The routinization of lay expertise: A diachronic account of the invention and stabilization of an open-source artificial pancreas

Abstract

Embodied health movements (EHMs) advance their agendas by mediating the production, circulation, and revision of biomedical knowledge. To do this, their constituents become lay experts by blending their embodied experience of illness with self-taught technical knowledge. However, it is unclear how lay expertise is routinized within EHMs, and consequently, to what extent it can be made durable in long-term partnerships with credentialed experts. I follow the OpenAPS community—a group of people with type one diabetes who engineered an open-source ‘artificial pancreas’—from their inception in the transient #WeAreNotWaiting movement to their research collaborations with endocrinologists and detente with the FDA. I argue that OpenAPS user-contributors formalized their expertise in three steps: First, they broke the OpenAPS algorithm into modules so that prospective users must become experts to assemble it. Second, they lowered this barrier to entry by facilitating the socialization of new user-contributors with a training ritual. And third, they intervened in the strained endocrinologist-patient relationship. These tactics—restricting membership, reproducing expertise, and realigning interests—won the respect of credentialled experts who saw themselves in the OpenAPS community’s image. While not all EHMs follow this trajectory, this case demonstrates that lay expertise can mature and assume new institutional forms without relying on commercialization or patronage.

Keywords

lay expertise embodied health movements diabetes open source DIY biology biohacking

Managing type 1 diabetes (T1D) is an arduous task. To fill the gap left by an imperfectly functioning pancreas, patients must monitor their blood glucose levels and offset them with proportional doses of insulin. Continuous glucose monitors (CGMs) and insulin pumps, respectively, simplify these tasks, but rely on their users as intermediaries. For decades, biomedical engineers have sought to ‘close the loop’ between the CGM and the insulin pump (see Figure 1), automating insulin delivery (AID). Their efforts culminated in the FDA’s 2016 approval of Medtronic’s 670G, the first ‘artificial pancreas’. However, a group of patients had beaten Medtronic to the punch. They collaborated online under the moniker ‘#WeAreNotWaiting’ to develop an open-source artificial pancreas system that they abbreviated as ‘OpenAPS’ in 2015.

Figure 1.

Diagram of an artificial pancreas.

Since then, OpenAPS has attracted thousands of user-contributors committed not only to health and wellbeing, but to the ideals of medical autonomy and transparency. Users report that the software—which runs on a ‘rig’ that bridges compatible CGMs and insulin pumps—is user-friendly, customizable, and non-paternalistic. Advanced features like ‘AdvancedMealAssist’ and ‘AutoSensitivity’, which were not available on commercial devices at the time, free patients to enjoy their day-to-day lives with fewer diabetes-related tasks. Early clinical research has demonstrated that people with T1D using OpenAPS have lower HbA1C scores (a measure of average blood sugar in the preceding three months) than those using alternatives approved by the Food and Drug Administration (FDA) (Shahid & Lewis, 2022). The community’s ambition goes beyond OpenAPS; early adopters have consulted for medical device corporations, inspired iOS and Android spin-offs, and contributed their CGM data to collaborate with endocrinology researchers.

Using OpenAPS (often called ‘looping’) requires the reconfiguration of CGMs and insulin pumps. On the one hand, tinkering can be an empowering intervention that Jansky and Langstrup (2022) call ‘device activism’. On the other hand, ‘hacking’ medical devices voids their warranties and, worse, defies the FDA’s restriction on the use of older CGMs to monitor blood glucose in real-time. The agency warns that an improperly assembled artificial pancreas might deliver a fatal dose of insulin, though no incident has ever been reported. Advocates for OpenAPS counter that people with T1D are experts. They bring to OpenAPS their ability to administer insulin within a narrow margin of error—most T1D patients deftly complement a once-daily basal dose of long-acting insulin with several precise boluses of rapid-acting insulin at mealtimes. To use the OpenAPS algorithm, they learn to hack insulin pumps, interpret CGM data, and code in Python. In contrast, Fox (2017) alleges that products like Medtronic’s 670G make for ‘dumb patients’ reliant on ‘remote and anonymous clinical expertise with its own agenda’ (p. 146). Proponents of OpenAPS argue that the risk introduced ought to be compared to the substantial risks already posed by CGMs and insulin pumps alone.

This deference to patients’ expertise has opened OpenAPS to two critiques. First, some claim that the algorithm’s promise for managing diabetes is limited to those with sufficient intellectual capital. Farrington (2017) worries: ‘The technical acumen, motivation, and financial wherewithal necessary to use the system inevitably exclude those who do not possess these resources.’ With this barrier to entry in mind, Hatch et al. (2019) conclude that access to OpenAPS falls victim to the same ‘social inequalities that also define corporate pathways’.

Second, without oversight, patients’ expertise is neither standardized nor certified. Naïve patients curious about building a ‘do-it-yourself artificial pancreas’ (DIYAPS) may be vulnerable to quacks masquerading as experts online. What is to stop an outspoken member of the OpenAPS community from greenlighting a new version of the software before it is proven safe? Greene (2016) asks: ‘How do we decide who is competent enough to deal with the risks of technological auto-experimentation and who is not?’ Credentials, standards, and regulations are necessary, these critics argue, to reduce the risks inherent in innovation.

I argue that OpenAPS user-contributors have succeeded in putting these concerns to rest by routinizing their lay expertise in three steps: First, senior OpenAPS contributors divided their algorithm into discrete Python-coded modules hosted on GitHub. In this modular form, they reasoned, their algorithm would be protected as speech and not FDA-regulated as ‘Software as a Medical Device’. To use OpenAPS, newcomers had to perform their expertise by assembling the algorithm from these building blocks. The ‘expertise barrier’ posed by this DIY setup closed the OpenAPS community to dilettantes and protected overeager new users from themselves (Parthasarathy, 2010). It also bolstered the OpenAPS community’s credibility in the eyes of onlookers who were comforted by the competency its users were made to demonstrate. These membership criteria, in other words, amounted to ‘boundary work’ that garnered the OpenAPS community ‘the authority to call [themselves] scientists and to claim scientific legitimacy’ (Gieryn, 1983, p. 789).

Second, early OpenAPS contributors set out to lower this barrier to entry. They invited novices to undergo a training ritual that would teach them the skills necessary to assemble the algorithm and use it to automatically control their insulin pumps. This ritual—a three-phase tutorial—reproduced the expertise that was once limited to less than a dozen pioneering OpenAPS contributors. Users who completed this training and immersed themselves in OpenAPS forums became both experts in the technical aspects of AID and inured to the group’s open-source ethos. Those with the time and desire to do so might further advance to become members of the ‘core set’ who debated new features (Collins, 1981). The core set cordoned off controversies that might overwhelm new users or otherwise undermine the community’s credibility.

Finally, OpenAPS contributors recruited endocrinologists as allies. To do this, they amassed anonymized data from their CGMs into a cache called the OpenAPS Data Commons. These minute-by-minute blood glucose data, which once disciplined users to engage in ‘self-reflexive self-management’ and ‘self-surveillance’ became a source of collective epistemic power (Lupton, 2016). OpenAPS contributors gave endocrinologists access to the Data Commons and analysed it alongside them as research collaborators. At first, the knowledge products that sprung from these joint ventures served as records of OpenAPS’s efficacy that convinced clinical endocrinologists to begin advising patients using open-source AID systems. Eventually, the breadth of the Data Commons allowed these ‘collective experiments’ to incorporate T1D patients’ concerns about quality of life, independence, and comfort (Kempner & Bailey, 2019; Rabeharisoa & Callon, 2004). Patient-centered research was a break from existing work in endocrinology that, indebted to the Diabetes Control and Complications trial that began in 1982, saw patient compliance as a hurdle to be cleared rather than as a variable to be understood.

In each successive step, OpenAPS contributors restricted whom among them they recognized as experts, reproduced their expertise in neophytes through ritual training, and realigned patients’ and endocrinologists’ attention to matters of mutual concern. I trace the OpenAPS community’s arrival at each of these moments, beginning in 2013, when the first loopers hacked their CGMs and insulin pumps, and culminating in 2020, when the CREATE randomized controlled trial in New Zealand signaled their algorithm’s acceptance into mainstream endocrinology (Burnside et al., 2020). This diachronic account demonstrates that lay expertise emerges with an unsteady footing and must be routinized lest it fall. Initially, lay experts combine their moral authority and autodidactic knowledge to catch the attention of their credentialed counterparts. Yet, this fusion is unstable; new members may introduce conflicting expert claims, or outsiders may co-opt the moral agenda of the embodied health movement (EHM), leaving its lay experts in disarray. To persist, EHMs must carve out a field and maintain control of its boundaries, recruit new lay expert members, and ally with adjacent stakeholders. These ‘closure mechanisms’ (Collins, 1981) garner EHMs the respect of credentialed experts who demarcated their authority in similar ways. If speaking the language of experts earns their lay interlocutors a seat at the table, mimicking experts’ epistemic hegemony earns them longevity.

DIY biologists, biohackers, or health social movement?

