The challenge of inclusive coproduction: The importance of situated rituals and emotional inclusivity in the coproduction of health research projects

Previous studies highlight that coproduced research initiatives are influenced by how individuals interact together and that group inclusivity amongst diverse members is cru-cial. However, it is not fully understood how inclusivity is sustained over time, particularly through routine encounters. Our study examines how coproduction occurs through routine and ritualistic patterns of everyday practices, which have the potential to facilitate sustainable and inclusive research initiatives. Using ethnographic data with four applied health research projects, we explored how everyday rituals generate and sustain inclusivity. Informed by interactional ritual change theory, we identify two types of interlinked inclusivity: relational, individuals routinely engaging together; and emotional, the feeling of being included. The process of producing and maintaining both types requires ongoing reflexivity from members. Groups with sustained inclusivity build interpersonal momentum through situated practices that enable them to mitigate external pressures and internal disagreements. Where groups experience a breakdown in inclusivity, they also experience a loss of momentum that makes them vulnerable to disintegration and collapse. Building and sustaining inclusivity are worked out through everyday interactions and - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - operate as a feedback loop that sustains the cohesiveness of the network and supports coproduction of knowledge. we through beyond the narrative We demonstrate importance ways how these and and provide and Specifically, this reports findings from four applied research to examine how interactions can support or hinder and how over to facilitate coproduction and implementation of health research. Comparisons across four cases highlight the variety in how coproduction is realized in practice, common social mechanisms that supported network sustainability and the mechanisms that undermined the sustainability of networks.


| INTRODUCTION
Over the last decade, a coproduction narrative has come to redefine health policy and research. It is widely acknowledged, for example, that patient and public groups should play a more meaningful role in the deliberation and formulation of policy options to ensure both their legitimacy and relevance to service users (Martin, 2008). In the field of health research, coproduction is promoted as bridging the "translation gap" between research and practice, where it is suggested that decision-makers, frontline professionals, service users, and other stakeholders should play an active role in problem definition, methodological design, data collection and analysis, and the application of evidence into practice (Smith & Ward, 2015). Coproduction facilitates the sharing of resources and expertise in ways that is often inhibited by organizational or occupational boundaries (Greenhalgh, 2010).
Beyond the idealized coproduction narrative, there remain many barriers to fostering a research environment in which diverse stakeholders can contribute to research, especially where there are varying priorities, inequalities in status, and different methods of practice (Flinders, Wood, & Cunningham, 2016). Previous studies highlight that coproduced research initiatives are influenced by how individuals interact and work together (Evans & Scarbrough, 2014;Flinders et al., 2016). In particular, the literature advocates the importance of inclusion and meaningful partnerships between all stakeholders (cf. Cooke, Langley, Wolstenholme, & Hampshaw, 2017;Rycroft-Malone et al., 2016;Rowley, Morriss, Currie, & Schneider, 2012). Taking the issue of inclusion as our focus, we suggest that there are two limitations within the translational research literature. First, the majority of studies focus on establishing coproduction initiatives and do not always explore how these initiatives develop over time and sustain a sense of inclusivity. Second, it is not known how the localized or situated interactions between actors facilitate a sense of inclusivity. Taking a decentered approach (Bevir, 2013), we address these two issues through looking beyond the aspirational narrative of policy to examine how different stakeholders, each with distinct histories and interests, interact in the processes of attempting to coproduce research. We demonstrate the importance of attending to the ways situated practices take the form of interactive rituals, how these rituals (re)create shared traditions and meanings and provide the basis for inclusivity, especially in the face of ambiguity and dilemmas. Specifically, this paper reports findings from four applied health research initiatives to examine how everyday interactions rituals can support or hinder inclusivity and how inclusivity is sustained over time to facilitate the coproduction and implementation of health research. Comparisons across four cases highlight the variety in how coproduction is realized in practice, common social mechanisms that supported network sustainability and the mechanisms that undermined the sustainability of networks.

