Your Paradigm Is Not Neutral

I caught myself doing it again last week. Someone asked whether my study would be qualitative or quantitative, and I started answering about methods before I had said a word about what I believe is real. That is exactly the problem. The method question is downstream. The paradigm question is upstream. And the paradigm you choose is not a neutral tool you pick off a shelf. It carries commitments about ontology, epistemology, methodology, and evaluation that shape what you can see and what you cannot.

Three paradigms organize most of the IS research I read: positivist, interpretive, and critical realist. Each one has its own answer to four questions. What is real? How can we know it? How do we investigate? And how do we judge whether the study was any good?

Positivism assumes an objective reality that exists independently of anyone observing it. You can discover regularities through systematic measurement. The researcher stays detached, knowledge is value-free, and the goal is explanation and prediction. Methods are variance-based: surveys, experiments, SEM, secondary data analysis. You evaluate with reliability, internal and external validity, construct validity, and replicability. Popper (1959) anchors the logic here. He argued that science advances by falsification, not by verification. No number of confirming observations can prove a universal claim, but one counterexample can disprove it. This is his demarcation criterion: a theory is scientific only when it makes predictions that could in principle be refuted. For positivist IS work, this means hypotheses must be testable. Orlikowski and Baroudi (1991) documented that IS research has been dominated by these assumptions for decades. The strength is cumulative structure. The limitation becomes visible when you look at a research stream like TAM or UTAUT: repeated tests of the same nomological net, each adding moderators, look progressive from inside but may be degenerating in Lakatos's sense because each new study mostly accommodates what was already known.

Interpretivism assumes that reality is socially constructed. Multiple realities exist as actors understand and enact them, and those understandings are the primary object of inquiry. The researcher does not stand outside. Knowledge is co-produced with participants. The goal is Verstehen, understanding from the actor's point of view. Methods include ethnography, hermeneutics, grounded theory, and field-based case study. But here is the trap I keep seeing: people equate interviews with interpretivism. That is wrong. You can conduct interviews in a positivist study, with a structured protocol and coded responses that test predefined relationships. The paradigm is not the method. It is the set of assumptions about reality and knowledge that the method serves. Klein and Myers (1999) give the evaluation framework for interpretive IS research with seven principles: the hermeneutic circle, contextualization, interaction between researcher and subjects, abstraction and generalization, dialogical reasoning, multiple interpretations, and suspicion. These are not reliability and validity in a different font. They are a distinct logic of what counts as good work. Sarker, Sarker, and Sidorova (2006) show what interpretive research looks like in IS when they use Actor-Network Theory to read a business process change failure. They do not reduce the failure to variables. They trace it through problematization errors, weak interessement, parallel enrollment, and irreversible inscription of interests.

Critical realism is its own paradigm, and I want to be careful here because the most common mistake is calling it a hybrid of positivism and interpretivism. It is not. CR has a stratified ontology with three domains. The real is mechanisms, structures, and powers. The actual is events generated when mechanisms fire. The empirical is the subset of events we actually observe. Wynn and Williams (2012) explain that observed regularities may not reveal underlying mechanisms because mechanisms can be activated, suppressed, or counteracted by contextual conditions in open systems. Epistemologically, knowing reality is possible but always fallible and theory-laden. CR rejects positivist naïve realism and interpretivist constructionism in equal measure. Methodologically, CR uses retroduction: reasoning backward from observed phenomena to the mechanisms that could have generated them. Mingers et al. (2013) argue that CR gives ontological depth that positivism cannot deliver while still respecting the social construction of knowledge that interpretivism stresses. Evaluation for CR is explanatory power. Does the account identify plausible mechanisms? Does it specify contextual conditions that activate or suppress those mechanisms? Does it show that mechanism plus context could have generated the observed pattern? This is a different question from reliability and validity, and a different question from the seven principles.