Scholars have labeled OpenAPS contributors as ‘DIY biologists’ or ‘biohackers’. Instead, I consider the OpenAPS community to be a health social movement (HSM) or, more specifically, an ‘embodied health movement’. Brown et al. (2004) introduce the latter concept—which resembles Rabeharisoa et al.’s (2014) ‘evidence-based activism’—to describe HSMs that advance their agendas by mediating the production, validation, circulation, and application of facts, such as HIV/AIDS activists (Epstein, 1996), fibromyalgia patient groups (Barker, 2002), and the breast cancer movement (Barker & Galardi, 2011; Klawiter, 2008).

EHMs are distinct from ‘DIY biologists’ and ‘biohackers’ in that they build solidarity among social ‘networks based on shared beliefs’ (della Porta & Diani, 1998, cited in Brown et al., 2004). Brown et al. (2004) argue that EHMs use collective action tactics borrowed from mass political movements of the late 20^th century to intervene in a newly ‘scientized’ (Morello-Frosch et al., 2006) and ‘medicalized’ (Conrad, 2005) world. The OpenAPS community is best characterized as an EHM rather than as a do-it-yourself community; its constituents congregate in online platforms to demand innovation in diabetes medical devices and, when ignored, use their network to take matters into their own hands. Yet, definition of an EHM does not fit the OpenAPS community perfectly. Traditional social movements are defined by their opposition to institutions or corporations (Melucci, 1980; Touraine, 1985) while the OpenAPS community came to ally with medical device manufacturers and regulators. The DIY label, too, could be applied to the OpenAPS community despite the fact that its contributors deem it individualizing as it is embraced by other solidaristic groups. The ‘activists design collective’ that Sánchez Criado et al. (2016) study, for example, prototype wheelchair ramps as part of a broader ‘independent-living movement’. OpenAPS contributors and users might best be described as a collaboration-seeking group that sits at the intersections of several social movements including the #WeAreNotWaiting, free software, and broader T1D movements, each with their own agendas.

I describe the OpenAPS community as an EHM despite these divergences, because solidarity serves as a foundation for their epistemological claims. Through collective mobilization, EHMs like the OpenAPS community are able to furnish what Epstein (1995, 2023) calls ‘lay expertise’. Lay experts combine self-taught textbook knowledge with embodied knowledge to create a chimera that is greater than the sum of its parts. In Epstein’s (2023) words, lay expertise is hybrid in that it ‘traverses the boundary between official knowledge and its multiple “others”, rather than being located purely on the “alternative” side of the divide’. Lay experts use these dual competencies to pursue concrete goals and challenge the values that frame credentialed expertise. In these campaigns, Lay expertise coalesces as an alternative credential, admitting its purveyors into new venues. HIV/AIDS activists, for instance, solidified their advisory role in clinical research design in community advisory boards (CABs). These hybrid venues facilitate collaborations between concerned publics, academic researchers, and policymakers (Callon et al., 2011). The—often hot-and-cold—relationships between stakeholders in these forums introduce what Parthasarathy (2010, p. 358) calls ‘new policy making logics’ that ‘[challenge] how knowledge is framed and contextualized’.

The hybrid nature of lay expertise and the emergent forums in which it thrives raise two problems. First, public forums like CABs tend to reward autodidacticism. Newcomers must be self-taught in the language of experts; individual HIV/AIDS activists earned a seat at the table by assiduously self-learning virology and immunology (Epstein, 1996). This onus risks the ‘devolution of responsibility to individuals, who become expected to fend for themselves by seeking out and assessing information’ (Epstein, 2023, p. 27). Ideally, this alienation is offset by the ‘fundamentally collective character’ (Epstein, 2023, p. 13) of lay expertise and the propensity of EHMs to ‘transform a personal trouble into a social problem’ (Brown et al., 2004). Within a cohesive EHM, expertise flows seamlessly from veterans to new members. If distance grows between generations, however, conflicts over expert claims will threaten outward credibility. In other words, lay expertise without a means to reproduce itself is unstable. For the OpenAPS community, the stakes of solidarity are high: Contributors’ DIY ethos predisposes them to dissensus, so a deliberate system of social support is necessary to recruit and retain less-savvy users.

Lay expertise also invites concern over meritocratic standards. Without credentials, it is difficult to discern those who have earned the privilege of authority from imposters. Experts without institutional backing are scrutinized and may be dismissed outright, especially in high-stakes environments. Lay expertise, without affiliations or endorsements, is vulnerable to distrust from credentialed experts and peers alike. To address this, EHMs must vet new members and erect barriers to entry not unlike those set by formal actors like licensing boards and universities. As their community grew, Senior OpenAPS contributors sought to protect inexperienced users from experimental features. Amateurs, they considered, might encounter a critical glitch if they defaulted to using the unpolished code of intermediate or advanced contributors.

These are the same concerns that trouble onlookers such as Farrington (2017) and Hatch et al. (2019). The goals of the OpenAPS community are honorable, they might concede, but how do they protect patients? How do new OpenAPS users come to trust the software without the credential-backed legitimacy of the FDA? More generally, how is lay expertise accredited, reproduced, and integrated into broader epistemic networks? In other words, how is it routinized?

That lay expertise must be routinized to endure makes it not unlike Weber’s (1963[1922], 1978[1921]) ‘charisma’, the unstable form of authority found, for example, in many cults. It is worth extending this comparison further: Both charisma and lay expertise are emergent, both appeal to substantive (rather than instrumental) values, and both, resultantly, clash with prevailing rational-legal authorities. These similarities extend to EHMs’ and charismatic cults’ uncertain trajectories: As charismatic cults expand, they face internal pressures from factions vying for rank. These strains come to a head in crises of succession that transpire upon the charismatic leader’s death. To routinize charisma, meritocratic standards are implemented within the cult-cum-bureaucracy to stratify followers across a skill-based hierarchy and to select a new leader. EHMs, likewise, must rationalize their volatile lay expertise by resolving members’ conflicting knowledge claims and allying with formal outsiders. Although EHMs may not face crises of succession or depend as much on the special qualities of individuals, their efforts to reproduce members’ lay expertise stems from the same need to ensure continuity. The adjacent concepts of bureaucratization and formalization also apply to the OpenAPS community insofar as contributors have established formal internal structures, pioneered parallel startups, and earned posts on advisory boards (Staggenborg, 2022).

The best way to study the routinization of lay expertise is to shift the focus from the individual ‘hacker’ to the group at large and to trace that group over time. I analyse the OpenAPS community across three dimensions: First, I describe their internal protocols, routines, and rituals. Far from being ‘doing-it-yourself’, OpenAPS user-contributors discipline one another in a hierarchy not unlike the surgeons-in-training that Bosk (2003) studies. At the bottom of the hierarchy are new users yet to be socialized into the OpenAPS ethos. They must undergo a tutorial before earning the quasi-credential required to start looping. At the top of the OpenAPS pecking order are a ‘core set’ of experienced contributors (Collins, 1981). They absorb uncertainty to prevent the greater community from erupting in conflict and to ensure the safety of the algorithm. This system of social control and closure holds OpenAPS coders to a high standard of expertise and, in doing so, legitimates the system to outsiders.

Second, I complement my internalist account of the OpenAPS community with attention to its context. OpenAPS contributors intervene in a milieu of actors—regulators, firms, and endocrinologists—whose interests are misaligned. The conflicts among them are crystallized in CGMs and insulin pumps, which act as ‘boundary objects with agency’ (Fleischmann, 2006). Initially, these devices limit users’ actions to a script enacting the FDA’s paternalism (Woolgar, 1990). Yet, the OpenAPS community asserts their power by subverting their CGMs and seizing their data. OpenAPS contributors offer this data to endocrinologists who, subsequently, cement the former’s legitimacy and agenda. The OpenAPS community became durable, in part, by entangling themselves in this technoscientific network as ‘obligatory passage points’ (Callon, 1986).

Finally, I track the OpenAPS community as it evolves over time. The temporalities of EHMs have puzzled social scientists. Many EHMs crystalize as NGOs or as conduits for pharmaceutical advertising (O’Donovan, 2007). The OpenAPS community began in the urgency of the #WeAreNotWaiting movement, culminated in the invention of a glucose control algorithm, formalized in research collaborations with endocrinologists, and evolved into parallel startup companies. Lay expertise reveals itself to be dynamic at each stage, unfolding in new institutional forms that often bear little trace of the misnamed ‘hackers’ that instantiated it.

Research design

I began my research by gathering and closely reading texts from formal and informal bodies of literature. My archive of informal sources included online forums where OpenAPS users seek technical support, such as TuDiabetes.org, the Google Group ‘OpenAPS Dev’, and GitHub’s ‘issues’ tab for OpenAPS. In these forums—amounting to over 500 threads—novices sought help with assembling, troubleshooting, and customizing OpenAPS. Experienced users also posted on these forums. From their interactions, I gleaned a sense of the ‘closure mechanisms’ that the OpenAPS community deploys to resolve technical disputes (Collins, 1981). I also treated these texts as sources of background information, as ethnographic artifacts, and as narratives that structured the OpenAPS community’s sense of identity.