| BACKGROUND
There is little consensus in how coproduction is defined (Orr & Bennett, 2012), and variations can be seen across the worlds of policy-making, service delivery, and knowledge creation. For some, it can be broadly defined as collaboration between "providers" and "end users" with the goal of promoting "user influence and ownership" of policies and services (Durose, Needham, Mangan, & Rees, 2017;Ostrom, 1999;Pestoff, 2014, p. 385). The narrative of coproduction in research extols a move away from "experts" generating and disseminating unidisciplinary knowledge, to a situation where a variety of stakeholders, problem-owners, interdisciplinary experts, and end-users collaborate to coproduce contextrelevant evidence (Gibbons et al., 1994;Rowley et al., 2012). Coproduced research "involves the close interaction of many actors throughout the process of knowledge production" (Gibbons et al., 1994, p. vi). In the field of health services research, coproduction is often advocated because evidence production does not always appear to reflect the needs of service providers and users and because there are often long delays in translating evidence into practice (Rowley et al., 2012). The purported benefit of coproduction is that knowledge is generated in collaboration with those who will utilize this knowledge, making it closer to context and bridging the all-important translation gap (Rycroft-Malone et al., 2016).
The coproduction of health service research therefore sees a multiplicity of stakeholders collaborating in the seamless generation, translation, and implementation of new evidence, including patients, practitioners, commissioners, managers, and local authorities. Nonacademic involvement in health research can vary from being on an advisory board to user-led research (Mathie et al., 2014). The public and patient involvement (PPI) agenda in healthcare has successfully established the principle that patient and public perspectives should be involved in policy-making, service planning, and evidence production; although the extent to which PPI is realized meaningfully remains open to debate (Martin, 2008). However, meaningfully engaging diverse stakeholders in research remains challenging as there may be varying levels of experience, multiple ontological and epistemological differences, and competing values, pressures, and goals (Flinders et al., 2016).
Coproduced research can be vulnerable to what Flinders et al. (2016, p. 269) describe as an "expectation gap," where stakeholders and research team members have different expectations of what will occur in the project and how and when it will occur. Health research processes, for example, are often time-consuming and have limited immediate impact on services; as such, they can appear out-of-step with the changing evidence needs of service leaders or the evolving experiences of service users. In such situations, sustaining the involvement of diverse stakeholders remains challenging, with the potential consequence that the very power and hierarchical differences coproduction seeks to equalize may actually be reinforced (Flinders et al., 2016). The challenges of ensuring inclusivity of stakeholders is arguably one of the main factors influencing the success of coproduction in contributing to public service innovations (Fung, 2015).
To counter these challenges, previous studies have identified a number of factors that influence the effectiveness of collaborative work, including leadership, geographical proximity, stakeholders' knowledge and experience, resources, navigating between differing epistemological bases, managing relational complexity, opportunities for feedback, and the pay-off for involvement (Baker & Irving, 2016;Brown et al., 2016;Currie, Enany, & Lockett, 2014;Fitzgerald & Harvey, 2015;Fung, 2015;Rycroft-Malone et al., 2016).
Of particular relevance to our work is the idea that coproduced research should be inclusive of relevant stakeholders. For example, Brown et al. (2016, p. 9) highlight that a "culture of inclusivity" is crucial, whereas Rycroft-Malone et al. (2013, p. 24) argue that a lack of belonging has a "negative impact on engagement." In this regard, the idea of inclusive coproduction has close similarities with research and theory on situated and social learning, especially the "communities of practice" concept, which describes how people engage in collective learning through shared domains of activity (Lave and Wenger, 1991). Significant here is the idea of active participation in shared practices, many of which can have a taken-for-granted quality as people build up shared repertoires and routines as they seek out and share information in the processes of problem-solving. In this sense, coproduction resembles a form of social learning where stakeholders come together to both learn from one another and, in doing so, contribute to the coproduction of new knowledge. Nonetheless, most studies exploring coproduction do not define inclusivity but implicitly associate it with bringing relevant individuals together to share knowledge, resources, and expectations (Brown et al., 2016;Kislov, Waterman, Harvey, & Boaden, 2014;Rycroft-Malone et al., 2016). These studies do not fully explore how inclusivity is generated and sustained. For instance, Kislov, Harvey, and Walshe (2011) acknowledge that it is unclear how multiprofessional teams transition to "genuine" communities that share practice and have a sense of belonging. Similarly, Evans and Scarbrough (2014, p. 125) contend that knowledge can be translated through a "continuous and incremental process" of "routine, day-to-day practices" between individuals from various communities, which they describe as a "blurring approach," but they do not explain how this process works in practice.
Our paper looks beyond the narrative of coproduction found in health policies, patient advocacy movements, and health research, to look closer at how inclusive stakeholder participation is created (or undermined) through the situated practices of coproduction. Following the idea that inclusivity and involvement are often realized through the formation of shared repertoires and routines (Lave and Wenger, 1991), our study focuses on the importance of rituals as characterizing and shaping the collaborative practices of coproduction. Our focus on shared rituals offers not only new insight about the localized practices of coproduction but also, as we suggest, a basis for explaining how inclusivity is realized.
Ritual theory deals with the way symbolic meanings, beliefs, and norms are transmitted and reinforced through relative stable patterns of social interaction or routine (Goffman, 1967). Collins (2004) has developed the concept of Interaction Ritual Chain (IRC) as a way to analyse the temporal, spatial, and affective connections between everyday rituals. For Collins (2004), rituals consist of involved social actors interacting together within a shared spatial and temporal situation, where there is a boundary to legitimate participation and where participants have a shared sense of purpose and emotional experience. Significantly, he sees rituals as social situations of mutually focused emotion and attention that produce a shared rhythm, symbols of group membership, and feelings of unity. Successful rituals foster solidarity amongst participants, the emergence of shared symbolic meanings and beliefs, a sense of belonging, and importantly, emotional energy (EE), or positive feelings associated with belonging to a community. For Collins, rituals do not occur in isolation but are linked together through a "chain" of interactions, which are made possible through the sense of belonging, shared symbols, and importantly, positive emotions. Conversely, unsuccessful rituals undermine a sense of belonging and are associated with negative emotions, thereby discouraging future participation. In terms of sustainability, rituals that consistently produce feelings of confidence and unity encourage individuals to continuously work together (Summers-Effler, 2004). For instance, Vertesi (2012) notes that through everyday encounters, such as team meetings, a collaborative approach facilitates unity and innovative solutions to challenges.
This research highlights the necessity of observational research over time to understand the importance of these interactions.
IRC theory, therefore, offers a distinct perspective in thinking about the situated practices of coproduction. It complements existing research on the importance of maintaining stakeholder inclusivity, by highlighting the importance of everyday routines and rituals in fostering a sense of belonging and, importantly, EE. Our study examines how coproduction occurs through everyday situated interactions in a recursive pattern that enables sustainable research initiatives. Over the life of a research project, these interactions relate, for example, to aspects of study design and planning, research governance, sampling, data collection and analysis, and dissemination, which are organized through a variety of interactive settings.
With coproduced research, the gap in our understanding is how individuals use their interactions to negotiate different beliefs and practices to collectively implement the intervention. Our research assesses how coproduced research is conditioned by the emergence of group unity and a shared sense of belonging amongst researchers, clinicians, and stakeholders. By taking a microinteractional approach and using observational methods, we question what inclusivity looks like in practice, how it is sustained over time, and how it influences the coproduction of health research.

