Just Data Research Matters The Deep WIth

Conceptualizing better data work in non-profits: Prompt 6 – Infrastructural inversion

In their seminal book on classifications and infrastructures, Sorting things out: Classification and its consequences, Geoffrey Bowker and Susan Leigh Star conclude:

We have argued in this book that it is politically and ethically crucial to recognize the vital role of infrastructure in the “built moral environment.” Seemingly purely technical issues like how to name things and how to store data in fact constitute much of human interaction and much of what we come to know as natural. We have argued that a key for the future is to produce flexible classifications whose users are aware of their political and organizational dimensions and which explicitly retain traces of their construction.


The only good classification is a living classification.

Bowker, G. C., & Star, S. L. (1999). Sorting things out: Classification and its consequences. MIT Press, p. 326.

Let’s unpack. The book suggests that classification systems – animal taxonomies, for example, can come to play similar socioeconomic and political roles to other infrastructures, such as water, power, or Internet. That is, if these classification systems become ubiquitous enough, in use across multiple communities of practice (simply, a set of relationships between people doing things together – e.g. doctors, social workers, hair stylists, frisbee golfers), they can gain the properties of more familiar infrastructures, and can be similarly studied.

Infrastructures of any form, according to these two, share a common set of characteristics:

  • Embeddedness: Sunk into other structures, social arrangements, technologies.
  • Transparency: Does not need to be re-invented, invisibly supports tasks.
  • Reach or scope: Reach beyond a single event or site of practice.
  • Learned as part of membership: Taken-for-grantedness of artifacts and organizational arrangements is necessary for membership in a community of practice.
  • Links with conventions of practice: Is shaped and shapes the conventions of a community of practice.
  • Embodiment of standards: Plugs into other infrastructures and tools in standardized fashion.
  • Built on an installed user base: Wrestles with the inertia of an installed base.
  • Becomes invisible upon breakdown: [[Good infrastructures become naturalized and thus invisible]].
  • Is fixed in modular increments, not all at once or globally: Infrastructures are big, layered, and complex, and thus never changed from above.
Bowker, G. C., & Star, S. L. (1999). Sorting things out: Classification and its consequences. MIT Press, p. 238.

To return to our opening quote, data systems in non-profit organizations can become naturalized, and adopt the characteristics of infrastructures listed above. As this happens, the political and ethical quandaries encountered in the creation of the infrastructure are driven to the wayside and often forgotten through a process of clearance, or intentionally destroying history prior to the current instantiation of the infrastructure.

There is one very obvious infrastructure that structures social work: the Diagnostic and Statistical Manual, or DSM. The DSM fits the above criteria for an infrastructure, an idea which several studies support. Other related classifications might include things like crisis assessments, common intake forms, or in work with youth, the developmental assets and standardized test scores. While less hegemonic in their power than the DSM or test scores, these others travel across communities of practice and have power in defining organizational activities, outcomes, and data systems.

Convergence is the reciprocal and recursive process through which classification systems (always idealized abstractions) come to shape and be shaped in response to social worlds. The “built moral environment” referred to in our quote notes this: that the process of convergence twists and torques the lives of people to fit into classifications, often in violent and harmful ways.

Bowker and Star propose infrastructural inversion as the process of moving invisible infrastructures toward visibility, opening them up for study and possibly for change (though changing infrastructure is hard, it is possible). While there’s much to learn about how infrastructures work, that has been well detailed by others. Here, I want to focus on how one can reveal the work of infrastructures in shaping our lives and work (and especially, data work) and then what we might do about it (the purpose of this blog series).

There are four threads Bowker and Star suggest we follow if we want to invert the tendency of infrastructure to become invisible:

  • The ubiquity thread: No infrastructure can stand alone. To what other infrastructures/classifications does this connect? Rather than this single classification, here, are there ways we can understand this classification as part of a web of classifications and their surrounding infrastructures?
  • The materiality and textures thread: How do we perceive the world we have classified? What texture has been given to the world as a result of these classification systems?
  • The indeterminate past thread: what is history, relative to this infrastructure? What has it done to shape the ways we can understand history? Is it possible to do so without it?
  • The practical politics thread: How did we arrive at this classification? What discussions, debates, or power brought us here? What does it make visible/invisible?