I complemented these online dialogues with an analysis of formal instruction manuals, articles, and books authored by contributors and their collaborators, critics, and boosters. The most accessible of these materials is Dana Lewis’s (2019b) monograph, Automated Insulin Delivery, which recounts the author’s contributions to the #WeAreNotWaiting movement for an audience of prospective loopers. I also studied dozens of peer-reviewed articles in journals that included the New England Journal of Medicine, Acta Diabetologica, and the Journal of Diabetes & Metabolic Disorders. These articles were authored by either OpenAPS contributors, endocrinology researchers, or collaboratively by both, and spanned topics from glycemic control to circadian rhythms.

In the second phase of my research, I conducted interviews with OpenAPS contributors, endocrinologists, and biomedical engineers. During interviews with OpenAPS contributors, I highlighted four important areas of questioning: alienation and autonomy, mutual support and care, standards of expertise, and barriers to entry. I listened to their worries about replacing the FDA’s slow, deliberate, regulatory process with the home-grown system I had read about online. I closed the interviews by inquiring about users’ trust in the FDA and in OpenAPS. With clinicians and researchers, I discussed the risks of OpenAPS, the difficulties of integrating the algorithm into clinical practice, and the usefulness of the Data Commons for answering questions of scientific import. When activists’ accounts diverged, I reconciled them by conducting follow-up interviews and requesting archival documentation. It is worth noting that while credentialled experts were skeptical about OpenAPS, OpenAPS contributors themselves were also critical in their support of the algorithm and its use.

Together, my analysis of these interviews and texts amounted to an ‘epistemography’, a description of how knowledge—or, in my case, expertise—is ‘made and certified’ (Dear, 2001, p. 130). In this spirit, I sought to understand how OpenAPS contributors earned expert status, resolved technical conflicts, and inured new recruits to the community’s internal status order. I then followed Fleischmann’s (2006) advice to treat the OpenAPS algorithm as a ‘boundary object with agency’ that was initially shaped by lay experts, endocrinologists, device manufacturers and regulators and subsequently came to exert agency over their interactions. By studying this case from the perspective of the algorithm itself, I was able to see how lay expertise and its routinization were also constrained by forces external to the OpenAPS community.

Inception

In 2013, Dana Lewis was frustrated with her continuous glucose monitor. Living with T1D, Lewis used her CGM—which clings to the skin with an adhesive—to manage her insulin dosing. If the device detected hypoglycemia, she drank a juice box. If it detected hyperglycemia, she injected a bolus of insulin using an insulin pump. At night, the CGM was supposed to wake Lewis if her blood glucose dropped too low—an imbalance that can be fatal. However, for Lewis, a heavy sleeper, the alarm was too quiet. Despite her and others’ complaints, the CGM’s manufacturer did not raise the alarm’s volume. Lewis—who has since been profiled in Wired, The Guardian, and Vice—felt limited in her role as a patient:

I was frustrated by being told to wait. I was living with the problem then, that day, that night, and every night for the rest of my life. And what could I do about it? Nothing. I was ‘just’ the patient and the ‘user’ or ‘consumer’ of the device, with no option to change medical devices to better suit my needs. (Lewis, 2019b, p. 4).

In protest, Lewis and her partner, a network engineer, attempted to make their own alarm. First, they adapted software from John Costik—who they met on Twitter—that could access data from Lewis’s CGM (Lewis, 2019a). With data in hand, they could display Lewis’s blood glucose status in real time on an external monitor, alert emergency contacts if she became hypoglycemic, and wake her with a loud, custom alarm.

Lewis’s innovation was part of the broader #WeAreNotWaiting movement that cropped up in multiple online forums. On Facebook, members of a group called ‘CGM in the Cloud’ similarly found a way to take control of their CGM data—by extracting it from their devices—and use it to meet their needs. Many used the workaround, which came to be called Nightscout, to monitor their children’s blood glucose levels remotely—it gave them peace of mind and some autonomy from the devices and healthcare teams that constrained their daily routines.

Only once she had appropriated her CGM data did Lewis recognize its full potential. She could use the real-time CGM data, together with rapid-acting insulin like Fiasp, to fine-tune her basal insulin rate. If her CGM reported hypoglycemia, she could taper her basal insulin rate; if it reported hyperglycemia, she could top it up. The equipoise of insulin and glucose remained the same, but real-time CGM data lent Lewis greater control. After several iterations, Lewis and Scott Leibrand combined real-time CGM data with user-inputted information about basal rate adjustments and carbohydrate intake to predict blood glucose levels into the future. With these forecasts, Lewis could proactively administer insulin boluses or adjust her basal rate to address fluctuations in her blood sugar. What started as a simple string of code that transferred CGM data to an external monitor became what endocrinologists call a ‘glucose control algorithm’.

Lewis used her glucose control algorithm to first create an ‘open loop’. Her CGM and algorithm generated insulin dosing recommendations, which she evaluated and manually transferred to her insulin pump. To ‘close the loop’ between the CGM and insulin pump, she needed only to remove herself as an intermediary. But, like CGMs, insulin pumps are shielded from unauthorized use. To subvert this restriction, Lewis collaborated with Ben West, who had recently reverse-engineered an insulin pump and could exploit a vulnerability in its firmware. When combined, Lewis’s control algorithm could instruct West’s ‘hacked’ insulin pump with little or no user intervention, effectively replicating the workings of a functional pancreas. Lewis called their invention the ‘open-source artificial pancreas system’ or OpenAPS (Lewis, 2019b).

Part of these DIY loopers’ success can be attributed to the duality of their lay expertise (Epstein, 1995). DIYers came to this project with an awareness of the pain points of commercial CGMs and insulin pumps, the acuity to administer the right dose of insulin at the right time, and the fear of sleeping the night relying on a hushed CGM. They combined this experiential knowledge (Borkman, 1976) with self-education in endocrinology, Python programming, and data science, competencies that resemble those of the lead-users described in innovation studies (von Hippel, 2017). Early loopers possessed both the experience of users and the technical skill of producers. This duality accelerated the feedback loop between inventor and consumer that is critical for innovation (Pinch & Bijker, 1984). It also occasionally resulted in conflicts between the jointly held priorities of technical rigor and user-friendliness: for instance, when the implementation of new insulin on board curves threatened the common-sense nature of the algorithm, as discussed below.

OpenAPS user-contributors also benefited from a lack of FDA oversight and, in some cases, a disregard for the restrictions the agency placed on their medical devices. Since at least the 1980s, the FDA has regulated how people with T1D can use CGMs and insulin pumps. In the early years, finger-stick blood glucose strips were the focus of skepticism; regulators cautioned that imperfect strips might mislead patients into making reckless decisions. In our interview, Lewis recalled:

If you go back to diabetes history, when fingerstick meters first came out, there was a big fight from doctors who said, ‘Patients cannot do this at home because they'll know what their blood sugar is and they'll do something, they might kill themselves.’ Well, fingerstick testing is the default now.

With more recent technology, the FDA is even more risk averse. As CGMs complemented fingerstick meters for many patients, the FDA forced CGM manufacturers to limit patients’ access to the device’s data. In order to receive FDA approval, CGMs like Lewis’s had to either save glucose data in a restricted format or delay its release. Additionally, the FDA restricts insulin pumps from receiving commands from unauthorized devices (when Ben West first controlled his insulin pump remotely, he exploited a software ‘bug’ that only afflicts a minority of devices). The FDA embargoes data to, in Woolgar’s (1990) words, ‘configure the user’. They limit the device’s use to a set of predetermined parameters to prevent reckless misuse.

FDA scrutiny of OpenAPS user-contributors was not focused on their misuse of insulin pumps and CGMs, but rather their glucose control algorithm itself, considered as of 2013 to be ‘software as a medical device’ (SaMD). Lewis first encountered an FDA representative at a 2014 industry conference. In that conversation, the official intimated that Lewis, if she shared her code, would be illegally distributing a class III medical device. Fearing sanctions, Lewis avoided sharing her early ‘open loop’ code.

As Lewis made progress on OpenAPS, however, she decided to share her code with a broader audience. Lewis and her peers proactively took steps to bolster their credibility in the eyes of the FDA. Initially, they used many of the same ‘credibility tactics’ popularized by HIV/AIDS activists (Epstein, 1995). For instance, they combined their autodidactic knowledge of endocrinology with a moral plea for autonomy. Just as HIV/AIDS activists pointed to the paradox of the FDA’s insistence on restricting investigational drugs from patients at the end of their lives, OpenAPS users lamented a similar contradiction between the self-responsibility necessary to manage T1D and the FDA’s paternalism. Lewis explained: ‘Diabetes is weird because when you’re diagnosed, [you’re told] “here’s a lethal drug, don’t kill yourself with it, but you’re in charge, you do everything yourself.”’ OpenAPS contributors also consulted with attorneys who advised them that their code is protected speech and, further, that their individual actions would not fall under the FDA’s regulatory purview. To cement this first amendment defense, early contributors divided their Python code into discrete modules on GitHub. In that form, they reasoned, OpenAPS would merely be a text—not regulated as a SaMD—until assembled by prospective users. Ben West recalls that the free software movement used a similar strategy to avoid intellectual property disputes:

[L]ore from the open source community … says that the software itself is an act of speech. It's ephemeral, and until you … configure all that ephemeral source code onto an actual device … it's still just ephemeral and only at that moment. We thought that the ‘free speech’ distribution would provide us with benefits, particularly as you say, with the FDA in falling outside of their regulatory reach.

