Groupware fails on social organisation, not on technical capability

CSCW's founding move was to make cooperative work, and the articulation work that sustains it, the unit of analysis. The second of ten units asks what that move leaves open: when a system built to support that work collapses, what caused the collapse? The answer the field reached in its first decade is that the cause is social. Three things predict groupware failure. The first is the distribution of incentives. The second is the critical mass a medium needs before it is worth anyone's effort. The third is the gap between prescribed and actual work. Adoption is therefore an organisational achievement no interface feature can substitute for.

That is not a claim that technology is beside the point. The variables deciding whether a system is used at all, by whom and at what cost lie outside the artefact, in the incentives, interdependencies and routines it meets. The reframing moves leverage from the interface to the organisation, and poses the question the third unit takes up in the signal that does the coordinating: what is left for design to do?

Failure starts with who does the work and who benefits

Grudin's account is where the thesis is built. He reviewed what he called repeated, expensive groupware failures. He argued they were not technical. They were failures to anticipate eight challenges arising from the difference between software written for a person and software written for a group (Grudin, 1994). The first is a disparity between who does the work and who receives the benefit: a shared calendar asks the engineer to maintain a record that chiefly serves the manager. His remedy is organisational as much as technical: design the process of use alongside the technology, and give every participant a direct benefit.

The claim is an argument from accumulated cases rather than a controlled comparison, and every setting it draws on is a working-age office. Markus and Connolly refined it into a distinction. Groupware whose benefit is collective and conditional on other people's use faces adoption problems no single-user design heuristic solves (Markus and Connolly, 1990). That separates a tool which fails ergonomically from one which fails structurally: no usability work repairs a system whose value depends on colleagues with no reason to join in.

A medium that needs everyone is fragile until everyone has arrived

Markus gave the structural failure its name. Interactive media, she argued, have two properties most innovations lack (Markus, 1987). Widespread use creates universal access. That access is a public good, and nobody can be prevented from enjoying it even after contributing nothing. Use also entails reciprocal interdependence, so earlier users are influenced by later users as much as the reverse. A medium with those properties is unusually vulnerable to start-up problems and to discontinuance. Adoption of a discretionary shared medium is not the sum of individual decisions.

If the mechanism is right, one remedy is to remove the discretion, and the field tested that remedy directly. Grudin and Palen asked whether online meeting schedulers in two large organisations needed a managerial mandate before use became universal (Grudin and Palen, 1995). Their interview evidence pointed away from the mandate. Near-universal use arrived without one. Three things carried adoption: a product that was versatile and easy to use, organisation-wide infrastructure, and peer pressure that built among members over time.

Resistance is a position about power, not a defect in the user

Markus set out three theories of resistance. People resist because of factors internal to themselves, because of poor design, or because design features interact with the organisational context of use (Markus, 1983). Her case is a financial system introduced to one division of the Golden Triangle Corporation in 1975. Divisional accountants had kept their own transaction data and reported summaries upward. Afterwards every transaction carried a corporate account code and the system produced the summaries, so control of the data moved to the corporate accountants. They resisted, and difficulty of use was secondary to the real objection, a change to how managerial accounting would be done (Markus, 1983). One accountant who had championed the system at headquarters began resisting it after a transfer to a division, locating the cause in her position rather than her temperament. The case rests on one firm described after the fact, with no comparison site.

Hirschheim and Newman reached the same place from the other direction. Surveying the literature on resistance and working through a case study, they concluded that resistance is a complex phenomenon which defies simple prescriptions (Hirschheim and Newman, 1988). Together the two block the reading that non-adoption is user error: a field that codes resistance as irrationality keeps prescribing training for a problem whose cause is a threatened distribution of work.

The work an interface encodes is not the work people do

Bullen and Bennett analysed several groupware installations. They found the technology rarely used as designed. Adoption depended on fit with the coordination routines already in place, and unanticipated uses and non-uses were the norm (Bullen and Bennett, 1990).

Bikson and Eveland named the process. They argued that groupware is reinvented as it is implemented. A sociotechnical frame explains which reinventions hold: outcomes depend on fit between tool and work group rather than on features (Bikson and Eveland, 1996).

