employees with internal and external parties results in unique rather than universalistic contexts.¹⁵² An organization may attempt to command-and-control policies on open sourcing and/or private sourcing, but collaborative arrangements will be made in one of the many situational horizontal (and interorganizational) contexts.

Pluralism in paradigms has been applied in organizational research is a way of accommodating multiple worldviews.¹⁵³ Alternative approaches to inquiry bring different assumptions about (i) ideology, (ii) ontology, and (iii) epistemology. These assumptions are summarized for modern paradigm approaches, multiparadigm approaches and postmodern paradigm approaches in Table 3.2.

As research that has started from an inductive method on a phenomenon, this study on open sourcing while private sourcing follows the goals of a multiparadigm approach, with (i) an accommodating ideology, (ii) a stratified ontology, and (iii) a pluralist epistemology.¹⁵⁴

Building descriptive theory through multiparadigm inquiry takes a multiparadigm research approach. This should be not confused with multiparadigm theory building that attempts to link conflicting paradigm insights.¹⁵⁵ This choice may leave unresolved paradoxes with theoretical contradictions and oppositions embedded in complex traditions (Poole & van de Ven, 1989, p. 564). In the design of an inquiring system as unbounded systems thinking, this should encourage “sweeping in” of additional knowledge.¹⁵⁶

3.4.2 Interplaying across multiple paradigms encourages fuller synthesis for future development

Having selected an approach of multiparadigm research, there is a choice of metatheoretical positions that include (i) paradigm incommensurability, where each paradigm is developed and applied separately; (ii) paradigm integration, assessing and synthesizing a variety of contributions ignoring differences between competing approaches and underlying assumptions; and (iii) paradigm crossing in sequential, parallel or bridging approaches. This research on OSwPS takes a metatheoretical position of (iv) paradigm interplay that “refers to the simultaneous recognition of both contrasts and connections between paradigms and, thus, to both the differences and similarities between paradigms that are emphasized by the parallel and bridging strategies, respectively” (Schultz & Hatch, 1996, p. 534). This enables cross-fertilization across diverse perspectives without demanding integration.¹⁵⁷

Mixing inductive reasoning with a discussion of paradigms leads to issues of incompleteness. Starting from paradigmatic categories (e.g. sociological theories as functionalist, interpretive, radical humanist and radical structuralist (Burrell & Morgan, 1979)) is deductive. Alternative paradigmatic categories (e.g. by discursive features as normative, interpretive, critical and dialogic (Deetz, 1996); or by systems approaches as functionalist, interpretivist, emancipatory and postmodern (Jackson, 2000)) have added to lists for consideration for deductive inquiries. This study on open sourcing while private sourcing leans towards an explanationist view “acknowledging the use of pragmatic virtues in reasoning and embraces the idea that theoretical explanation indeed plays a role in scientific inference” rather than the Spartan view that “rejects pragmatic virtues and accepts truth as the only virtue in scientific inference (Ketokivi & Mantere, 2010, p. 317).

In normative approaches to systems analysis, the term “perspective” has generally been preferred over “paradigm”. The Multiple Perspective Concept is consistent with principles in the design of an inquiring system that is a metainquiring system.¹⁵⁸ In sociotechnological systems – specifically technology assessments – three perspectives are proposed: a technical perspective (T), an organizational perspective (O), and a personal perspective (P).¹⁵⁹

Multiple perspectives can be approached from a variety of philosophies and cultures. Based in Chinese philosophy, the WSR approach to interventions view sociotechnical systems as constituted by wu (objective existence), shi (subjective modeling) and ren (intersubjective human relations) (Gu & Zhu, 2000). When the American TOP (T, O, and P perspectives) is compared with the Chinese WSR (wuli-shili-renli), the multiperspective systems approaches each map the world and follow with methodological treatments differently, and yet can provide opportunities for mutual learning (Z. Zhu, 2002). In systems research, increased sensitivity to the cultural dimension (e.g. west vs. east) in methodologies has been seen as overlooked in a world of deepening globalization (Z. Zhu, 2010). In addition to the recognition of diversity in groups – as diverse perspectives, diverse interpretations, diverse heuristics, and diverse predictive models (Page, 2008) – the “simultanenous localization and globalization driven by information technology inevitably invites a variety of perspectives” (Linstone, 2010, p. 697).

In aspiring to the design of a metainquiring system, care is taken to not overstep its inductive nature with OSwPS. The crossing of philosophical distinctions follows the spirit of multiparadigm research and paradigm interplay, yet the perspectives may be neither orthogonal nor complete.