OpenAPS contributors quickly realized that breaking OpenAPS into modules also made it safer. Users with the technical competency to download the algorithm from GitHub, assemble its fragments, and adapt it to their own hardware, Lewis and West posited, could also manage the risks inherent to a DIY system. OpenAPS contributors with whom I spoke resisted the characterization that they deliberately restricted their code: It was always available for anyone with internet access to download. The first OpenAPS users, however, faced an intentionally arduous task to compile the algorithm. Lewis used a Lego metaphor to describe the initial setup to me:

[W]e shared it in a way that it was like giving somebody a 1,000-piece Lego kit without the instructions, with a picture on the box that says like, ‘Hey, you can do this thing, but you have to do it yourself. Here's the 1,000 Lego pieces. Go figure it out.’ So, the first couple of people who found us via social media and word of mouth were people who are a little more technically oriented.

As the OpenAPS community’s ranks grew, contributors lowered this barrier to entry by drafting instructional documentation. Still, however, the authors of these manuals omitted key examples to vet new users. In the first draft of their documentation West, Lewis, and Leibrand explained that ‘there are a few things … that are not included in this guide, and intentionally so in order to ensure that you have full intent and autonomy in building your system for yourself’ (Lewis et al., 2016, p. 2, cited in Hatch et al., 2019, p. 425).

This group-level instantiation of what Parthasarathy (2010) calls an ‘expertise barrier’, enabled by OpenAPS’s modular form, earned the OpenAPS community legitimacy in the eyes of critics: Farrington (2017) conceded in the Lancet that ‘users who can implement the system are also likely to possess sufficient know-how to address any problems that arise.’ The same reasoning may have persuaded Courtney Lias, the director of the Division of Chemistry and Toxicology Devices at the FDA, to decline (several years earlier) to punish the parents of children with T1D for accessing their CGM data (Hurley, 2014). Additionally, roughly three quarters of endocrinologists with favorable views of DIYAPS limited their confidence to ‘certain trusted patients’ (Palmer et al., 2020, p. 864).

Yet the barriers that protect unskilled patients from tampering with their medical devices also prevent OpenAPS from achieving its lofty, democratic goals. OpenAPS may be customizable, user friendly, and empowering, yet—as my interviews with early contributors revealed—it can be vexing and impenetrable to those who have never coded in Python. Jansky and Langstrup are right to ‘stress that structural inequalities are weaved [into] device activism’ (Jansky & Langstrup, 2022).

To remedy this, OpenAPS contributors sought to implement mechanisms of social support for prospective users beyond instructional documentation. Lewis and her collaborators—a group which had by this time grown to about a dozen—took advantage of the modular form of OpenAPS itself to train novices. The chunks of OpenAPS, they reasoned, could be used as the ‘building blocks’ of expertise. In the embodied practice of constructing OpenAPS one module at a time, new users would be trained to read Python code and use GitHub.

To formalize recruits’ induction, OpenAPS contributors devised a three-phase tutorial: In ‘phase zero’, prospective users acquire compatible hardware and translate their CGM data into a legible, reliable form. In ‘phase one’ the quasi-students upload their CGM data to a rig and monitor it in real time. Finally, in ‘phase two’, users allow OpenAPS to automate the release of insulin. During the tutorial, recruits are also socialized into the community’s ethos of mutual care. OpenAPS documentation, a mostly technical instruction manual, prescribes the role that new users ought to play in the emergent community:

While formal training or experience as an engineer or a developer is not a prerequisite, a growth mindset is required to learn to work with the ‘building blocks’ that will help you develop your OpenAPS instance. Remember as you consider this project that this is not a ‘set and forget’ system; an OpenAPS implementation requires diligent and consistent testing and monitoring to ensure each piece of the system is monitoring, predicting, and controlling as desired. The performance and quality of your system lies solely with you (#OpenAPS Community, 2017).

This community of contributors believes in ‘paying it forward’, and individuals who are implementing these tools are asked to contribute by asking questions, helping improve documentation, and contributing in other ways (#OpenAPS Community, 2017).

In this passage, intermediate loopers are reminded that their ultimate responsibility for their own safety need not prevent them from coaching less-senior users. In fact, users who have recently acquired lay expert status are impelled to reproduce their expertise by writing instructional guides, building lively forums, collecting FAQs, and publishing blog posts. In the same way that self-help literature organized the ‘vast and dissimilar symptoms’ of fibrom;yalgia (Barker, 2002, p. 280), online forums transmuted OpenAPS users’ mutual desire for independence into collective care and solidarity. When their insulin pump cannulas clogged or CGMs reported missing data, users collectively troubleshot the problem. In her book, Lewis recognizes the collective effort that underwrites her expertise. Giving advice to future OpenAPS users, she writes:

Even armed with all of this knowledge, and years of experience, you may still find yourself in need of help from time to time. You need help troubleshooting your new device, you can’t find documentation or answers to your question and the support line for your system is closed, etc. Thankfully, the diabetes community is incredibly helpful, friendly, and supportive! There are dozens (if not hundreds) of channels where you can find help and support: forums such as TuDiabetes or Beyond Type 1, Twitter, Facebook Groups, Reddit, etc. (Lewis, 2019b, p. 68).

As the OpenAPS community grew, the sheer number of new users introduced two strains on these structures of social support: First, no two users had the same set of devices. Hundreds of CGM and insulin pump permutations made it hard for senior OpenAPS users to train neophytes. Second, experienced users began to conceive advanced features. Updates like AdvancedMealAssist and AutoSensitivty stirred conflicts that contributors needed to resolve without confusing newcomers. Together, these growing pains threatened the community’s cohesion and increased the risk of technical failure.

OpenAPS user-contributors solved these problems by bifurcating: A ‘core set’ of contributors, to borrow Collins’s (1981) phrase, broke off to tweak an unstable, beta version of OpenAPS. They communicated in long-form posts on GitHub and the ‘OpenAPS Dev’ Google Group. The remaining, larger group of users relied on a stable ‘master branch’ of the code. They received real-time support from senior users on messaging platforms like Slack and Gitter. This two-tiered structure followed the common open-source coding convention of protecting the public-facing product from ongoing development behind-the-scenes.

In one instance, the core set debated the best way to estimate insulin on board (IOB)— the volume of insulin lingering in one’s body. IOB is a function of insulin pharmacokinetics; IOB will peak 30-90 minutes after injecting rapid-acting insulin and, one hour later, will be negligible. Glucose control algorithms like OpenAPS rely on a running estimate of IOB to calculate insulin boluses. If a user’s IOB is high, fewer additional units of insulin are necessary to offset blood sugar.

Two camps emerged: The original contributors, including Leibrand, started with the axiom that the activity of rapid-acting insulins like Humalog, Novolog, and Fiasp increases linearly until it reaches a peak, after which it decreases on the same linear path. This model, which became known as the bilinear decay function, was then mathematically manipulated—integrated and inverted—to produce an ‘IOB curve’ (orange lines, Figure 2). The second camp of contributors thought this approach was too deductive. Instead, they collected data that came from either insulin pumps, which estimate IOB, or from published ‘clamp studies’, which are endocrinologists’ gold-standard for evaluating insulin pharmacokinetics. This inductive approach was robust, but the complex polynomial IOB curve that it produced was intractable.

Figure 2.

OpenAPS insulin “activity” and IOB curves.

To quell this dispute, the first camp introduced what Collins (1981) calls a ‘closure mechanism’. In Collins’s case, physicists with the most compelling ‘rhetorical and presentational devices’ won the dispute over gravitational wave experiments. The debate over IOB modeling similarly hinged on a closure mechanism exogenous to the merits of the proposals themselves. Namely, the linear decay function prevailed because of its elegance and ease-of-use. Chastising the convoluted math of an inductivist, one user wrote:

Again, I’m not trying to bash your choice, but I do have to say I'm not comfortable using an equation I don’t understand that seems to basically have the same shape as another to calculate an insulin dose. I’d much rather use an equation I fully comprehend and that has a basis in chemical kinetics.

For the OpenAPS community, as for physicists, the core set concealed expert conflict to bolster legitimacy. It worked by separating the expert field into a frontstage and backstage (Hilgartner, 2000). On the frontstage, OpenAPS appeared to be a stable product. The well-tested ‘master branch’ bared no trace of the conflicts brewing backstage over yet-to-be-released features. Users in the audience found the algorithm to be ‘more uniformly compelling’ than they would have if they were privy to the skirmishes behind the curtain (Collins, 1981). Backstage, the core set debated with one another to reconcile their dual commitments to user-friendliness and technical rigor encompassed by the community’s hybrid lay expertise. In November 2017, this dialectical tension was resolved when the inductivists and deductivists compromised on a curvilinear insulin decay function with a single exponential term (green lines, Figure 2).

To pull back the curtain is not to say that the backstage work of core sets is deceptive. Core sets are pragmatic. They bracket conflicts so that experts can weigh in on time-sensitive issues. For the OpenAPS community, the two-tier workflow protects less-experienced users from underdeveloped features. The core set also accelerates innovation—features like MealAssist, advanced MealAssist, and AutoSensitivity were not available on commercial artificial pancreases at the time, and would not have been feasible without a core set.