Dourish supplies the warning that keeps this honest. Ethnography is often treated as field investigation whose evaluative criterion is the implications it yields for design, arguing that implications for design may not be the best metric and may fail (Dourish, 2006). A mismatch between design and practice is evidence about how work is organised, so converting it into a feature list discards the analysis.

The same artefact produces different organisations

Barley's study answers what the tool does if it does not determine the outcome. Observing computed tomography scanners enter different radiology departments, he found that the same technology led to different organisational structures, because it provided an occasion for structuring rather than a determinant (Barley, 1986). Structure followed the local negotiation of roles, so the artefact is causally active without fixing what follows.

Orlikowski gave that relationship a theory. Technology is both a product of human action and a medium that structures it. Neither the view of it as an external force nor the view of it as pure strategic choice is complete (Orlikowski, 1992). Her later practice lens went further: people enact structures through recurrent use, so the same artefact produces different structures in different hands (Orlikowski, 2000).

Leonardi supplied the mechanism at the level of routine. Human and material agencies, he argued, are imbricated, so whether a technology affords or constrains action depends on the routines it meets (Leonardi, 2011).

Contested question: does the social account leave design anything to do?

The literature agrees that adoption is social and disagrees about whether anything remains for design. On the first side the levers are organisational. Grudin located the remedy in the process of use and a benefit for every participant (Grudin, 1994). Markus and Connolly showed interdependent tools carry an adoption problem no heuristic touches (Markus and Connolly, 1990). Grudin and Palen found near-universal use arriving without a mandate (Grudin and Palen, 1995). The designer has less leverage over adoption than the group's pressure on itself.

On the second side, that reading underestimates the artefact. Barley showed a technology reorganising work (Barley, 1986). Orlikowski showed the artefact structuring action (Orlikowski, 1992). Leonardi showed material agency acting through routine (Leonardi, 2011). Bikson and Eveland showed reinvention producing different outcomes from one tool (Bikson and Eveland, 1996). Dourish cuts against both camps if either becomes a checklist (Dourish, 2006).

What would settle the disagreement is a comparison the field has not run. Hold the artefact constant and vary the incentive structure and the mandate, then measure both use and the work outcome. Alternatively, hold incentives constant and vary the design. Grudin and Palen come nearest, but they observed adoption in two settings rather than assigning the regime, so they cannot separate mechanism from setting.

The strongest counter-argument is that the artefact is a party to the outcome, not a bystander. Barley's scanner reorganised the technical work of radiology, and Leonardi's imbrication gives material agency a role that design choices help determine (Barley, 1986; Leonardi, 2011). The defensible version of the thesis is therefore narrower: social organisation decides whether a benefit is realised, while the artefact decides which reorganisations are available to realise it.

Rate the evidence behind each reading. The organisational reading rests on a review of accumulated cases (Grudin, 1994), one conceptual distinction (Markus and Connolly, 1990) and one interview study of two organisations whose regime the field did not assign (Grudin and Palen, 1995). The artefact reading rests on three single-setting observational studies and one account of reinvention (Barley, 1986; Orlikowski, 1992; Leonardi, 2011; Bikson and Eveland, 1996). While the dispute is open, a designer should design the process of use and a direct benefit for every participant.

Where the argument is heading

The direction of travel runs past go-live. Wagner and Newell found that repairing a troubled enterprise resource planning project took continuous social work to reconcile the technology with local practice (Wagner and Newell, 2006). Read beside Leonardi's account of routines (Leonardi, 2011), the finding relocates adoption from a decision taken once to maintenance carried on indefinitely. Thirty years on, the founding challenges were retested rather than assumed. Duckert and Bjorn (2024) observed groupware use in two companies doing hybrid office work and found the disparity between who does the work and who benefits persisted, with groupware now infrastructure.

The sources this unit rests on

Each note states what the document did and found. Identifiers were resolved on 2 November 2026.

