The IT Artifact Is Not Optional

Fifty-six percent. Sarker et al. (2019) coded the IS research literature and found that 56% of published papers are what they call Type I, meaning purely social. The technology is present as a setting, not as a theoretical actor. All the constructs come from psychology, management, or organizational behavior. You could pick up the entire research model, drop it in a non-IT context, and nothing would change. Only 13% reach Type IV, the category where social and technical factors genuinely interact and where IT does real theoretical work. I read that number three times. More than half the field publishes work where the IT artifact is decorative.

Benbasat and Zmud (2003) named the problem first. They argued that IS has drifted away from its core, toward generic organizational and behavioral topics that reference disciplines study better. Their remedy was simple: put the IT artifact back at the center. Stop publishing papers that happen to use an IT setting but could have been published in a management journal without changing a word. The paper sparked a decade of debate, and the debate produced a lot of words and not a lot of movement in the numbers.

Orlikowski and Iacono (2001) gave us the diagnostic tool. They classified how IS papers treat the IT artifact into five views. The Nominal view names IT but never theorizes it. The system appears in the title or the context section, then vanishes from the model. The Computational view focuses on algorithms and processing logic, treating IT as a technical object divorced from organizational life. The Tool view treats technology as an instrument that serves user goals, which sounds reasonable until you realize the technology itself is underspecified. The Proxy view substitutes IT for something else, using "ERP implementation" as a stand-in for organizational capability, which confuses the artifact with what it represents. The Ensemble view is the one that actually works. It treats IT as embedded in social practice, tightly linked with human action, theorized as part of the causal story. Orlikowski and Iacono found that 88% of the ISR papers they reviewed did not take the Ensemble view.

That number stuck with me. 88% not Ensemble. 56% Type I. These are different papers, different coding schemes, different years, and they tell the same story. The IT artifact is named, mentioned, situated, and thenignored in the theoretical machinery of most IS research.

I think of what Sarker et al. call the Transport Test as the sharpest version of this diagnosis. Take your entire research model and transport it from an IT context to a non-IT context. Move it from information security to physical security at an elementary school. If nothing in the model needs to change, your paper is Type I. The IT was never doing theoretical work. It was wallpaper. This test is brutal because it is so easy to apply. I have started running it mentally on every paper I read now, and the failure rate is high.

Three practical reasons explain why most papers cluster at Type I. Survey data from human subjects is easier to collect than technical instrumentation. Borrowing a theory from psychology or management takes less effort than developing a new one that accounts for the artifact. And these papers are easier to publish because reviewers and editors already know the template. The result is a discipline that produces knowledge portable to any other field, which sounds like a strength but is actually an identity problem. If your findings work identically in a school, a hospital, and a tech company, you have not discovered anything specific to information systems.

Sarker et al. also identify three dangers of losing the sociotechnical axis. Uneven emphasis: the field over-concentrates at Type I and Type VI, leaving the middle sparse. Limited relationship variety: when social and technical factors do appear together, researchers almost exclusively study fit or joint optimization, ignoring contextual, inscribed, imbricated, disharmonious, and role-reversal relationships. And instrumental focus: approximately 91% of IS research studies efficiency and productivity outcomes. Only 9% considers humanistic outcomes like well-being, fairness, or ethics. That last number is what makes me think the problem is not just about theory but about what the field values. I wrote about why "use" is the wrong construct for agentic AI, and the same flattening happens here. When every relationship is fit and every outcome is efficiency, the field is not studying sociotechnical systems. It is studying social systems with a technology prop.

Faulkner and Runde (2019) give us a way to think about what the IT artifact actuallyis, not just what it is not. A digital object is defined by having a bit string as one of its components. That bit string sits on material bearers, servers, chips, drives. Most digital objects are hybrid: they have both material and non-material components. But Faulkner and Runde push further. Digital objects occupy social positions. A social position is a set of rights and responsibilities that exists before the object arrives. When you design an app, you are designing it to fill a position that already exists in the social structure. The position carries expectations about what the object should do, what rights it has, and what obligations it holds. Positions can atrophy when the object becomes obsolete, like floppy disks. They can mutate, like the camera position when digital photography arrived. New positions can emerge, like fitness trackers creating an activity monitor position that never existed before. This matters because it connects the technological specification of an artifact to its social role. The bit string is not just code. It is a thing that occupies a position in a structure of rights and responsibilities.

Burton-Jones et al. (2021) offer four strategies for moving forward. Replace means introducing new constructs to substitute for inadequate ones. Reformulate means applying new logics to existing constructs without abandoning them. Extend means adding boundary conditions to existing theories for new contexts. Envision means adopting entirely new ontological assumptions about how the world works. These are not methodological pleas to study technology more. They are concrete theorizing moves. When I wrote about delegation replacing use for agentic systems, I argued that Baird and Maruping (2021) are doing Reformulation, not Extension, because the foundational assumption changes. The old assumption is that the human acts and the system is operated. The new assumption is that the system has agency and the human manages the transfer of tasks. You cannot extend TAM to cover that. You have to reformulate the construct.

The identity question is simple to state and hard to resolve. What makes IS research different from organizational behavior? If the answer is "we study technology," then the technology has to show up in the theory, not just in the setting. Orlikowski and Iacono gave us the five views and showed that most papers fall short. Benbasat and Zmud named the drift. Sarker et al. quantified it and gave us the Transport Test. Faulkner and Runde gave us a way to make the artifact specific instead of vague. Burton-Jones et al. gave us strategies for theorizing forward. The productivity paradox also lives in this space: we keep measuring IT spending without measuring whether the artifact actually changes how work gets done.

The 56% number is not a curiosity. It is a diagnosis. The Transport Test is not a thought experiment. It is a standard. And every time I run it on a paper and the model survives the transport, I know I am reading organizational behavior with a Wi-Fi password.