Crucially, though, the core set was permeable enough to include contributors with heterogenous devices. They made OpenAPS compatible with Medtronic, Dexcom, and Abbott CGMs. For the first time, patients could easily combine insulin pumps and CGMs from different manufacturers on a ‘bring-your-own-device’ basis. Cross-compatibility became a strength of the system. In our interview, Ben West recalls that,

people wanted to bring their own device. [For example,] I want to bring my mobile device from last year, or from five years ago or … the one that they just sold me yesterday. I want to bring that device into my system that I use for my medical therapy. That's not something that the medical industry does very well, but it's something that Nightscout and the open-source projects do handle very well because that's precisely the need that a lot of people want solved. [I]t's not just the tech users or the super users that want to bring their own device, that's actually a common need.

OpenAPS’s interoperability is hailed as the technical fruit of its contributors’ unconventional model of innovation. Yet, I argue, the community’s greater triumph lies in the respect it earned from the FDA and its new users. To assuage both parties, OpenAPS contributors formalized their lay expertise in two ways: First, by creating a mechanism to reproduce the expertise of seasoned contributors among nonexpert newcomers. And second, by establishing a de facto meritocracy within their community that limited the risks of innovation to only the most capable lay engineers. Together, these new standards of expertise gave shape to the OpenAPS community. What began as a constellation of individual patients engaging in ‘do-it-yourself’, activities became a durable community that dispatched social control, social support, and social closure.

Expanding the reach of the #WeAreNotWaiting movement

In their challenges to the FDA and in the credibility tactics that they used to appease credentialled experts, DIY loopers mimic antecedent HSMs. HIV/AIDS activists were also Janus-faced: They ‘learn[ed] the foreign language of biomedicine’ while simultaneously subverting FDA clinical trial rules with guerilla tactics, like importing unapproved therapies and splitting experimental treatment doses (Epstein, 1996).

HIV/AIDS activists, however, went on to formalize their relationship with the FDA in durable forums like CABs. Today, CABs provide input on US clinical trial protocols, recruitment, and data analysis. HIV/AIDS activists also pressed to reform FDA rules, like those pertaining to investigational drug exemptions (IDEs), that carried the fruits of their organizing into the future and for new disease constituencies.

In contrast, OpenAPS contributors distanced themselves from credentialled experts in the early years. Instead of partnering with the FDA, endocrinologists, and medical device manufacturers, they defended their autonomy from regulators, refused to perform the ‘sick role’ (Parsons, 1975), and reconfigured medical devices. If the OpenAPS community continued on this path, onlookers feared, they would be unable to impact policy, as HIV/AIDS activists had, or to distribute their technology to a wider audience. OpenAPS contributors faced a critical juncture: continue to adopt an adversarial posture and face ‘diminished possibilities of holding traditional experts or governments to account’ (Epstein, 2023) or partner with credentialed experts.

Enrolling endocrinologists

Lewis, Leibrand, West, and other members of the #WeAreNotWaiting movement chose the second path. They started by allying with endocrinologists. Unlike the FDA, clinicians are points of first contact for T1D patients. Their purpose is, in part, to ‘indoctrin[ate] … patients into the ways and means of personal disease management’ (Mauck, 2010, p. 8) by surveilling their glycemic control. Over time, CGMs and insulin pumps have simplified this task by extending surveillance into patients’ homes and simplifying their compliance with insulin regimens (Fox, 2017). OpenAPS inverted this relationship, giving users greater control over their medical devices in exchange for an upfront investment in building technical expertise. In doing so, OpenAPS also unsettled endocrinologists’ clinical practice. To counsel loopers, endocrinologists had to learn about OpenAPS themselves, if only enough to implicitly endorse it. To enroll endocrinologists as allies, the OpenAPS community intervened in clinical research. They conducted clinical trials using their own data—collected in a cache they called the OpenAPS Data Commons—that earned the trust of endocrinologists and, eventually, became instrumental for future studies of glycemic control.

The regime of T1D care into which OpenAPS intervened began in the 1980s, with the invention of the insulin pump. The landmark Diabetes Control and Complications Trial found that the ‘intensive therapy’ enabled by this device reduced the long-term risk of retinopathy, nephropathy, and neuropathy. Yet, insulin pumps were not in universal use. For subjects without an insulin pump, the DCCT defined intensive therapy as ‘three or more daily insulin injections’ (DCCT Research Group, 1993, p. 977). To meet these standards, endocrinologists counseled patients to self-monitor blood glucose (SMBG) using paper logs. Patients were taught to extract blood with a lancet, blot it with a reagent strip, and analyse it with a reflectance photometer (also called a glucose meter). They recorded these blood glucose readings, along with their commensurate insulin doses, at mealtime. This diaristic routine was optimistically dubbed ‘patient-centered self-management’.

From its inception, however, intensive therapy was impractical. As DCCT investigators noted, their protocol required ‘an expert team of diabetologists, nurses, dietitians, and behavioral specialists’ to monitor patients and facilitate their compliance with the regimen (DCCT Research Group, 1993). At home, patients found ‘the approach impractical and inconsistent with their other needs’ (Mauck, 2010). The DCCT study team concluded that new strategies or devices would be needed ‘to adapt methods of intensive treatment for use in the general community at less cost and effort’ (1993, p. 984).

Their wishes were fulfilled, in part, when the FDA approved the first ‘professional’ CGM in 1999. This device replaced the daily telephone calls from nurse diabetes educators that DCCT found efficacious but laborious. Instead, the professional CGM surveilled patient adherence in the background:

[The sensor] is placed on the patient’s arm in the clinician’s office. Patients leave it on for 14 days, going about their normal daily routine (no finger stick calibrations are required), after which time, they return to the office and the physician removes the device and downloads the data for assessment. (Thompson, 2016)

At the advent of the #WeAreNotWaiting movement, CGMs and insulin pumps had advanced but were still bound to the DCCT model. New CGM sensors were more accurate, but wary of asking too much of patients. They ‘configured’ their users (Woolgar, 1990) by, for instance, limiting Lewis from raising the volume of her hypoglycemia alarm, or parents from remotely monitoring their children’s glycemia. Through these restrictions, Fox argues,

[r]esponsibility for self-management is removed from users, replacing an ‘expert patient’ (Shaw & Baker, 2004) with sophisticated understanding of his or her disease and its management with a ‘dumb patient’ who merely has to wear the device and follow any instructions it provides to the user (e.g. to inject an additional insulin bolus if blood glucose goes too high). A collaborative relationship between patient and professional is now replaced with a much more traditional relationship in which the patient is passive and the active relations in the assemblage are the device, its manufacturer and the medical specialist (Fox, 2017, p. 142).

The #WeAreNotWaiting movement reasserted its members’ expertise by contravening FDA rules dating to the DCCT era to use data from Dexcom G4 sensors for AID. OpenAPS users’ initial investment in building OpenAPS was repaid in fewer bolus calculations and meal-time insulin top-ups.

Endocrinologists were divided on OpenAPS’s safety and wary of advising patients with a DIYAPS. In an interview study of endocrinologists, Palmer et al. (2020, p. 863) found that, when asked if DIYAPS was safe, providers responded ‘I don’t know’ (39%), that DIYAPS was ‘moderately safe’ (30%) or that it was ‘safe’ (22%). Critical to expanding the user-base of OpenAPS, then, was convincing endocrinologists of the software’s safety and to train them in advising patients who opt to use OpenAPS.

To do this, OpenAPS user-contributors ushered in a new paradigm of research that broke with the DCCT. They began by documenting in online forums their experiences using OpenAPS. Contributors testified to the algorithm’s stability, ease-of-use, and safety. But user success or failure was a function of their savvy as much as the algorithm’s sophistication. Endocrinologists picked up on this, cautioning that OpenAPS is ‘used by a highly selective, motivated and technology-adept cohort’ (Melmer et al., 2019).

Their findings were also epistemologically flawed. At best, OpenAPS contributors, like the ‘Clusterbusters’ that Kempner and Bailey (2019) studied, produced knowledge collectively by ‘iteratively and recursively refashioning their self-experiments in collaborations with and in response to feedback from similarly positioned experiments’ (Kempner and Bailey, 2019, p. 3). At worst, they wrote testimonials that were idiosyncratic and nonreplicable. Their anecdotes may have been adequate to garner the trust of adventurous patients, but fell short of convincing endocrinologists.

By June 2016, the core set sought to surpass ‘collective self-experimentation’ by organizing a rigorous community survey. Eighteen users reported their HbA1c scores, sleep quality, and incidents of hyperglycemia or hypoglycemia on a standardized survey. These data became the basis for a poster that Lewis and Leibrand presented at the annual meeting of the American Diabetes Association and a paper that they co-authored in the Journal of Diabetes Science and Technology (Lewis et al., 2016). Their analysis found that participants spent a vast majority of their time using OpenAPS in a state of euglycemia—the target range for blood glucose—and benefited from improved ‘peace of mind’. The study inspired Lewis and her co-authors to publish further:

I learned how to write an academic paper, it wasn’t that hard. They let me do it, I'm going to keep doing it, just because nobody was studying this, so I had to study it myself and provoke other people to study it and write about it, which has now happened.