• Grudin, J. T. (1994). Groupware and social dynamics. doi:10.1145/175222.175230 Argues that repeated, expensive groupware failures were not technical but failures to meet eight challenges created by the gap between software for a person and software for a group, and recommends designing the process of use alongside it. The unit's foundation.
• Grudin, J. T. and Palen, L. A. (1995). Why groupware succeeds: discretion or mandate? doi:10.1007/978-94-011-0349-7_17 Interview studies of online meeting schedulers in two large organisations, asking whether adoption required a managerial mandate. Near-universal use was achieved without one, carried by versatile, easy-to-use software, organisation-wide infrastructure and peer pressure that built over time. It tests the unit's mandate mechanism directly and finds a mandate unnecessary.
• Markus, M. L. (1987). Toward a critical mass theory of interactive media. doi:10.1177/009365087014005003 Proposes a critical mass explanation for interactive media, built on universal access as a public good and reciprocal interdependence between earlier and later users, and predicts vulnerability to start-up failure and discontinuance. It supplies the unit's second mechanism.
• Markus, M. L. (1983). Power, politics, and MIS implementation. doi:10.1145/358141.358148 Sets out three theories of resistance, based on internal factors, poor design, and the interaction of design with organisational context, and argues through a case that resistance is often a rational response to a shift in power. The unit's reason for refusing to read non-adoption as user error.
• Markus, M. L. and Connolly, T. (1990). Why CSCW applications fail: problems in the adoption of interdependent work tools. doi:10.1145/99332.99368 Distinguishes groupware that benefits the individual user from groupware whose benefit is collective and conditional on others' use, showing the second faces adoption problems no single-user heuristic solves. It makes the incentive mechanism a property of the tool.
• Bullen, C. V. and Bennett, J. L. (1990). Learning from user experience with groupware. doi:10.1145/99332.99362 Analyses several installations and finds the technology rarely used as designed, with adoption depending on fit with existing routines and unanticipated use and non-use the norm. Evidence that use is reinvented rather than installed.
• Bikson, T. K. and Eveland, J. D. (1996). Groupware implementation: reinvention in the sociotechnical frame. doi:10.1145/240080.240355 Argues that groupware is reinvented as it is implemented and that a sociotechnical frame explains which reinventions hold, with outcomes depending on fit between tool and work group rather than on features.
• Orlikowski, W. J. (1992). The duality of technology. doi:10.1287/orsc.3.3.398 Develops a structurational model in which technology is both a product of human action and a medium that structures it, rejecting equally the view of technology as an external force and the view of it as pure strategic choice. The unit's resolution of the two readings.
• Orlikowski, W. J. (2000). Using technology and constituting structures. doi:10.1287/orsc.11.4.404.14600 Proposes a practice lens in which people enact structures through ongoing use, so the same artefact produces different structures in different settings.
• Barley, S. R. (1986). Technology as an occasion for structuring. doi:10.2307/2392767 Observes computed tomography scanners entering different radiology departments and finds the same technology produced different organisational structures, making it an occasion for structuring rather than a determinant. Evidence that the artefact is causally active without fixing the outcome.
• Leonardi, P. M. (2011). When flexible routines meet flexible technologies. doi:10.2307/23043493 Argues that human and material agencies are imbricated, so whether a technology affords or constrains depends on the routines it meets, and the same artefact can enable new practice in one setting and reproduce old practice in another. The unit's mechanism-level statement of fit.
• Dourish, P. (2006). Implications for design. doi:10.1145/1124772.1124855 Argues that implications for design may not be the best metric for evaluating ethnographic investigation and may fail, because the metric strips the account of its analytic value. It keeps the mismatch finding from collapsing into a feature list.
• Hirschheim, R. and Newman, M. (1988). Information systems and user resistance: theory and practice. doi:10.1093/comjnl/31.5.398 Survey the literature on resistance and work through a case study, concluding that resistance is a complex phenomenon which defies simple prescriptions. It supports the refusal to treat non-adoption as a psychological defect.
• Wagner, E. L. and Newell, S. (2006). Repairing ERP: producing social order to create a working information system. doi:10.1177/0021886305284290 Studies the repair of a troubled enterprise resource planning project, arguing through social ordering that a working system required continuous social work to reconcile the technology with local practice and that consensus need not be sought in every conflict. The organisational work continues after go-live.
• Duckert, M. and Bjorn, P. (2024). Revisiting Grudin's eight challenges for developers of groupware technologies 30 years later. doi:10.1515/icom-2023-0039 Retested Grudin's challenges through ethnography of hybrid office work in two companies and found the social and organisational challenges still present, with groupware now infrastructure.