What constitutes the activity of “researching MIS.” What ought to constitute the activity of researching MIS? How we answer these questions depends, in turn, on how we answer some logically antecedant questions: What is MIS? What is research? What d

What constitutes the activity of "researching MIS"? What ought to constitute the activity of "researching MIS"? In this essay I will venture some answers to these two questions.

The answers to these questions will depend, in turn, on answers to some related questions:

Certainly, there would be differences in how different MIS researchers answer these questions – and the differences themselves would all be significant to any effort aimed at empirically establishing what "researching MIS" is. In this essay, I contribute to such an effort by offering my own answers to these questions.

An answer to the question, "What constitutes ‘researching MIS’?," depends on what we mean by "MIS" in the first place. And what we mean by "MIS" is, in turn, a non-trivial matter because its empirical referent is itself always in a state of change.

MIS begins where computer science ends (Lee, 1991a). Computer scientists deserve accolades for developing and delivering ever more advanced forms of information technology – hardware technology, software technology, and network technology. However, because no technology implements itself, there is more to MIS than just information technology. In a sense, an information system is an instanciation of information technology, where the same information technology can be instanciated in different ways. One dimension of MIS, therefore, is that it involves not just information technology, but also its instanciation: there are the rich organizational and political processes whereby a given set of information technology is instanciated and there are the rich organizational and political processes pertaining to the continual managing, maintaining, and changing of the information technology so as to sustain the instanciation.

Robert Zmud, Editor in Chief of MIS Quarterly, refers to another commonly accepted dimension of MIS in his description of MIS Quarterly’s Theory and Research section. He describes MIS research, at least as published in MIS Quarterly, as "either enhancing existing theory or building new theory regarding the management of information systems in organizations" (1995: v). A significant element in his characterization is that (unlike computer science) MIS involves the information system and the organization. I take this a step further: an information system and its organizational context each have transformational effects on the other. They are more like the reagents that react to and change each other’s properties in a chemical compound than the inert elements that retain their respective properties in a chemical mixture. The reactive process is what leads some MIS researchers to describe MIS phenomena as "emergent" (Markus & Robey, 1988). In the same spirit, socio-technical systems theory makes the claim that separate efforts to optimize the technical system alone and the social system alone will not only lead to a global suboptimum, but can even be infeasible in the first place. Equivalently stated, the idea is that the same information system can be a success in one organization but a failure in another, while the same organization can experience success with one information system but failure with another; hence, the information system and the organizational context must be studied, understood, and managed together, not separately. Overall, therefore, another dimension of MIS is that it involves, as reactive and inextricable elements, both an information system and its organizational context.

A subset of the previous dimension is so significant that it deserves its own classification as a separate dimension. It derives from information technology’s being an intellectual technology, as opposed to an industrial technology (Curley & Pyburn, 1982). The latter, like a drill press or a steam engine, typically has a fixed set of functionalities. The former, however, has functionalities that are not fixed at the outset, but can be innovated endlessly, depending on its interaction with the intellect of the human beings who implement and use it. Information technology is an intellectual technology. In the process of its instanciation into one or another information system, a given set of information technology becomes subject to the shaping influence of the intellects of its implementors and users, who can end up creating an information system that the inventors of the supporting information technology never had in mind. Even more important, the information technology, once instanciated, can react with and extend the intellects of its implementors and users, who can then turn their transformed intellect to innovating even additional functionalities for the information technology that, once instanciated, can again react with and extend the intellect of its implementors and users, who can then...

Implicit across the three dimensions above, but worthy of being made explicit, is that no MIS researcher is or even should be an objective, disinterested scientist. MIS researchers seek to contribute to the documentation, innovation, or illumination of better ways in which people in organizational contexts can use, manage, and maintain (or, in short, "instanciate") information technology. Unlike some behavioral scientists who seek to avoid Hawthorne effects, MIS researchers want Hawthorne effects (Jönsson, 1991: 389); we want our observations and theories ultimately to make a difference, in good ways, to what it is that we are observing. MIS involves what our own research subjects themselves consider to be a profession and corporate function. Unlike the situation for most behavioral scientists, the relationship between MIS researchers and their research subjects is that the latter are constituents to whom the former are responsible and, to some extent, must serve. Without MIS as a profession or corporate function, there would be no raison d’être for MIS research.

As a whole, these four interrelated dimensions of MIS – that MIS involves not just information technology, but also its instanciation; that MIS involves, as reactive and inextricable elements, both an information system and its organizational context; that MIS involves information technology as a form of intellectual technology; and that MIS involves the activities of a profession or corporate function – are integral to the essence of what "MIS" is. Clearly, the four mentioned dimensions (there are likely more) differentiate MIS from, and move it beyond, the empirical domains traditionally and appropriately researched by computer science, organizational behavior and theory, economics, operations research, and other so-called "reference disciplines." (Indeed, because of this, I no longer view them as reference disciplines, but instead, at best, as "contributing disciplines.") As for what form "researching MIS" currently takes and what form it ought to take, especially with regard to the four dimensions – I may appropriately venture an answer only after attending to another antecedent question.

In this section of the essay, I shall momentarily suspend discussion of "MIS" and focus instead on "research." Numerous research traditions were already in existence long before MIS ever came into being and certainly before MIS researchers began to draw upon them. A discussion of research traditions is therefore appropriate.

