Opening Remarks
Statistical Equivalence
Models with excellent fit tell us little about the data-generating mechanisms. Fried (2020, p. 275)
Fried (2020) argues that a well-fitting statistical model cannot establish that a construct exists or show where to intervene in a causal system, because different causal structures can produce the same data. Bendoly & Oliva (2025) take this “statistical equivalency” as one of the fundamental reasons why research needs well-reasoned theoretical arguments: when phenomena are complex and mechanisms are not obvious, data alone do not tell us what is real and what is an artifact of computation.
The first two days dealt with how a manuscript is positioned and written. This unit turns to its foundations: how the manuscript produces knowledge, which logic of theory development it follows, and which assumptions about reality and knowledge it rests on. Reviewers rarely ask about these foundations directly; they notice when the parts of a manuscript do not fit together.
Learning Outcomes
After completing this unit, you will be able to:
- Explain the functions of theory and its building blocks.
- Distinguish deductive, inductive, and abductive logics of theory development.
- Classify your manuscript along the three pathways of empirical research in theory development.
- Analyze the alignment of paradigm, theory, method, and claims in your manuscript and in your conversation.
Theorizing
Why Theorizing Matters
Theories exist along a continuum from vague hunches to detailed accounts of causal mechanisms (Weick, 1989).
- Early theorizing creates constructs and narratives for an observed phenomenon.
- Theorizing matters most where phenomena are not straightforward and mechanisms are not obvious (Bendoly & Oliva, 2025).
Weick (1989) describes theory construction as “disciplined imagination”: researchers generate conjectures and select among them with explicit criteria. Theories are never finished products; they range from first hunches and typologies to fully specified causal models. Bendoly & Oliva (2025) build on this view: the initial phases of theorizing often consist of defining constructs and narratives that account for an observed phenomenon, and later phases specify mechanisms and boundary conditions.
For a manuscript, the position on this continuum determines what kind of contribution is realistic: a paper that introduces and defines a new construct makes a different contribution than a paper that tests a fully specified model.
Functions of Theory
Three functions of theory in empirical research (Bendoly & Oliva, 2025):
- Inspiring work: theory serves as a seed for new inquiries, even when it is still unrefined.
- Motivating design: theoretical arguments are the warrants for hypotheses or propositions.
- Aiding interpretation: theory helps to make sense of new observations and patterns.
Bendoly & Oliva (2025) describe theory as the vehicle for making sense of things. As a design motivator, theory provides the arguments that precede claims: well-formulated hypotheses are logically sound, grounded in previous knowledge, and empirically testable. As an interpretation aid, theory helps researchers rationalize new observations and patterns, which in turn inspires further work. The three functions correspond to different stages of a research process, and a single paper usually emphasizes one of them.
Building Blocks
| Element | Level | Description |
|---|---|---|
| Construct | Abstract | A concept chosen or created to explain a phenomenon; not directly observable |
| Variable | Observable | An observable entity that can take two or more values |
| Proposition | Abstract | A statement of relationship among constructs |
| Hypothesis | Observable | A statement of relationship among variables |
Bacharach (1989) defines theory as a system of constructs and variables in which constructs are related by propositions and variables by hypotheses. The system is bounded by the theorist’s assumptions about values, time, and space. Bacharach proposes two criteria for evaluating theories: falsifiability (can the theory be shown to be wrong?) and utility (does it explain and predict?).
Markus (2008) distinguishes concepts from constructs. Concepts extend over actual and possible cases and can travel across disciplines; a construct is a concept chosen or created to explain a specific phenomenon. Distance, for example, is a concept that can explain social separation, gravitational influence, or the influence between societies separated by time; as a construct, it is defined for one of these phenomena. The distinction connects to the construct paragraphs from the second day: a construct needs a definition that fixes its meaning for the phenomenon at hand.
Types of Theory
| Type | Purpose | Example question |
|---|---|---|
| I. Analysis | Describe and classify | What kinds of digital platforms exist? |
| II. Explanation | Explain how and why, without precise prediction | How do professionals deal with opaque AI results? |
| III. Prediction | Predict without explaining | Which firms will adopt AI? |
| IV. Explanation and prediction | Explain and predict, with testable propositions | Why and to what extent does CIO presence increase AI orientation? |
| V. Design and action | Prescribe how to do something | How should an AI-based decision support system be designed? |
Gregor (2006) distinguishes five types of theory by their primary goal. The types are interrelated: analysis provides the classifications on which explanation and prediction build, and design theories draw on explanatory theories. Gregor also distinguishes views of theory: theory as a lens for viewing the world (formal testing not envisaged), theory as testable statements of relationships among constructs, and theory as prescriptions for practice.
Knowing which type your manuscript develops helps to set expectations correctly: a type II paper does not need to predict, and a type V paper is evaluated by the utility of its design knowledge.
Pathways
Logics of Inference
| Logic | Direction | Conclusion |
|---|---|---|
| Deduction | From theory to specific predictions | Follows necessarily if the premises are true |
| Induction | From observed instances to a generalization | Probable, strengthened by further observations |
| Abduction | From a surprising observation to the best explanation | Plausible, a conjecture to be examined further |
Deduction derives specific, testable claims from general theoretical statements; if the premises are true and the reasoning is valid, the conclusion must be true. Induction generalizes from observed instances; its conclusions are probable and can be overturned by new observations. Abduction starts from an observation that existing theory cannot explain and proposes the explanation that would make the observation expected; its conclusions are conjectures that require further examination.