| Study design
This research used an ethnographic comparative case study approach with four applied health research projects, each explicitly using coproduction approaches to generate knowledge. Ethnography aims to develop a "rich" descriptive account and interpretative understanding of social and cultural processes, usually involving direct observations of the everyday social interactions of a community (Fetterman, 2009). For our study, this involved observations of both the formal and informal routines and situations of the research process, from the early stages of project design and start-up through access negotiations and data collection to analysis and dissemination. Following an ethnographic case study approach, the study aimed to describe the particularities and complexities of coproduction as manifest in this sequence of happenings (Stake, 1995) of different project teams, with the goal of understanding how the particular localized rituals of coproduction produce distinct effects in terms of their EE, sense of belonging, and in sustaining stakeholder inclusivity. Through developing four distinct in-depth accounts, the study then moved to develop interpretative comparisons to explain the similarities and differences between research projects (Eisenhardt & Graebner, 2007;Stake, 1995). All appropriate (U.K.) ethical approvals were obtained.
Four "cases" were purposively selected on their explicit use of coproduction; differences in their clinical area and intervention; their implementation and mobilization strategy; and range of stakeholders involved in their project description. These were identified through a longstanding research partnership between local health and care providers and three regional universities. All initiatives explicitly followed a coproduction philosophy. A description of the case studies appears in Table 1. To help ensure confidentiality and anonymity, all individual and project names are anonymized, and individual roles are noted as "research team" or "stakeholder" in fieldnote extracts. All appropriate (U.K.) ethical approvals were obtained. given to how initiatives were organized, how often they met, where they met, membership and continuity of membership, whether and how networks disagreed and resolved tensions, what was discussed, how they engaged with wider stakeholders, and whether member feedback was actioned. As part of the observations, opportunities were taken to talk with project team members about their experiences and reflections of group activities to clarify and extend the observational data. Observations were recorded in a field journal before summary notes were typed up and shared with the study team for reflection and preliminary analysis.