Typically, definitions of research refer to formal procedures, such as those of statistical inference, interviewing, construct validation, and so forth. Research certainly involves these procedures, but a point requiring emphasis is that all research is a "social construction" (Berger & Luckmann, 1967). To plunge us into what I mean by social construction, I begin by taking us to the philosopher and sociologist Alfred Schutz’s discussion of multiple realities (1973: 207), in which he refers to the work of the psychologist William James:

Certainly, "research" would include the world of science and the world of ideal relations, though it is not restricted to these two worlds. However, before I add to the list of worlds and subworlds that "research" includes, it would be instructive to examine one such world in detail so as to illuminate what I mean by "social construction."

I have argued (1994) that Euclidean geometry (which is a subworld of the world of ideal relations, mentioned above) is a socially constructed reality (cf. Schutz, 1973: 127-128). The world of Euclidean geometry does not exist in the physical world of nature. It is, strictly speaking, a fiction; it is something that people have constructed. Furthermore, the people who carry knowledge of it, including its originators, can come and go, but the object that we call Euclidean geometry persists. Like a physical object, the world of Euclidean geometry can retain the same form across the different individuals who encounter it. In this sense, the world of Euclidean geometry is objective, not subjective. It is an objective, socially constructed reality.

Moreover, suppose I am reading a Euclidean research paper. Its meaning is not restricted to the Euclidean argument that its author is making, but also involves the entire socially constructed apparatus that comprises Euclidean geometry – its axioms, theorems, symbols, and logic, all of which transcend what is in the research paper and all of which were in existence prior to the author’s writing of the research paper. The research paper itself is just one possible instanciation of this apparatus. Upon grasping this socially constructed reality, I become Euclidean myself and I become able to identify any inconsistencies and suggest improvements in the research paper, thereby even transcending the author’s own understanding. This means not only that I can appropriate the research paper or what the author had in mind, but also that the research paper and the socially constructed world behind it (here, the world of Euclidean geometry) can appropriate me. Upon this appropriation, I am no longer an independent individual exercising free will (if indeed I ever was), but I become an agent of the socially constructed world of Euclidean geometry. Other individuals, placed in the same situation as me, would be likewise appropriated and transformed. We could then (and only then) engage one another in a dialogue about this Euclidean research paper, collectively enact some meanings for it, and — after some time, effort, and perhaps struggle — reach a shared understanding of it. Moreover, rather than being agents under the total control of the socially constructed Euclidean world that has appropriated and transformed us, we could still, as a group, eventually render change in this world, perhaps by contributing to its collection of theorems or by offering resolutions to any logical inconsistencies or gaps that we identify in it.

Note that, in the above discussion, we can substitute "physics," "mathematics," "experimental and quasi-experimental design," "sociology," "economics," "anthropology," etc., for "Euclidean geometry," and the argument would be the same. The point is that every research world – both the product of research (a body of knowledge) and the process of research (individual and group activities that involve application of research procedures) – is socially constructed. No research procedure (which I see as a form of technology) applies itself, but necessarily involves individuals, their activities, and their institutions in its instanciation. (There are additional parallels between research procedures and information technology to which I will return in the essay’s concluding section.) A person appropriates a research world while the research world appropriates the person, with the result that the person becomes an agent of the research world while the research world becomes susceptible to advancement by those who have appropriated it. Research, therefore, is not an entity that has an existence independent of knowing subjects; research (even research in the natural sciences) is a human creation and social activity.

With regard to the natural sciences, I reject the notion that the laws of nature govern nature. I believe that the "laws of nature" have no existence independent of human beings, but instead are constructed and continually revised by human beings so that, as human-made fictions, the "laws of nature" end up conforming to nature, rather than vice versa. Because of the fictional status, the term "theory" is more appropriate than "law." I believe that this argument is no less true for the social sciences. It is the behavior of society (which includes people, groups, and organizations) that governs the human-made fictions known as the theories of social science; it is hardly the case that there are any "laws" or even theories of social science that govern the behavior of society. Research, in my view, consists of individual and group activity aimed at the construction of theories, whether they are theories about nature or theories about individuals, groups, organizations, and society. As such, theories are only theories: they are fictions; they are social constructions. They are not already "there," waiting to be discovered; rather, we invent them.

Researchers interested in investigating MIS phenomena have drawn primarily from just two already existing worlds of research. They are the research worlds of positivism and interpretivism.

1. The World of Positivism

Among social scientists, positivism is often known as the "natural-science model" of research. A social scientist who appropriates the world of positivism proceeds to implement, in his or her own research, an image (as held by followers of positivism) of how research proceeds in physics, biology, and other natural sciences. Descriptions of the world of positivist research sometimes use the terms "ontology" and "epistemology" (and also "methodology," as distinct from "method"); these terms are from the philosophy of science, which itself originated and advanced positivism as a model or an account of what science is. However, I will avoid these terms in this essay. In my experience, most researchers who do positivist research do not know what these terms mean, nor do their actions suggest that they are putting into use the concepts underlying these terms. Furthermore, the positivist model of what science is, at least as articulated by positivist philosophers of science, has been shown to be infeasible and thereby been discredited.¹ (In this light, it is ironic that positivism remains the predominant world of research among scholars not only in academic departments of MIS, but other fields of management studies and social-science research as well.) Therefore, in my characterization of the world of positivism, I will refer more to positivism-in-practice than to the formulation of positivism by the philosophy of science.

In practice, positivism’s natural-science model of social-science research involves three interrelated sets of logic: the rules of formal logic, the rules of experimental and quasi-experimental design, and the rules of hypothetico-deductive logic. The natural-science model maintains that natural-science research complies with all three sets of rules. Social-science research, in complying with the same three sets of rules, would be following the natural-science model.