The philosophy of science behind these positions has its own lineage. Kuhn (1962) complicates Popper's clean story. Normal science is puzzle-solving inside a paradigm's rules, not constant testing of the paradigm itself. Anomalies build up, crisis follows, and a revolutionary paradigm shift replaces one set of commitments with another. Kuhn calls this incommensurability: paradigms cannot be compared by a neutral standard, because each defines its own evidence and meaning. Lakatos (1970) sits between Popper and Kuhn. A research programme has a hard core that researchers protect by convention, and a protective belt of auxiliary hypotheses that absorbs refutations. A progressive programme predicts novel facts that get confirmed. A degenerating programme only patches up what was already known. I use Lakatos when I need to evaluate whether a research stream is actually moving forward or just defending itself. The long TAM line looks different through this lens. Each new moderator or extension protects the hard core, which is the belief that beliefs shape intentions and intentions shape use. The programme may still be progressive if it generates genuinely new predictions, but it edges toward degeneration when each study mainly explains what was already observed.

Burrell and Morgan (1979) gave us the most cited map. Two axes: subjective versus objective on one side, and regulation versus radical change on the other. Functionalist sits at objective and regulation. Interpretive at subjective and regulation. Radical Humanist at subjective and radical change. Radical Structuralist at objective and radical change. They also push an incommensurability thesis: you cannot combine elements across paradigms because each defines its own assumptions and mixing them produces logical contradiction. Deetz (1996) corrects this map. He argues that Burrell and Morgan describe the field from inside the functionalist paradigm, which naturalizes that position as the default. His alternative uses consensus versus dissensus on one axis, and elite or a priori versus local or emergent on the other. This yields four discourses: Normative, Interpretive, Critical, and Dialogic. The Dialogic discourse is the one missing from the original map, and that absence is the heart of the critique. Dialogic inquiry makes room for users, communities, and grassroots actors who resist, appropriate, and reshape technology in ways that neither functionalist prediction nor top-down critical theory can capture.

Lee (1991) is the bridge paper that has changed how I think about this. He does not collapse positivism and interpretivism into one thing. He stacks three levels of understanding. Subjective understanding is actors' everyday meanings. Interpretive understanding is the researcher's field-based interpretation of those meanings. Positivist understanding is formal propositions assessed with falsifiability, logical consistency, relative explanatory power, and survival. One IS study can move through all three. The paradigms do not merge. Their logics operate at different layers. This is a more careful position than saying paradigms are incommensurable, which would freeze the field into isolated camps that cannot talk to each other. It is also more careful than saying we should all just be pragmatic and pick whatever method works, which risks sliding into unreflective eclecticism where no assumptions are examined.

Goles and Hirschheim (2000) push further with paradigm pluralism and pragmatism. Pluralism means using methods from different paradigms when the research question warrants it. Pragmatism means choosing methods by research question rather than by paradigm loyalty. Their argument is that the paradigm wars in IS have outlived their usefulness. Mingers et al. (2013) give philosophical ground for this: if mechanisms operate at multiple levels, different methods reveal different aspects of reality. The key distinction is between paradigm and method. A case study is a method. It can be conducted inside positivist, interpretive, or CR commitments depending on what the researcher assumes about reality and knowledge. Yin's case studies are positivist. Klein and Myers support interpretive case study. Wynn and Williams support CR case study. The method label does not settle the paradigm question.

I think what bothers me about how paradigms get taught is the flattening. People memorize a two-by-two, learn to say "qualitative is interpretive and quantitative is positivist," and move on. But a case study in the positivist tradition tests hypotheses about causal relationships in bounded contexts. A case study in the interpretive tradition generates thick description of meaning-making in context. A case study in the CR tradition retroducts from observed outcomes to the mechanisms that generated them, with explicit attention to contextual conditions that activated or suppressed those mechanisms. Same method, three different logics, three different evaluation criteria. If you cannot name which paradigm your case study operates in, you have not chosen a paradigm. You have inherited one without examining it.