| Data collection
In addition, 39 semistructured interviews were carried out across the four projects (Burnside 9; Terwilliger 12; Quimby 10; Marshall 8), undertaken with lead researchers (4), methodologists (2), project researchers (9), PPI representatives (3), health professionals (10), service managers (6), and project administrators (5). Interviews were digitally recorded and transcribed verbatim and included questions on how participants became involved with the research initiative, how they experienced being involved, what motivated them to continue involvement over time, what they felt worked well, and what was challenging.

| Data analysis
Interpretative data analysis was carried out concurrently with data collection, with identified issues informing ongoing data collection (Corbin and Strauss, 1991). In keeping with case study approach, the aim of the analysis was to first describe and explain the individual cases, but where empirical themes and questions identified in one case, they were asked of others (Stake, 1995). Clarke developed reflective summaries and earlier interpretations of field data that were shared with the other authors at monthly team meetings. This provided the primary basis of case description and comparative analysis, with a focus on telling the story of the routines and rituals of individual project teams, before moving to think about similarities and differences. In these, and subsequent, data analysis meetings alternative interpretations were explored by the authors and provided the basis for subsequent data collection. As the study progressed, the aggregated body of observational, documentary, and interview data were subject to qualitative coding (Braun & Clarke, 2006). This was initially undertaken by Clarke, and again codes and extracts of data were reviewed by all authors in regular meetings, where particular attention was given to the internal consistency and relationships between codes and themes. Through this process, analysis focused on categorizing the project team composition, the configuration project meetings and events, the patterns of interaction with project events, the instances of conflict and disagreement, and the reactions and reflections of stakeholders to their participation. The interview data were analysed, for example, in terms of how participants described their involvement, whether they felt valued, frustrated, and their perspective on what worked well and areas that could have been improved. Through ongoing review of descriptive code and themes, each author developed interpretations for why cases evolved as they were observed and how they varied, with IRC used as a heuristic framework for elaborating and extending interpretations. Four key themes became apparent including the structure of the projects and common rituals, generating inclusivity, sustaining inclusivity and barriers to inclusivity.