Rules of Formal Logic In the natural-science model, mathematics is the ideal form in which a researcher may express a theory. A researcher may use the rules of mathematics (the rules of algebra provide an excellent illustration here) not only to relate different propositions to one another in her construction of a theory, but also to derive new theoretical propositions from existing ones. Expressing a theory in mathematical form thereby facilitates a researcher’s efforts to bring about the internal logical consistency of the theory. A researcher could also apply the rules of mathematics for the purpose of checking the validity of the derivations and other operations already performed. Where a researcher expresses a theory in words instead of in mathematics, the researcher may still use the rules of formal logic (of which the rules of mathematics are a subset) for the same purpose; syllogistic logic (e.g., "All men are mortal," "Socrates is a man," "Socrates is mortal") exemplifies this.

Whereas theoretical propositions can be expressed verbally, theoretical propositions expressed mathematically are more easily managed, if only for the reason that researchers are more likely to be already familiar with the rules of mathematics (particularly algebra) than with the rules of formal logic in general. The easeful use of mathematics as a tool for rendering logical consistency in a theory is a likely reason that academic journals in management and other social-science disciplines can give the impression, to practicing managers and other non-academic researchers, of being journals of applied mathematics.

Logical consistency, while a necessary quality of a theory, is not enough to make a theory. Consider that systems of non-Euclidean geometries and systems of imaginary numbers are logically consistent, but do not (and need not) have any empirical referents. In addition to being logical, a theory must also be empirical.

Rules of Experimental and Quasi-Experimental Design Overall, this second set of rules pertains not so much to how a researcher may properly relate a theory’s propositions to one another (which still needs to be accomplished for the theory to be logical), but to how a researcher may properly relate a theory’s propositions to the "real world" or empirical referent that the researcher is crafting the propositions to explain (so that the theory may also be empirical).

In the world of positivism, researchers often express theories in terms of independent and dependent variables. A theory could posit that as an independent variable X (say, user friendliness of an information system) increases, the dependent variable Y (say, system usage) increases. Of course, this raises the issue of how to "operationalize" a measure in the first place (for instance, how might a researcher establish that his or her way of measuring X is indeed a good measure of the user friendliness of an information system?) – an issue for which Straub (1989) provides an excellent treatment. Still, when a researcher empirically tests the theory and observes that, when X increases, Y does too, how can the researcher (and her colleagues) be sure that this was not simply a coincidence? Perhaps sometimes Y will increase even when X does not; perhaps, over time, Y increases by itself anyway, regardless of whatever changes do or do not take place in X. Or perhaps Y does not always increase when X does, but only when an increase in X is accompanied by the presence of another factor, Z. Indeed, perhaps the researcher should posit that it is changes in Z, rather than changes in X, that are associated with changes in Y. In short, when empirically testing for relationships theorized to exist among different factors, a researcher working in the world of positivism must somehow remove, control, or otherwise account for the potentially confounding influences of all other factors. Laboratory experiments (ideally, involving a researcher’s varying of just one factor in a "treatment group" while also involving the researcher’s holding constant of other factors in a "control group"), statistical experiments (involving "statistical treatments" and "statistical controls" implemented through such multivariate tools as regression, LISREL, and PLS), and natural experiments (involving a researcher’s identification, in the field, of factors that are naturally held constant while other factors naturally vary) all exist for this purpose. Whatever form of experimentation, the rules of experimental and quasi-experimental design (Campbell & Stanley, 1963; Cook & Campbell, 1979; Nagel, 1979) help the researcher to ensure that an empirical test of a relationship theorized between particular variables is indeed an empirical test of just this relationship and no other.

The need to make theories empirical, and not just logical, is the reason that the research published in academic journals in management and other social-science disciplines routinely employ statistical experiments and laboratory experiments. Statistical experiments are common also for the reason that they obviously involve mathematics, hence facilitating the positivist requirement of logical consistency in a theory. Furthermore, the popularity of statistical experiments is also explained, in part, by the popularity of surveys and questionnaires, which readily supply the data needed to conduct statistical experiments.

The Rules of Hypothetico-Deductive Logic In the world of positivist research, does a researcher test a theory through induction or deduction? I argue that it is the latter.

Using induction, a researcher would generalize across "n" observations in order to arrive at a theory. Proponents of induction maintain that the greater "n" is, the more general or "generalizable" the resulting theory. As common sensical as this inductive procedure may appear, it is flawed. The flaw is that induction itself is not empirically justifiable. Any attempt to establish the empirical validity of induction ultimately applies induction itself, thereby leading to an infinite regress in reasoning (Popper, 1968, citing Hume). In particular, is the proposition, "inductive inference leads to valid theories," itself empirically valid? To justify it, we would marshal "n" examples in which inductive inference leads to valid theories. However, this would create a new problem: we would be applying inductive inference to justify itself. Then, to justify this application as valid, we would marshal "n" examples of such applications... Hence, whereas induction may be useful for generating a theory, it cannot contribute to the testing or justification of the theory. Even the discipline of statistical inference has distanced itself from the notion of induction. For instance, students in introductory statistics courses learn that the action of increasing a sample size does not increase the probability that a statistically inferred proposition (such as a confidence interval) is correct; instead, increasing a sample size would only have the effect of enhancing what statisticians call the "level of confidence" or "statistical significance" that they associate with the sample.