In their text, they follow these threads for classifications that have become infrastructures, like the ICD, or International Classification of Disease, and the infrastructures of apartheid South Africa. In social work, it is particularly interesting to follow infrastructures like the Diagnostic and Statistical Manual (DSM).

The DSM is a particularly interesting and powerful example of an infrastructure because it is used by mental health practitioners, researchers, and insurance agencies – so nearly everyone talking about mental health does so in the context of the DSM. It therefore deeply shapes mental health practice. Even attempts to subvert or evade it, like using the mildest diagnostics for billing insurance, require consideration of, and response to, the DSM.

The DSM is a classification system – it helps us to diagnose a client’s symptoms as a particular form of mental illness, and through so doing, to categorize them appropriately. The DSM is an infrastructure too, fitting all the criteria listed above.

Infrastructures are very difficult to change, and tend to change very slowly and piecemeal. A tracing of the establishment of the DSM as an infrastructure reveals many issues. The DSM was created by the American Psychological Association and initially ignored. By the 3rd version, it was becoming embedded into practice and then into insurance billing. By the 4th edition, it was established as infrastructure, meaning many institutions and communities of practice relied on it for different functions. This meant that, moving into the 5th edition, very little could change, despite 25 million dollars and hundreds of volunteer experts spending thousands of hours trying to do so! Big ambitions became little change as many groups resisted changes that might impact diagnoses upon which affinity groups and progams were built, insurance companies didn’t want to pay for, and so on.

A couple of authors have traced the infrastructural dimension of the DSM (in my references) and offer a sound infrastructural inversion that suggests ways we might come to see the impacts of the DSM, the challenges it poses to offering good mental health supports, and the difficulties involved in trying to change it for the better. They are well worth reading to understand how one might approach the work of infrastructural inversion.

Our opening quote offers us a path forward, if only in the form of difficult starting questions. How do we create classification systems flexible enough to accommodate the complex lives of living beings? How do we make it possible to always see the political, ethical, and values-based decisions we are making? Can we leave traces of the creation of this classification so users can understand that choices were made in its creation, and see ways forward to doing other things?

This methodology or prompt, like the others in this series, doesn’t provide any easy answers or even a set of rules to follow. There are two approaches though that are possible in this case. One can use infrastructural inversion (and I highly recommend reading either an example from the book or something else from the below examples) to remove an existing infrastructure from its reified, naturalized, and ahistorical setting, and move it into our consciousness for thinking and consideration. One could do so in combination with other approaches in this series – what might a “participatory action infrastructural inversion” look like?

It is also possible to bring some of this awareness and techniques to the process of creating classifications that will become infrastructure. The standard classification turned infrastructure becomes a near-dead object in the sense that it ceases to make significant change, and instead becomes a force for the maintenance of the status quo. Users become beholden to it, and it twists and torques on the lives it intersects. If the only good classification is a living one, how do we keep classifications alive? Can we create them with sufficient flexibility to grow and change? Can we build in patterns of iteration that enforce significant growth and prepare users to accommodate it? Can we maintain a connection to, and intentionally work to build awareness of, the process of creating and evolving this system (the Reggio model does this with learning, for example)? How do we attend to the ways this classification system torques and twists the lives of those it impacts?


Bowker, G. C., & Star, S. L. (1999). Sorting things out: Classification and its consequences. MIT Press.

Cooper, R. (2015). Why is the Diagnostic and Statistical Manual of Mental Disorders so hard to revise? Path-dependence and “lock-in” in classification. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences, 51, 1–10.

Nelson, A. D. (2019). Diagnostic dissonance and negotiations of biomedicalisation: Mental health practitioners’ resistance to the DSM technology and diagnostic standardisation. Sociology of Health & Illness, 41(5), 933–949.

What I’m Reading

One reply on “Conceptualizing better data work in non-profits: Prompt 6 – Infrastructural inversion”