Lewis prepared for future studies by recruiting other OpenAPS users to a third-party platform called ‘Open Humans’ where she hosted the OpenAPS Data Commons (Greshake Tzovaras et al., 2019). The platform allows users to anonymously contribute measurements from their CGM coupled with their demographic characteristics. As of 2022, the Commons comprises ‘the largest freely available diabetes-related dataset with over 46,070 days’ worth of data and over 10 million CGM data points’ (Shahid & Lewis, 2022).

Internally, the Data Commons served as another mechanism through which OpenAPS user-contributors built solidarity. CGM data once served as a disciplinary tool—a high-stakes version of the self-tracking that Lupton (2016) laments. Like pedometers or calorie trackers, the CGM directed users’ attention inward, to inculcate hygiene and self-responsibility. When collectivized, however, CGM data became a powerful source of cohesion and social support. User-contributors came to see hyperglycemia and hypoglycemia as collective problems rather than individual ones. They took pride in their shared commitment to looping and its aggregate benefits.

Externally, the Data Commons lent the OpenAPS community epistemic strength in numbers. With far more than 18 participants, Lewis’s new research notes carried more cachet: One study of OpenAPS using Commons data found that ‘the glycemic benefit of DIYAPS is in reducing hyperglycemia without compromising the low occurrence of hypoglycemia’ (Zabinsky et al., 2020). As academic researchers encountered these findings, they sought access to the Data Commons for their own research: In Nutrients, Shahid and Lewis (2022, p. 20) write, ‘the OpenAPS Data Commons dataset has great potential for further applied research and development’.

With the Data Commons, the OpenAPS community flipped the script. In the same way that the FDA ‘configured the users’ of CGMs and insulin pumps (Woolgar, 1990), limiting them from exporting or analysing CGM data, the OpenAPS community used the now-emancipated data to reorient the priorities of research (Rabeharisoa & Callon, 2004). They established two rules surrounding data use: First, they permitted researchers to use the data only after their proposals are vetted by a panel of OpenAPS volunteers. Second, they required applicants to publish their research in an open-source format and to inform the OpenAPS community of their preliminary results in six months’ time. With the Data Commons, loopers, like HIV/AIDS activists, inserted themselves as ‘obligatory passage points’, through whom credentialled experts must pass to benefit from the wealth of their Data Commons (Callon, 1986, cited in Epstein, 1996, p. 253). The platform allowed OpenAPS contributors to concretize their experiential knowledge in the form of ‘observations, accumulated day after day … through practices that resemble scientific practices’, which Akrich and Rabeharisoa (2023, p. 106) find is also advantageous for patient associations.

Members of the OpenAPS community often took an active role in analysing data from the Commons. Collaborative papers, like one co-authored by Lewis and published in Nutrients, introduced concerns about quality of life to research on AID (Shahid & Lewis, 2022). These questions were quickly picked up in larger clinical trials. Even randomized-controlled trials (RCTs), notorious for their narrow focus on efficacy, began to include measures of users’ diligence, their access to healthcare, and their reliance on family support. In New Zealand, a team of endocrinologists and patients initiated the first RCT of a DIYAPS dubbed CREATE (Community deRivEd AutomaTEd insulin delivery). RCTs are expensive and lengthy, but without them, the FDA and the broader scientific community are skeptical of novel medical devices. Endocrinologists with whom I spoke turned a corner on OpenAPS upon hearing of CREATE: The algorithm, they reconsidered, might place patients in the center of their own care.

Although regulators and endocrinologists came to tolerate and even collaborate with the OpenAPS community, the pathway to this relationship was not without obstacles. Endocrinologists were hesitant to treat, and regulators to implicitly endorse, patients who might implicate them in the unauthorized use of CGMs. Yet, after publishing journal articles, compiling a strong safety record, and most importantly, cultivating the technical expertise of new users, the OpenAPS community was successful in aligning the interests of credentialed experts with their own.

Encounters with industry

In a prospectus describing the goals of the New Zealand trial, the investigators lament the prohibitive cost of commercial artificial pancreases like the 670G. They state their intention to study ‘an algorithm developed through open-source innovation [that] would facilitate the development of an AID system at a fraction of the cost’ (Burnside et al., 2020, p. 1616). Many OpenAPS contributors likewise hoped that their legacy would culminate in commercialization. They held out hope that a medical device firm would build on the innovations of OpenAPS. This faction of the #WeAreNotWaiting movement saw knowledge projects like CREATE as a step toward broader investment in a commercial artificial pancreas.

To accelerate this goal, OpenAPS contributors adopted three strategies. The earliest group of contributors, true to their open-source ethos, shared their work freely with firms. Lewis, despite her initial hostility to companies like Medtronic and Abbott for their restrictive designs, recognized their importance for bringing a closed-loop system to the broader market. In an interview with The Medical Futurist, Lewis reiterated that the movement’s goal was to push manufacturers to act more quickly. The decision to share their algorithm using the MIT license would give firms a head start. Lewis said:

We decided we will best achieve this by supporting the community through an open-source approach, but not by distributing a medical device. To us, this seems to be making a bigger difference to the community than if we had launched yet another commercial APS effort. (The Medical Futurist, 2016)

Lewis recognized that while her DIYAPS gained traction among a technically ambitious minority, it would have its best chance at global impact if it was copied by a medical device manufacturer. In an interview with me, West echoed this sentiment. He sees commercialization as a necessary next step to get OpenAPS into more hands:

Right now, one of the things that’s holding up more people from benefiting from things like Nightscout is the fact that it's DIY—you can’t just go in the App Store and press a button and have it all done right.

A second set of the early OpenAPS contributors chose to initiate parallel startup companies. In Europe, several OpenAPS contributors took advantage of the ubiquity of Bluetooth-enabled insulin in their home countries. (The European Medicines Agency—and the regulatory bodies of its member states—allow insulin pumps to receive commands over Bluetooth.) Their OpenAPS spinoff, AndroidAPS, lets users build their own closed-loop system using an off-the-shelf insulin pump. Users are also reassured by AndroidAPS’s recent decision to become a nonprofit: A full-time development team—not a gaggle of volunteers—adopts features from OpenAPS after they have been tested. The benefit of AndroidAPS is its user-friendliness: Users can build the app with Android Studio on their smartphones and synchronize it with their insulin pumps and CGMs. It is worth noting here that the variation in regulatory regimes between the US, where most OpenAPS contributors operate, and Europe, the home of AndroidAPS, was sufficient to split the #WeAreNotWaiting movement into two communities that administer the same software under different organizational schemes. This cleavage suggests the importance of a ‘fit’ between the trajectory of the routinization of lay expertise and its sociolegal context.

In a third type of encounter with industry, medical device firms have called on OpenAPS contributors for advice. In one case, a French company asked Lewis to improve how users experience their platform. Lewis took the opportunity to push for patients’ ability to access their own data in real time:

[O]ne of the things that we've talked a lot about with the French company and with the other companies has been, ‘How do you help people decide to trust? How do you give a user information?’ And I think, encouraged by the regulatory environment, companies have traditionally [said], ‘Well, we’re not going to show people information because they might do something with it, and it’s risky because they could do the wrong thing.’ This company is unique because they told us what they were doing, they invited us in for conversations and asked questions and deeply wanted to learn.

As a consultant, Lewis taught the firm how to gain the trust of patients. She persuaded them that OpenAPS users’ ability to exercise control over their CGM data contributed to their higher-than-average health outcomes. The French company was receptive to Lewis’s suggestion and open to designing a transparent user interface that respects patients’ expertise.

In pursuing these three strategies, members of the OpenAPS community use their lay expertise to advise the same firms that they once criticized. Here, lay expertise refers in part to the concurrent application of experiential knowledge to improve the usability of APSs, technical knowledge to ensure the safety of algorithms, and regulatory and economic knowledge to bring a product to market (Akrich & Rabeharisoa, 2023; Epstein, 1996, 2023; Rabeharisoa & Doganova, 2021). This full circle moment renders obsolete the ‘intellectual opportunity structure’ that once energized OpenAPS contributors—in this case, the lack of interoperable, customizable devices for managing T1D (Waidzunas, 2013). Yet, unlike the Irish health advocacy organizations that O’Donovan (2007) studies, that were hijacked by pharmaceutical industry funding, the OpenAPS community remains a self-sufficient community with a glucose control algorithm of their own. In their willingness to advise firms who they might otherwise see as competitors, OpenAPS user-contributors further inscribe their niche as purveyors of a cutting-edge algorithm for patients who choose it over one-size-fits-all commercial systems. Outside of this niche, OpenAPS contributors’ consultations for industry make their expertise durable not in freestanding forums like CABs, but in new relationships between medical device manufacturers, regulators, and patients. This realignment leaves their customizable DIYAPS intact and creates inroads for future patient feedback on commercial medical device innovation. In other words, OpenAPS contributors are able to institutionalize their role in outside organizations in a way that resists the co-optation of their interests and the dissolution of their community. This balance of recognition and persistence is critical for the routinization of expertise.