I have argued (1989) that the logic of research in the natural-science model is not inductive, but deductive. In his classic text, Copi describes the hypothetico-deductive logic in use by the natural sciences (1986: 483): "Few propositions of science are directly verifiable as true... For the most part they concern unobservable entities, such as molecules and atoms, electrons and protons, chromosomes and genes." As a result, justification in natural-science research is indirect rather than direct. "The pattern of indirect testing or indirect verification consists of two parts. First, one deduces from the proposition to be tested [the proposition being the theory] one or more other propositions capable of being tested directly [these later propositions being the predictions]" (p. 486). In the terminology of logic, a theory’s predictions are its conclusions. "Then these conclusions are tested and are found to be either true or false." The researcher then compares what the theory predicts and what she actually observes. "If the conclusions are false, any proposition that implies them [namely the theory] must be false also. On the other hand, if the conclusions are true, that provides evidence for the truth of the proposition being tested, which is thus confirmed indirectly" (p. 486). Note, however, that when the conclusions or predictions turn out to be true, the truth of the theory (from which the prediction originated) is not proven, but is only "corroborated," "supported," or "confirmed" in the instance of this single test. A different instance, involving a different set of conditions in another experiment, would result in yet additional predictions, which would open up the same theory to yet another opportunity for its falsification. Thus, in positivism’s natural-science model, the ever-present possibility for contradictory evidence to surface in a subsequent test requires that a theory always be considered falsifiable. I regard the widespread characterization of theories, even in the social sciences, as falsifiable, testable, refutable or disconfirmable, as an indication of the widespread extent to which the deductive testing of theories is practiced.

In a hypothetico-deductive framework, a researcher can achieve generalizability in a theory not so much by increasing the number of observations or data points, but by increasing the variety of the experimental tests to which he or she subjects the theory. The more varied the sets of experimental conditions in which the theory can predict correctly, the more generalizable the theory. It is the greater range in experimental conditions, rather than the greater number of observations per se, that leads to greater generalizability in a theory.

Summary Positivism is, in practice, a socially constructed world, populated with social-science researchers whose shared beliefs include what I have portrayed and described, above, as the three sets of interrelated logic. A researcher who studies and appropriates what the three sets of logic refer to would, in turn, be appropriated by the world of positivism and thereby become its instrument or agent. The result is that this researcher, and others who likewise experience the appropriation, end up inventing theories that conform to positivism’s natural-science model of social-science research.

[Return to beginning of this section, "What is Research?"]

2. The World of Interpretivism

With Ojelanki Ngwenyama, I have argued (forthcoming) that, whereas the world of positivism has shaped much social-science research, its allegiance to natural-science research as the normative model for social-science research has limited its applicability. A dimension of the subject matter that social scientists examine, but not the subject matter that natural scientists examine, is what the field of phenomenology calls the "life world," which refers to, among other things, the world of consciousness and humanly created meanings. "Unlike atoms, molecules, and electrons, people create and attach their own meanings to the world around them and to the behavior that they manifest in that world" (Lee, 1991b: 347, referring to Schutz, 1973). Atoms, molecules, electrons, and other objects of natural-science inquiry do not "mean" anything to each other (Schutz, 1973: 59). However, people – who are integral to the subject matter of the social sciences – do mean something to each other. In this way, the world of humanly created meanings, however "subjective" they may be, is an integral part of the subject matter that the social scientist is studying. Because of this, "the social scientist must collect facts and data describing not only the purely objective, publicly observable aspects of human behavior," which constitutes a subject matter to which the procedures of positivism’s natural-science model are indeed well suited to studying, "but also the subjective meaning this behavior has for the human subjects themselves," which constitutes a different subject matter whose study requires procedures that have no counterparts among those of the natural sciences (Lee, p. 347). In contrast to the world of positivism, the world of interpretivism gives explicit recognition to the "life world." Not originating in the natural sciences, interpretivism involves research procedures such as those associated with ethnography (from anthropology), participant observation (from sociology), history, and hermeneutics, all of which give explicit recognition to the world of consciousness and humanly created meanings. In most interpretive approaches, a central idea is "mutual understanding" – the phenomenon of one person’s reaching an understanding (i.e., "interpreting") of what another person means, whether it is a person engaged in everyday life taking a "natural attitude" (Schutz, 1973: 59, 208) to understanding another person in everyday life or it is a person engaged in scientific research taking a calculated "scientific attitude" (Schutz, 1973: 63, 249) to understanding everyday people in their everyday lives.

Unlike in the world of positivism, a researcher in the world of interpretivism avows that the researcher herself is the instrument of observation. Sanday writes (1979: 528): "Fieldworkers learn to use themselves as the principle and most reliable instrument of observation, selection, coordination, and interpretation. Ethnography as Metraux... says, ‘depends on this highly trained ability to respond – and to respect that response – as a whole person’" (emphasis added). In the world of positivism, researchers typically write their papers in the third person or the passive voice, as if their knowledge could exist independently of knowing subjects; in the world of interpretivism, researchers more often write their papers in the first person, making no pretense that their interpretive knowledge is inextricably linked to themselves as the human instruments of observation. Indeed, to the extent that a researcher seeks a mutual understanding of another person, the researcher may, of course, do so only as a person; in other words, the researcher must use himself as an instrument of observation.

Of the different interpretive approaches (for bibliographies, see Myers, 1996), I am most familiar with hermeneutics. I will turn to hermeneutics in this discussion to bring out some concepts that I believe are common across different interpretive approaches.

Hermeneutics originally referred to the interpretation of ancient texts, such as the Bible. Today, interpretive scholars take a hermeneutic approach to the interpretation of human and organizational behavior. The motivating idea behind hermeneutics is that reading a text provides the model for reading human behavior. In other words, the process of achieving an understanding of a text can provide insights to the process of achieving a mutual understanding with another person. In this model, people and organizations are a text, which we "read."