Most research projects combine the three logics: an abductive insight from a case study can be formalized into propositions and then tested deductively. Manuscripts benefit from stating which logic dominates, because reviewers apply different standards to a theory test than to theory building.
Three Paths
Three ways empirical research contributes to theory development (Bendoly & Oliva, 2025):
- Path A: Testing theoretical claims (deduction): experiments, quasi-experiments, and secondary data analysis refute or support claims derived from theory.
- Path B: Observation and explanation (abduction and induction): researchers observe real-world processes; regularities that existing theory cannot explain call for new constructs and explanations.
- Paths A and B: Intervention-based research: researchers apply theory to solve a real-world problem and refine the theory based on the outcome.
Bendoly & Oliva (2025) describe the paths for operations management, but they transfer to most empirical disciplines. Path A covers lab experiments (precision), field experiments (realism), natural and quasi-experiments, and causal inference from secondary data. Path B starts from observation: if observations align with existing theory, induction strengthens confidence in the theory; if they do not, abduction proposes new constructs and explanations. Intervention-based research (e.g., action research, design science) combines both: interventions are designed deductively from theory, and their outcomes modify the theory abductively.
Paradigms
Assumptions
| Assumption | Question | Poles |
|---|---|---|
| Ontological | What is the nature of reality? | Realism vs. constructivism |
| Epistemological | How can we know about it? | Positivism vs. anti-positivism |
| Methodological | Which procedures produce knowledge? | Deductive vs. inductive |
| Axiological | Which role do values play in research? | Value-free vs. value-laden |
Every study rests on assumptions about what exists, how it can be known, how knowledge is produced, and which role the researcher’s values play. These assumptions are rarely stated in a manuscript, but they shape the research question, the method, and the type of claim a study can make. The epistemology unit of the RP1 course offers a broader introduction for readers who want to revisit the foundations.
Paradigms
Three research paradigms in information systems (Orlikowski & Baroudi, 1991):
| Paradigm | View of reality | Aim | Typical evidence |
|---|---|---|---|
| Positivist | Objective, independent of the observer | Test theory, predict | Formal propositions, measurement, hypothesis tests, generalization from samples |
| Interpretive | Socially constructed through meanings | Understand meanings and contexts | Accounts of participants, in-depth case studies |
| Critical | Historically constituted, shaped by power | Critique and transform the status quo | Analysis of contradictions, conflicts, and structures of domination |
Orlikowski & Baroudi (1991) analyzed information systems research published between 1983 and 1988 and found that the field was dominated by the positivist paradigm. They argued for paradigmatic pluralism: interpretive and critical research address questions that positivist research cannot answer, and each paradigm has its own criteria for good research.
Since then, the field has diversified, and further positions have gained ground. Critical realism, for example, assumes a reality independent of our knowledge of it, but holds that the mechanisms producing observable events are often unobservable and must be inferred (Wynn & Williams, 2012). For a manuscript, the point is consistency: a paper can follow any paradigm, but its question, method, and claims have to fit the paradigm it follows.
Methods Framework
| Direct observation of object reality | People’s perceptions of object reality | Artificial reconstruction of object reality | |
|---|---|---|---|
| Axiomatic | Reason, logic, theorems; normative and descriptive modeling | ||
| Logical positivist / empiricist | Field studies, field experiments | Structured interviewing, survey research | Prototyping, physical modeling, laboratory experimentation, simulation |
| Interpretive | Action research, case studies | Historical analysis, Delphi, intensive interviewing, expert panels, futures and scenarios | Conceptual modeling, hermeneutics |
| Critical theory | Introspective reflection |
Meredith et al. (1989) classify research methods along two dimensions. The rational-existential dimension (rows) captures the epistemological structure of the research, from axiomatic reasoning to critical theory. The natural-artificial dimension (columns) captures the source of information, from direct observation of reality through people’s perceptions to artificial reconstructions such as models and simulations.
The framework helps to locate a manuscript and its conversation: which cells do the dominant methods of the conversation occupy, and which cell does your manuscript occupy? A manuscript in a cell the conversation rarely uses can make a methodological contribution, but it also has to explain its approach more carefully to reviewers.
Alignment
Paradigm → logic → method → claim → conversation
- The paradigm determines which claims count as knowledge.
- The logic of theory development determines which kind of claim the study makes.
- The method has to be able to produce that claim.
- The claim has to match what the conversation recognizes as a contribution.
Misalignment is a frequent reason for rejection, even when reviewers do not name it as such. Typical forms:
- An interpretive case study that claims statistical generalization (“our findings show that firms …”).
- A survey study that claims to explain processes it did not observe.
- A manuscript whose method is sound but alien to its conversation, without an explanation of why the conversation should accept it.
- Constructs defined in one paradigm and measured in the spirit of another.
Alignment does not require following the dominant approach of the conversation. It requires that the manuscript makes its choices explicit and that each element supports the next.
Homework
Note the strongest objection from the hot seat and one sentence on how your manuscript will address it. Bring both to the afternoon session; they enter your response matrix and your publication plan.