Conclusion

Prior accounts of ‘DIY biologists’—both in general magazines (Talbot, 2020) and in academic texts (Hatch et al., 2019)—have been either laudatory or derisive. Articles written in the former tenor recognize that cheap, individualized, and at-home solutions can chip away at healthcare policy failures. Writing for the New Yorker, Talbot (2020) comments that ‘given the profound flaws of the American health-care system, there is something hopeful about [DIY biology].’ Conversely, critics fret over the stakes of error and the alienating effects of ‘doing-it-yourself’. In the Oxford Handbook of the Sociology of Body and Embodiment, Hatch et al. (2019) warn that ‘[t]he notions of individuality and self-reliance that define maker culture and the DIYAPS movement are classic neoliberal values’. Articles of both valences portray members of DIY communities as self-sufficient mavericks who operate outside the strictures of regulation or credentials. OpenAPS contributors distance themselves from the ‘DIY’ label. They emphasize, in contrast, the collective nature of their pursuit. While much of their work is done at home, on a budget, and by hand, the stakes of reconfiguring medical devices require social support.

The #WeAreNotWaiting movement is also distinct from DIY biologists and biohackers in how participants bolster their expertise. The latter, as Ikemoto (2017) details, share a common training (many have PhDs) and a set of institutional connections with their credentialed counterparts in synthetic biology. These experts borrow credibility from universities, degrees, and tools like CRISPR-Cas9. The OpenAPS community, by contrast, built their credibility from the ground up. They begin by drawing boundaries around their community that demarcate their expert status. New users who wish to cross that expert threshold are socialized in a ritual not unlike a formal training or degree program. OpenAPS contributors’ similarity to credentialed experts even extends to the closure mechanisms that they use to resolve disputes (Collins, 1981). It is perhaps this isomorphism with their credentialed counterparts that led OpenAPS contributors to publish in academic journals and to consult for industry.

Working backwards, these findings imply that EHMs unable to routinize their lay expertise may fracture, dissipate, or pivot. If EHMs fail to reproduce their expertise to new generations of recruits, they risk succumbing to tensions between the self-taught and amateur members of their cadre. AIDS activists, as Epstein (1996) argues, experienced this strain as the ‘first wave of autodidacts’ neglected to develop ‘a coherent educational strategy that would bring larger numbers of activists into the arena of knowledge-assessment’ (293). Partly as a result, the movement bifurcated; ACT UP’s erudite Treatment and Data Committee cooperated with government officials from whom they sought incremental reform, while their ‘lay lay’ counterparts sustained their theatrical demonstrations and demands for radical change (Elbaz, 1992, p. 488). Early OpenAPS contributors like Lewis faced similar ‘pressures to democratize science’ from throngs of aspiring users that conflicted with ‘pressures to establish new hierarchies of expertise’ from the FDA and critics (Epstein, 1996, p. 294). Surpassing AIDS activists in this regard, Lewis balanced both demands by sharing OpenAPS widely yet guiding its users with the three-phase tutorial.

Conversely, a failure to restrict expert status—to establish these ‘hierarchies of expertise’—may leave an EHM similarly disjointed. The breast cancer movement, in contrast to AIDS activism, was adept at reproducing its lay expertise through programs like Project LEAD, ‘an intensive science training for activists’ under the auspices of the National Breast Cancer Coalition (Myhre, 2001, p. 98). These forums, Myhre argues, served as ‘avenues for the flow of lay expertise from medical mavens to the rest of the movement’ (p. 124). Yet, in the San Francisco Bay Area alone, Klawiter (2008) observes that the movement comprised three discrete cultures of action (COAs). These COAs competed for expert status, resources, and recognition. The ‘culture of patient empowerment’, for example, clashed with the ‘culture of early detection’ in its framing of breast cancer as a ‘body-altering, life-threatening source of suffering’ rather than as a ‘disease … that could be controlled through scientific research’ (211). The OpenAPS community avoided the problem of dissensus in part by dividing its membership into ‘development’ and ‘master’ branches. Yet, as time passes, many OpenAPS contributors have chosen to spearhead parallel startups. Discord and divergence are inevitable within any EHM and its closure (or aggravation) determines the durability (or transience) of lay expertise.

Finally, while EHMs must reproduce and restrict their lay expertise to maintain internal cohesion, these routinization mechanisms are insufficient to preempt external competitors or usurpers. For that, EHMs must realign the interests of external stakeholders with their own. The OpenAPS community and AIDS activists alike earned clout in their roles as consultants and CAB members, respectively. These close partnerships with the FDA and with industry led both movements to converge with their credentialed interlocutors. As ACT UP earned a position on NIAID’s Statistical Working Group, they came to tolerate placebos as ineluctable epidemiological tools. Likewise, OpenAPS contributors’ close relationships with endocrinologists led them to continue publishing in peer-reviewed journals. Conversely, when Nixon’s NIH usurped the Black Panther Party’s program to screen for sickle cell anemia, the Party pivoted. Whereas the Panthers once took pride in their ‘trustworthy “authentic expertise”’, Nelson (2011, p. 129) finds that they instead came to ‘emphasize [their] collaboration with medical professionals from “accredited hospitals”’ and noted that those who donated their time at People’s Free Medical Clinics had ‘received training in Sickle Cell Anemia testing from accredited hospitals’ (151). In this case, the Party’s ‘receding moral authority’ (151) drove the lay experts to replace their alternative credentials with literal ones.

These three tactics are visible at different moments when EHMs are placed in a diachronic horizon. At the outset, the OpenAPS community benefited from an ‘intellectual opportunity structure’ (Waidzunas, 2013) or ‘credibility gap’ (Epstein, 1995, p. 411), which, in this case, was a lag between the commercial introduction of accurate CGMs and the realization of their potential to empower users to independently manage their T1D. To address this gap, the OpenAPS community operated on an unsteady legal footing and relied, in part, on their iron-clad closure mechanisms. When medical device companies unveiled their own artificial pancreases and academic researchers took interest in AID, the OpenAPS community persisted by fashioning themselves as ‘obligatory passage points’ with the introduction of the Data Commons (Callon, 1986).

As conflicts over expertise continue to motivate social movements, it is important to recognize that lay expertise evolves (and devolves) over time and that its boundaries are constructed (and deconstructed) in response to internal and external pressures. Internally, lay expert movements must reconcile the tension between the autodidacticism of their first members and their need for a growing coalition that speaks with a single voice. Externally, lay experts must enroll allies and collaborators who shore up their claims to credibility without co-opting their agendas. In navigating these tensions, lay experts often (and perhaps surprisingly) converge with their credentialed counterparts; both rely on restricting their domain from outsiders, reproducing their expertise and its ethos among newcomers, and re-aligning the interests of adjacent fields to accommodate the novelty of their intervention.

Footnotes

Acknowledgements

The author kindly thanks Steven Epstein, Wendy Espeland, Carol Heimer, and Charles Camic for their mentorship and thoughtful feedback on this manuscript.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Clay Davis

Author biography

Clay Davis is a PhD candidate in sociology and Mellon Cluster Fellow in science studies at Northwestern University. He studies the way experts, both lay and credentialed, construct their credibility and resolve epistemic conflicts.

References

Akrich

Rabeharisoa

(2023). On the multiplicity of lay expertise: An empirical and analytical overview of patient associations’ achievements and challenges. In Eyal

Medvetz

(Eds.), The oxford handbook of expertise and democratic politics (pp. 103–133). Oxford University Press.

Barker

K. K.

(2002). Self-help literature and the making of an illness identity: The case of fibromyalgia syndrome (FMS). Social Problems, 49(3), 279–300.

Barker

K. K.

Galardi

T. R.

(2011). Dead by 50: Lay expertise and breast cancer screening. Social Science & Medicine, 72(8), 1351–1358.

Borkman

(1976). Experiential knowledge: A new concept for the analysis of self-help groups. The Social Service Review, 50(3), 445–456.

Bosk

C. L.

(2003). Forgive and remember (2nd ed). University of Chicago Press.

Brown

Zavestoski

McCormick

Mayer

Morello-Frosch

Gasior Altman

(2004). Embodied health movements: New approaches to social movements in health. Sociology of Health & Illness, 26(1), 50–80.

Burnside

Lewis

Crocket

Wilson

Williman

Jefferies

Paul

Wheeler

B. J.

de Bock

(2020) CREATE (Community deRivEd AutomaTEd insulin delivery) trial. Randomised parallel arm open label clinical trial comparing automated insulin delivery using a mobile controller (AnyDANA-loop) with an open-source algorithm with sensor augmented pump therapy in type 1 diabetes. Journal of Diabetes and Metabolic Disorders, 19(2), 1615–1629.

Callon

(1986). Some elements of a sociology of translation: domestication of the scallops and the fishermen of St Brieuc bay. In Law

(Ed.) Power, action, and belief (pp. 196–233). Routledge & Kegan Paul.

Callon

Lascoumes

Barthe

(2011). Acting in an uncertain world: An essay on technical democracy ( Burchell

, Trans.). MIT Press.

10.

Collins

H. M.

(1981). The place of the “core-set” in modern science: Social contingency with methodological propriety in science. History of Science, 19, 6–19.

11.

Conrad

(2005). The shifting engines of medicalization. Journal of Health and Social Behavior, 46(1), 3–14.