To illustrate this model, consider the above set of ten words. In the first sentence, the ten words mean one thing. In the second sentence, the same ten words mean another thing. And in the third sentence, the ten words mean something else entirely. The same word and even the same set of words can end up meaning different things. The meaning of an individual word and the meaning of the sentence as a whole are mutually dependent and, as a reader, I form my understanding of both simultaneously.

I apply the same model to human behavior. The same publicly observable behavior can have different meanings in different organizational arrangements. The meaning of an individual action and the meaning of the organizational setting as a whole are mutually dependent and, as an interpretive researcher, I form my understanding of both simultaneously.

Modeling interpretation on reading is useful for bringing out three points. First, interpretation, like reading, is not a mysterious, esoteric activity (Boland, 1985). This model promises the following: if we can read, we can interpret. Second, based on our experience in reading, we are already familiar with some of the basic elements involved in interpretation: meaning; its expression by one person; its understanding by another. Most important of all, our everyday practice of reading is, in itself, a demonstrative proof that interpretation works. Interpretation is "do-able," no less than reading is "do-able."

I have encountered at least twelve different, serious definitions of hermeneutics.² I will present just one hermeneutic approach that I derive (c.f. Davis et al., 1992) from the work of the historian of science, Thomas Kuhn. Consider the situation where I have a text whose author, and the audience for whom the author wrote, are from a time and culture far removed from my own. I have a dictionary and I know the grammar. How do I proceed? Kuhn offers the following advice (quoted in Bernstein, 1983: 132, emphasis added): "When reading the works of an important thinker, look first for the apparent absurdities in the text and ask yourself how a sensible person could have written them. When you find an answer... when those passages make sense, then you may find that more central passages, ones you previously thought you understood, have changed their meaning."

Accordingly, when I encounter a passage that I find absurd or irrational, I have a choice to make. Either I may draw a conclusion about the author – that she thinks in a way that she herself would consider absurd – or I may draw a conclusion about myself – that I have not succeeded yet in figuring out what the author meant. When doing a hermeneutic interpretation, I choose the conclusion about myself. A basic premise in this hermeneutic approach is that the author knows what she’s talking about.

The same premise applies when I am reading human behavior. Suppose I observe a systems developer who professes the value of persuading managers to use information technology, even though I see that he is resistant to using CASE tools in his own work. Either I could draw a conclusion about the subject – namely, that this person is acting in a way that he himself would consider absurd and irrational – or I could draw a conclusion about myself – namely, that I have not succeeded yet in figuring out what the subject himself means by his own behavior. Whether taking a hermeneutic approach in particular or an interpretive approach in general, I choose the conclusion about myself. A premise in hermeneutic and other forms of interpretation is that people know what they’re doing. Rosabeth Moss Kanter has expressed this more elegantly: "With Michel Crozier, I wanted to demonstrate that everyone is rational, that everyone within an organization, no matter how silly or irrational their behavior seemed, was reacting to what their situation made available, in such a way as to preserve dignity, control, and recognition from others" (Kanter, 1977: 291). What initially appears to me, in my role as an interpretive researcher, as irrational behavior therefore might not be irrational at all; it could instead be a rational response by rational people to the irrational conditions of their situation. In such cases, the task of the interpretive researcher is to discern, if not the rationality, then at least the rationale, behind apparently irrational behavior.

Returning to the situation where I have encountered a passage that makes no sense to me, how do I, as a reader, proceed? Again, this is nothing mysterious or esoteric. I would return to the text, keep on reading, and seek out more clues. Then I would try out alternative readings of the puzzling passage until I would see how a sensible person could have written them. But then, as Kuhn says, when this passage makes sense, other passages which I thought I had understood before might now change their meaning and create new puzzles. To resolve these new puzzles, I would return to the text and repeat this process.

This reasoning moves in a circle, but the reasoning is not circular. Palmer states (1979: 87): "We understand the meaning of an individual word by seeing it in reference to the whole of the sentence; and reciprocally, the sentence’s meaning as a whole is dependent on the meaning of individual words." This interlocking system, from which meaning emerges, is well known; it is the hermeneutical circle – which I demonstrated above in the diagram with the ten words and three sentences.

Correspondingly, in the situation where I am reading human behavior – in particular, where I have just observed a person acting in a way that seems absurd or irrational – how do I, as an interpretive researcher, proceed? I would return to the text itself – the text of this person’s actions, as well as the whole of the organizational setting of which these actions are a part. And I would use the idea of the hermeneutical circle here: I would come to understand any single action by relating it to the whole of the organizational setting; and reciprocally, I would come to understand the whole of the organizational setting by relating it to the individual actions. Just as when I am reading the text of a book, I would continue reading the text of an organizational setting and its individual actions until the apparent absurdities dissolve.

There are two more points regarding hermeneutic interpretation that require attention. First, whether the text I am reading is a book or a person’s behavior, I could extend the hermeneutical circle to the social context immediately surrounding the text. For instance, when I am trying to make sense of a biblical passage, I might take advantage of archeological knowledge about the ancient middle east. Likewise, when I am trying to understand the behavior of systems analysts in a corporate headquarters in Boston, I might take advantage of knowledge about the systems profession in the United States. With a larger hermeneutical circle, I would have additional passages on which to draw for clues to make sense of the passages that I find puzzling.

The second point is that, whenever I read, I necessarily approach the words of the text with my own pre-existing "dictionary" of definitions. Whether I am reading a book or I am reading human behavior, I necessarily form an understanding with the help of a pre-existing understanding that I carry with me and apply to the text. The "dictionary" that I bring with me is part of what I am as an instrument of observation.

Summary Interpretivism, no less than positivism, is itself a socially constructed world, populated with social-science researchers whose shared beliefs include the following four concepts. First, the subject matter of interpretive research involves the "life world," which includes humanly created meanings, whether those that are individually held or those that are shared by groups. This is a subject matter whose study requires procedures that have no counterparts among those of the natural sciences. Second, the researcher himself or herself must inevitably serve as an instrument of observation. Third, interpretation is iterative. In a sense, an interpretation is always "in process"; the possibility of encountering a "breakdown" (Agar, 1986) in one’s interpretive understanding (i.e., a newly appearing instance of absurd or irrational behavior, in the light of one’s most recently revised interpretation) is always present and is, in fact, to be welcomed as an opportunity for further refining the interpretation. Fourth, the validity or goodness of an interpretation can be assessed. There are numerous ways to do this (cf. Sanday, 1979: 529). With a good interpretation, any apparently absurd or irrational behaviors would no longer appear so. With a good interpretation, new observations would not surprise the observer. With a good interpretation, new observations would not surprise a different observer, to whom the interpretation has been communicated. With a good interpretation, an observer would be able to enter the organizational world of the observed human subjects, whereupon he or she could communicate with them.

How would I respond to a critic who insists that he or she does not understand or agree with my interpretation, and demands conclusive proof? I would respond that such a demand is inappropriate. After all, when I am reading, the burden is on me, the reader, to reach an understanding of the text. Likewise, if a critic is reviewing my interpretation of human or organizational behavior, the burden is on the critic to reach an understanding of my interpretation (where "understanding" neither implies nor requires agreement). Because the instrument of observation is the person, the failure of a critic to understand is a fault of the instrument, or the critic himself. For the critic who simply does not understand, the political scientist Charles Taylor offers the following advice (1979: 68): "Change yourself." In other words, the instrument of observation itself would require adjustment. Then, if the critic indeed so changes himself, he would become able to appropriate my interpretation (and, in turn, be appropriated by it); this would enable himself to identify any inconsistencies and suggest improvements in my interpretation, thereby even transcending it. This scenario applies to interpretive research in general, not just hermeneutics.

[Return to beginning of this section, "What is Research?"]
[Return to beginning of paper.]

The predominance of the world of positivism among MIS researchers is well known. This predominance received formal documentation by Orlikowski & Baroudi (1991) who noted that 96.8 percent of all the published MIS articles in their sample were positivist, with the rest being interpretive. Since 1991, there have been signs of increasing acceptance of the world of interpretivism by MIS researchers. The editorial board of MIS Quarterly, one of the top MIS research journals, has come to include ample representation from interpretive researchers. The International Conference on Information Systems, the leading annual conference for MIS researchers, routinely includes panels and paper presentations involving interpretive research. A relatively new and highly regarded journal, Accounting, Management, and Information Technologies, has the open editorial policy and practice of welcoming interpretive research. Another encouraging sign is that MIS doctoral students are showing and pursuing interests in interpretive research.

Still, researching MIS through positivism and interpretivism, while not incorrect, is incomplete.

Positivism’s natural-science model, which is of course well suited to studying the physical subject matters found in nature, is ill suited to capturing the four dimensions, discussed earlier, of what MIS is. With regard to the first dimension, positivism is well suited to studying information technology itself (as in the academic discipline of computer science), but the rich aspects of human and organizational instanciations of information technology have no counterparts in the physical subject matters of the natural sciences and therefore are elusive to positivist study. With regard to the second dimension, positivism posits relationships between factors, but the emergent nature of the transformational effects of an information system and its organizational context on each other can make such relationships unforeseeable to a positivist researcher. With regard to the third dimension, information technology’s being an intellectual technology similarly builds unpredictability into how information technology is instanciated, again impeding a positivist researcher’s attempt to foresee, formulate, and empirically test relationships. With regard to the fourth dimension, it is positivism’s natural-science model itself which insists that scientists not "contaminate" or "bias" their subject matter with their presence and their values – an insistence that is difficult, if not impossible, to reconcile with the fact that we MIS researchers have an interest and responsibility to serve our research subjects as our constituents and that we want our observations ultimately to make a difference, in good ways, to them, to their information systems, and to their organizations. Positivist research can and does contribute to explanations and diagnoses of existing MIS phenomena, but is an incomplete research approach in itself for addressing fully the four dimensions of MIS.

The practice of interpretive research can be integrated with the practice of positivist research (Lee, 1991b). Clearly, interpretivism can and does help to address the humanity and sociality that infuses the first three dimensions; however, interpretivism would only help to complete an understanding or diagnosis of existing MIS phenomena and, therefore, would still leave us wanting for a prescription of what ought to be done regarding the people, the information systems, and the organizations we are researching.