12.

Dear

(2001). Science studies as epistemography. In Labinger

J. A.

Collins

(Eds.), The one culture? (pp. 128–141). University of Chicago Press.

13.

della Porta

Diani

(1998). Social movements: An introduction. Blackwell.

14.

Diabetes Control and Complications Trial Research Group. (1993). The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. The New England Journal of Medicine, 329(14), 977–986.

15.

Elbaz

(1992). The sociology of AIDS activism, the case of ACT UP/New york, 1987-1992 [PhD dissertation]. City University of New York.

16.

Epstein

(1995). The construction of lay expertise: AIDS activism and the forging of credibility in the reform of clinical trials. Science, Technology & Human Values, 20(4), 408–437.

17.

Epstein

(1996). Impure science: AIDS, activism, and the politics of knowledge. University of California Press.

18.

Epstein

(2023). The meaning and significance of lay expertise. In Eyal

Medvetz

(Eds.), The oxford handbook of expertise and democratic politics (pp. 76–102). Oxford University Press.

19.

Farrington

(2017). Hacking diabetes: DIY artificial pancreas systems. The Lancet, Diabetes & Endocrinology, 5(5), 332.

20.

Fleischmann

K. R.

(2006). Boundary objects with agency: A method for studying the design–use interface. The Information Society, 22(2), 77–87.

21.

Fox

N. J.

(2017). Personal health technologies, micropolitics and resistance: A new materialist analysis. Health, 21(2), 136–153.

22.

Gieryn

T. F.

(1983). Boundary-work and the demarcation of science from non-science: Strains and interests in professional ideologies of scientists. American Sociological Review, 48(6), 781–795.

23.

Greene

J. A.

(2016). Do-it-yourself medical devices–technology and empowerment in american health care. The New England Journal of Medicine, 374(4), 305–308.

24.

Greshake Tzovaras

Angrist

Arvai

Dulaney

Estrada-Galiñanes

Gunderson

Head

Lewis

Nov

Shaer

Tzovara

Bobe

Price Ball

(2019). Open humans: A platform for participant-centered research and personal data exploration. GigaScience, 8(6), giz076.

25.

Hatch

A. R.

Gordon

J. T.

Sternlieb

S. R.

(2019). The artificial pancreas in cyborg bodies. In Boero And

(Ed.) The oxford handbook of the sociology of body and embodiment (pp. 413–430). Oxford University Press.

26.

Hilgartner

(2000). Science on Stage. Stanford University Press.

27.

Hurley

Beane

Renault

Wright

Cassidy

Knibbs

Barber

(2014, December 24). Diabetes patients are hacking their way toward a bionic pancreas. Wired. https://www.wired.com/2014/12/diabetes-patients-hacking-together-diy-bionic-pancreases/

28.

Ikemoto

L. C.

(2017). DIY bio: Hacking life in biotech’s backyard. U.C. Davis Law Review, 51(2), 539–568.

29.

Jansky

Langstrup

(2022). Device activism and material participation in healthcare: Retracing forms of engagement in the #WeAreNotWaiting movement for open-source closed-loop systems in type 1 diabetes self-care. BioSocieties. Advance online publication. https://doi.org/10.1057/s41292-022-00278-4

30.

Kempner

Bailey

(2019). Collective self-experimentation in patient-led research: How online health communities foster innovation. Social Science & Medicine, 238, 112366.

31.

Klawiter

(2008). The biopolitics of breast cancer. University of Minnesota Press.

32.

Lewis

(2019a). History and perspective on DIY closed looping. Journal of Diabetes Science and Technology, 13(4), 790–793.

33.

Lewis

(2019b). Automated insulin delivery: How artificial pancreas closed loop systems can aid you in living with diabetes. Amazon Digital Services LLC - KDP Print US.

34.

Lewis

Leibrand

, & #OpenAPS Community. (2016). Real-world use of open source artificial pancreas systems. Journal of Diabetes Science and Technology, 10(6), 1411.

35.

Lupton

(2016). The quantified self. John Wiley & Sons.

36.

Mauck

A. P.

(2010). Managing care: The history of diabetes management in twentieth century America [PhD dissertation]. Harvard University.

37.

Melmer

Züger

Lewis

D. M.

Leibrand

Stettler

Laimer

(2019). Glycaemic control in individuals with type 1 diabetes using an open source artificial pancreas system (OpenAPS). Diabetes, Obesity & Metabolism, 21(10), 2333–2337.

38.

Melucci

(1980). The new social movements: A theoretical approach. Social Sciences Information, 19(2), 199–226.

39.

Morello-Frosch

Zavestoski

Brown

Gasior Altman

Mccormick

, Brian Mayer Community & Environment and PolicySociology. (2006). Embodied health movements: Responses to a “scientized” world. In Frickel

Kelly

(Eds.), The new political sociology of science: Institutions, networks, and power (pp. 244–271). University of Wisconsin Press.

40.

Myhre

J.R.

(2001). Medical mavens: Gender, science, and the consensus politics of breast cancer activism [PhD dissertation]. University of California.

41.

Nelson

(2011). Body and soul: The black panther party and the fight against medical discrimination. University of Minnesota Press.

42.

O’Donovan

(2007). Corporate colonization of health activism? Irish health advocacy organizations’ modes of engagement with pharmaceutical corporations. International Journal of Health Services: Planning, Administration, Evaluation, 37(4), 711–733.

43.

#OpenAPS Community. (2017). revision 30b2443b. https://openaps.readthedocs.io/en/latest/#

44.

Palmer

Greeley

S. A. W.

Letourneau-Freiberg

L. R.

Naylor

R. N.

(2020). Using a do-it-yourself artificial pancreas: Perspectives from patients and diabetes providers. Journal of Diabetes Science and Technology, 14(5), 860–867.

45.

Parsons

(1975). The sick role and the role of the physician reconsidered. The Milbank Memorial Fund Quarterly. Health and Society, 53(3), 257–278.

46.

Parthasarathy

(2010). Breaking the expertise barrier: Understanding activist strategies in science and technology policy domains. Science and Public Policy, 37(5), 355–367.

47.

Pinch

T. J.

Bijker

W. E.

(1984). The social construction of facts and artefacts: Or how the sociology of science and the sociology of technology might benefit each other. Social Studies of Science, 14(3), 399–441.

48.

Rabeharisoa

Callon

(2004). Patients and scientists in French muscular dystrophy research. In Jasanoff

(Ed.) States of knowledge: The co-production of science and social order (pp. 142–160). Routledge.

49.

Rabeharisoa

Doganova

(2021). War on diseases: Patient organizations’ problematization and exploration of market issues. In Geiger

(Ed.) Healthcare activism: markets, morals, and the collective good (pp. 55–85). Oxford University Press.

50.

Rabeharisoa

Moreira

Akrich

(2014). Evidence-based activism: Patients’, users’ and activists’ groups in knowledge society. BioSocieties, 9(2), 111–128.

51.

Sánchez Criado

Rodríguez-Giralt

Mencaroni

(2016). Care in the (critical) making. Open prototyping, or the radicalisation of independent-living politics. Das Altertum, 10(1), 24–39.

52.

Shahid

Lewis

D. M.

(2022). Large-scale data analysis for glucose variability outcomes with open-source automated insulin delivery systems. Nutrients, 14(9), 1906. https://doi.org/10.3390/nu14091906

53.

Staggenborg

(2022). Institutionalization of social movements. In The Wiley-Blackwell Encyclopedia of Social and Political Movements. Wiley. https://doi.org/10.1002/9780470674871.wbespm113.pub2

54.

Talbot

(2020, May 18). The rogue experimenters. The New Yorker. Retrieved February 17, 2021, from https://www.newyorker.com/magazine/2020/05/25/the-rogue-experimenters

55.

The Medical Futurist . (2016). What is living with an artificial pancreas like? Retrieved April 30, 2021, from https://medicalfuturist.com/living-with-an-artificial-pancreas/

56.

Thompson

(2016). Bigfoot biomedical: Reimagining diabetes care with a novel AP system (Vol. 3, pp. 20–27). The MedTech Strategist.

57.

Touraine

(1985). An introduction to the study of social movements. Social Research, 52(4), 749–787.

58.

von Hippel

(2017). Free innovation. MIT Press.

59.

Waidzunas

(2013). Intellectual opportunity structures and science-targeted activism: Influence of the ex-gay movement on the science of sexual orientation. Mobilization: An International Quarterly. Advance online publication. https://doi.org/10.17813/maiq.18.1.8353777g2t72j408

60.

Weber

(1963). The Sociology of Religion ( Fischoff

, Trans.). Beacon Press. (Original work published 1922).

61.

Weber

(1978). Economy and Society ( Roth

Wittich

(Eds.), Fischoff

, Trans.). University of California Press. (Original work published 1921).

62.

Woolgar

(1990). Configuring the user: The case of usability trials. The Sociological Review, 38(1_suppl), 58–99.

63.

Zabinsky

Howell

Ghezavati

, et al. (2020). Do-it-yourself artificial pancreas systems for type 1 diabetes reduce hyperglycemia without increasing hypoglycemia. Diabetes, 69(Supplement 1), 988-P.