In a current research effort, Ojelanki Ngwenyama and I (forthcoming) have examined a research instance of the limitations of positivism and interpretivism. In investigating the richness that occurs in the managerial use of electronic mail, we reviewed the past MIS research on communication richness, which employs both positivist and interpretive perspectives. MIS research that follows positivism’s natural-science model has, predictably, conceptualized communication as if it were a physical process, where humanly created meanings are treated like a physical substance transported from one person to another through a conduit and where the receiving person is conceptualized as nothing more than a passive receptacle that unquestioningly accepts and comes to hold whatever pours or trickles into it. Interpretive MIS research that examines managerial communication has addressed more adequately the phenomenon of humanly created meanings, explicitly acknowledging the concept of "mutual understanding" between knowledgeable, intelligent, and acting human beings. Significantly, neither the positivist research nor the interpretive research has ventured into the realm of entertaining value judgments on the validity or rightness of what the observed managers are communicating in the first place. What should they be communicating, how should they be communicating, and what is the role of information technology in both? Both the positivist research and interpretive research stop short of these considerations; they only offer a portrayal at a distance. However, because MIS is a profession and corporate function, because we MIS researchers must treat the people whom we study not only as research subjects but also as our constituents, and because we MIS researchers want our observations ultimately to make a difference, in good ways, to our constituents, to their information systems, and to their organizations – MIS research ought to include more than just the worlds of positivism and interpretivism.

Certainly, different MIS scholars would offer different answers to the question of what forms "researching MIS" ought to take in the future. One plausible position is that MIS researchers ought to pursue positivism and interpretivism in the manner of pure or basic research, and that such basic research (similar to basic research in physics, astronomy, biology, chemistry, and other natural sciences) will eventually lead to one or another problem-solving application. The position I take is that MIS researchers ought to pursue another form of research that has not received as much attention it deserves: critical social theory (CST).

In my research with Ngwenyama (forthcoming), we note that, in Bernstein’s review of CST (1978: 179-185), he observes that it was the critical theorist Max Horkheimer who, in the 1930’s, coined the term "critical theory" so as to contrast it to "traditional theory," which refers not only to positivist research, but even includes interpretive research (such as the phenomenology of Edmund Husserl). Whereas positivist and interpretive researchers see themselves as observers or on-lookers whose research is completed when they have achieved a sound explanation or understanding of their subject matter, CST researchers believe differently; they believe that no researchers can simply be on-lookers, but that researchers themselves influence, and are influenced by, the social and technological systems they are studying, and that the responsibilities of researchers do not end with the development of sound explanations and understandings, but also extend to the critique of unjust and inequitable conditions in society from which people require emancipation. Excellent reviews of CST already exist both inside and outside the MIS research literature (Alvesson & Willmott, 1992; Held, 1980; Ngwenyama, 1991; Orlikowski & Baroudi, 1991; Tice & Slavens, 1983; White, 1988). However, a difficulty of researching MIS through a CST approach is that it awaits extensive development. Hirschheim & Klein (1994) offer the commentaries that the philosophy of neohumanism, on which CST is built, "is strong on utopian vision but short on principles for implementation" (p. 99) and that "critical social theory does not point to effective ways of handling the darker side of organizational life, which blocks the road to emancipation, in particular the distortions arising from vested interests and power" (p. 100). Hirschheim & Klein can be credited for taking an incremental step towards the needed implementation, thereby establishing a direction in which more such steps need to be taken.

Certainly there are other avenues that researching MIS may also take in the future. Action research (for a bibliography, see Baskerville, 1995), like CST, is a form of research that deserves more attention than it has received. Another potential avenue whose merits I have argued (1991a) is to model MIS research not so much on research in the positivist and interpretive sciences, but on research in the professions, such as architecture.

For a productive future for researching MIS, I believe that the effort of attending to a diversity of research approaches, their limitations, and their potentials is necessary, but insufficient. To conclude this essay, I return to two concepts with which I opened the essay: that our MIS research itself is a social construction and that MIS is distinguished by at least four dimensions.

To highlight the significance of research as being a socially constructed activity, I offer an analogy involving Kling & Scacchi’s classic notion of "the web of computing" (1982) which, in Hirschheim & Klein’s recent characterization, "reveals that information systems are like social institutions in that they are embedded in a complex web of social norms and practices" (1994: 100). Just as we MIS researchers are familiar with the notion of the web of computing, we may also entertain the notion of the web of researching. Our own activities of "researching MIS" involve social institutions (journals, conferences, schools of research thought, funding agencies) and are themselves embedded in a complex web of social norms and practices.

The notion of the web of researching presents us with a challenge. What is it about our social institutions in the socially constructed world of MIS research that encourages some forms of research but not others? The mere recognition of a new, needed form of research by one researcher or even a group of researchers would not be enough to launch the new research. In addition, what social and political actions would be needed to intervene in the web of researching so as to secure the new form of research? In what ways do MIS researchers themselves require emancipation? What are some equitable and emancipatory ways in which the activities of researching MIS may continue or should change, and exactly who would, or should, decide what is equitable and emancipatory?

Rich complications in meeting this challenge arise from the fact that our research procedures themselves are a form of technology. Research technology richly complicates researching researchers, just as information technology richly complicates researching MIS. The same four interrelated dimensions that characterize what MIS is, described at the outset of the essay, also characterize what researching is. First, researching MIS does not involve research procedures alone, but also the rich aspects of human and social instanciations of research procedures. Second, the emergent nature of the transformational effects of our research procedures and our research context on each other can pose difficulties to anticipating, studying, explaining, and understanding our own activities of researching. Third, the fact that research procedures are obviously an intellectual technology further builds unpredictability into how we end up instanciating our research procedures. And finally, MIS research itself is a profession or academic function, where we each have an interest and responsibility to serve one another as colleagues and where we want our self-reflections on our activities of researching MIS to make a difference, in good ways, to one another, to our research procedures, to our research institutions, and thereby eventually to our constituents – people who use information technology in organizations. As rich as these complications are, we must tackle them. If we do not even attempt to tackle them in researching ourselves, how may we profess to tackle them in researching others?

If we are able to investigate the web of computing among people in organizations, we are no less able to investigate the web of researching amongst ourselves. Both logical consistency and the golden rule require that if we profess the validity of research procedures for explaining, understanding, and prescribing for others, we must also profess the validity of the same procedures for explaining, understanding, and prescribing for ourselves. Ultimately, researching MIS requires researching ourselves.

1. Schön (1983: 48) quotes Bernstein (1978: 207): "There is not a single major thesis advanced by either nineteenth century Positivists or the Vienna Circle that has not been devastatingly criticized when measured by the Positivists' own standards for philosophical argument... Whatever one's final judgment about the current disputes in the post-empiricist philosophy and history of science... there is rational agreement about the inadequacy of the original Positivist understanding of science, knowledge and meaning." Schön himself adds: "Among philosophers of science no one wants any longer to be called a Positivist..."

2. See Palmer (1979: 33) for six definitions. See Tice and Slavens (1983: 297) for another six definitions.

Alvesson, M. and Willmott, H. (1992), Critical Management Studies (London and Newbury Park, CA: Sage).

Berger, P. L. and Luckmann, T. (1967), The Social Construction of Reality: A Treatise in the Sociology of Knowledge (Garden City, NY: Anchor Books).

Bernstein, R. J. (1978), The Restructuring of Social and Political Theory (Philadelphia: University of Pennsylvania Press).

Bernstein, R. J. (1983), Beyond Objectivism and Relativism: Science, Hermeneutics, and Praxis (Philadelphia: University of Pennsylvania Press).

Boland, R. J. (1985), "Phenomenology: A Preferred Approach to Research on Information Systems," in Mumford et al. (1985), 193-201.

Campbell, D. T. and Stanley, J. C. (1963), Experimental and Quasi-Experimental Designs for Research (Chicago: Rand McNally).

Cook, T. D. and Campbell, D. T. (1979), Quasi-Experimentation: Design & Analysis Issues for Field Settings (Chicago: Rand McNally).

Curley, K. F. and Pyburn, P. J. (1982), "‘Intellectual’ Technologies: The Key to Improving White-Collar Productivity," Sloan Management Review, 24: 31-39.

Davis, G. B., Lee, A. S., Nickles, K. R., Chatterjee, S., Hartung, R., and Wu, Y. (1992), "Diagnosis of an Information System Failure: A Framework and Interpretive Process," Information & Management, 23: 293-318.

Held, D. (1980), Introduction to Critical Theory: Horkheimer to Habermas (Berkeley: University of California Press).

Hirschheim, R. and Klein, H. K. (1994), "Realizing Emancipatory Principles in Information Systems Development: The Case for ETHICS," MIS Quarterly, 18: 83-109.

Kanter, R. M. (1977), Men and Women of the Corporation (New York: Basic Books).

Kling, R. and Scacchi, W. (1982), "The Web of Computing: Computer Technology as Social Organization," in Yovits (1982), 1-90.

Lee, A. S. (1989), "A Scientific Methodology for MIS Case Studies," MIS Quarterly, 13: 33-50.

—— (1991a), "Architecture as a Reference Discipline for MIS," in Nissen et al. (1991), 573-592.

—— (1991b), "Integrating Positivist and Interpretive Approaches to Organizational Research," Organization Science, 4: 342-365.

—— (1994), "Electronic Mail as a Medium for Rich Communication: An Empirical Investigation Using Hermeneutic Interpretation," MIS Quarterly, 18, 143-157.

Markus, M. L. and Robey, D. (1988) "Information Technology and Organizational Change: Causal Structure in Theory and Research," Management Science, 5: 583-598.

Mumford, E., Hirschheim, R., Fitzgerald, G., and Wood-Harper, A. T. (1985) (eds.), Research Methods in Information Systems (New York: North-Holland).

Ngwenyama, O. K. (1991), "The Critical Social Theory Approach to Information Systems: Problems and Challenges," in Nissen et al. (1991), 267-280.

Nissen, H.-E., Hirschheim, R., and Klein, H. K. (1991) (eds.), Information Systems Research: Contemporary Approaches & Emergent Traditions (New York: Elsevier Science).

Orlikowski, W. J. and Baroudi, J. J. (1991) "Studying Information Technology in Organizations: Research Approaches and Assumptions," Information Systems Research, 2: 1-28.

Palmer, R. E. (1979), Hermeneutics: Interpretation Theory in Schleiermacher, Dilthey, Heidegger, and Gadamer (Evanston: Northwestern University Press).

Popper, K. (1968 edn.), The Logic of Scientific Discovery (New York: Harper Torchbooks).

Rabinow, P. and Sullivan, W. M. (1979) (eds.), Interpretive Social Science (Berkeley: University of California Press).

Sanday, P. R. (1979), "The Ethnographic Paradigm(s)," Administrative Science Quarterly, 24: 527-538.

Schön, D. A. (1983), The Reflective Practitioner: How Professionals Think in Action (New York: Basic Books).

Schutz, A. (1973 edn.), Collected Papers, ed. M. Natanson (The Hague: Martinus Nijhoff).

Straub, D. (1989), "Validating Instruments in MIS Research," MIS Quarterly, 13: 147-169.

Taylor, C. (1979) "Interpretation and the Sciences of Man" in Rabinow et al. (1979), 25-71.

Tice, T. N. and Slavens, T. P. (1983), Research Guide to Philosophy (Chicago: American Library Association).

White, S. (1988), The Recent Work of Jürgen Habermas: Reason, Justice and Modernity (Cambridge and New York: Cambridge University Press).

Yovits, M. C. (1982), Advances in Computers, 21 (London and New York: Academic Press).