| Structure of research initiatives and common rituals
Each of the projects used different approaches for involving stakeholders. They each had a "core" research team consisting of the principal investigator(s), coinvestigator(s), researchers, research managers, administrator, and for some, a trial manager and statistician; and also a diverse group of more "peripheral" nonacademic stakeholders, including public involvement members, healthcare professionals, practice managers, commissioners, local authority staff, and third sector staff. In line with the principles of coproduction, all projects were configured on the explicit premise of academic and nonacademic stakeholders working collaboratively to coproduce knowledge relevant to the needs of local service providers, commissioners, users, and other constituents. As shown below, however, they varied to the extent that this was realized, especially whether more "peripheral" and "core" members participated in rituals through which they functioned as an inclusive community of practice (Lave and Wenger, 1991).
We observed that each project engaged in a number of common interaction rituals, typically focused on structured and unstructured meetings between "core" project members and "peripheral" stakeholders, as well as various study site visits and engagement activities. Although these varied in organization, focus, and frequency, across the four project teams, a number of prominent meetings were identified in which we observed common interactional or ritualistic features through which degrees of inclusivity were enabled (Table 2). For example, most of the projects groups met on university or hospital premises, and one (Marshall) held stakeholder meetings at a general practitioner (GP) practice. The frequency of meetings varied from twice a month to every 6 weeks. Burnside included public involvement members and a healthcare professional in their regular "core" team meetings every 6 weeks and had wider meetings for all stakeholders every few months. Quimby, and in the beginning, Terwilliger, engaged wider stakeholder members in their meetings, but these were rarely regarded as "core" members. When Terwilliger encountered difficulties with recruiting service users to their study, these more inclusive meetings became increasingly infrequent, and the "core" team began meeting more exclusively to address operational issues. With the exception of Terwilliger, the projects sustained regular membership over 3 years, though individual attendance fluctuated, particularly with Quimby and Burnside.
All rituals involved varying degrees of shared attention and emotion engagement that had the potential to produce to contribute to a sense of inclusivity. Many of these rituals were characterized by highly localized microinteraction between different people, but over time, these coalesced into collective patterns of interconnected interaction chains. For instance, the time before a meeting started was often a time for members to catch up and get to know one another through small interactions. Often, the emotional residue from these interactions would set the tone for the more formal meeting: The friendly atmosphere at the start of this meeting continued through subsequent interactions, including during the discussions of bringing patients in for their "baseline" assessments for the trial: It is remarked that on those days, their corridor will be very busy. "It'll be alright" a stakeholder says.
"That's the motto of this study," replies Martina. "It'll be alright," she repeats. Everyone laughs and jokes about getting t-shirts made.
Because these rituals were repeated throughout the research process, they became moments where inclusivity could be generated and sustained.
Our paper focuses on these meeting rituals because they were the most frequent sites for stakeholder involvement and also directly observable. Other rituals occurred but these others were rarely observed due to the difficulties of acquiring access, such as ad hoc meetings with a few project team members and stakeholders and clinic visits where the research focused of clinical interventions. The findings look, in particular, at how inclusivity was generated and then sustained through these ritual interactions, focusing primarily on the case of Burnside, Marshall, and Quimby as instances of positive inclusion, and then the experiences of Terwiliger as a negative case study.

| Generating inclusivity
In the early stages of project team formation, we observed a number of prominent interaction rituals that were designed to, and seemed particularly important for, generating an initial sense of inclusivity amongst stakeholders.
For example, each project team convened regular Project Meetings to oversee research progress, including standing items for "public involvement feedback," which were explicitly aimed at encouraging nonresearchers to engage in the research process. But as Brown et al. (2016) and Rycroft-Malone et al. (2016) note, a well-structured group that brings together individuals does not necessarily generate inclusivity and collaboration. In our case studies, project teams had to generate an environment where stakeholders could engage with the different stages of the research process. Facilitating engagement involved, for example, informing stakeholders of project developments, before and after meetings, and providing space for opinions to be expressed. During one Quimby meeting, we observed that the group discussed the difficulty of submitting an amendment to ethics in which stakeholders were invited to contribute to the task of amending the ethics application with the core research team: Through the common ritual of inviting stakeholder feedback, most interview participants reported feeling included and valued in their respective initiatives, with the exception of the Terwiliger project:

So, I think there's been an awful lot of learning off of each other that will be used outside of the [Quimby] research project. (Roberta)
Interviewees responded that this feeling of inclusion was linked to the opportunities to voice opinions, ask questions, provide feedback, and observe how suggestions were implemented by the research team. For example, during a Burnside meeting, the group discussed their website, again allowing space for expression of suggestions and opinions:

Stefania (research team) says they will have a mission statement on the website that includes things like
their values, ethos. Arun (stakeholder) also says that, "we will listen, we will keep information confidential, etc." Though Arun would not be interacting directly with patients, he used the term "we" rather than "you." Similarly,

Marshall regularly acted on input from their members:
There is some discussion about training practice nurses to go through the booklets with patients. Generating inclusivity could occur through most of the rituals observed in the meetings. At this early stage, it seemed important for the core project team to provide space for questions and feedback, to facilitate stakeholders having a direct impact on the project, and the willingness of all members, particularly the stakeholders, to question and voice their opinions. Through these ritual encounters, we observed how stakeholders could begin to form a sense of cohesiveness and belonging to their respective research projects.

| Sustaining inclusivity
As the project teams settled into more regular routines and meetings, we observed how interaction rituals could serve to sustain inclusivity amongst stakeholders, where rituals resulted in a sense of solidarity and positive emotions.
Sustaining inclusivity included using project meetings, events and site visits to proactive involve stakeholders regu- whether through a lack of recruitment, changing priorities, and funding reviews. When difficulties arose, the majority of projects used rituals to openly discuss and manage these challenges. These interactions were described as positive by most members:

And it is great to sit round a table when you have got so many different people with so many different ideas because then you can develop your study as a result of a conversation that you have had about one single thing and evolve it. (Tanya)
Most members from the different case studies felt comfortable discussing various aspects of the projects with stakeholders present: "there's never been any issues in terms of us feeling uncomfortable to raise certain things" (Nicole).
Conversely, other members reflected on an awareness of impression management to stakeholders with one member of Quimby comparing the process with a "swan […] paddling furiously underneath." Two members of Quimby talked about not wanting to bring "nitty-gritty" details to meetings and avoiding "mixed messages," only sharing about difficulties after they had been addressed.
At times, challenges were not easily addressed through group rituals (or, worse, rituals were used to conceal pressures). In such situations, the absence of interaction put pressure on group cohesion. For Marshall, the national policy of diabetes prevention and treatment changed, and it became clear that it would not be possible for the intervention to be commissioned in its current format. Because of the project's ongoing links with wider stakeholders and commissioners, they were in a position to modify their intervention to fit with national priorities and get commissioned: "[I]t's not going to be a wasted result it just means the commissioning is different, but the learning will still be exactly the same" (Jerome). Worries about recruitment were a feature of all projects, particularly for studies that relied on targets to enable the statistical interpretation of results. Burnside, for example, struggled to recruit patients from hospital emergency departments (ED), even after presenting and explaining the study to ED staff during a site visit: One staff member explains that the problem is they could never be sure that the patient did not have a health problem. So, they would run all the tests to not rule it out. Then they would pass these results on to the GP. So, the suggestion was to speak to the GP rather than the ED staff.

(ED Team visit)
Although some ED staff felt that the project could be useful, others cited a lack of time to discuss the study with patients and the overriding concern of treating medical emergencies. To overcome low recruitment from GPs and the ED, Burnside expanded their project's wider group of stakeholders, especially more GP practices and clinical staff in outpatient services. This include a GP with an interest in mental health who offered alternative methods of recruitment, which were quickly actioned resulting in significantly increased recruitment numbers.
Sustaining a sense of inclusivity amongst stakeholders necessitated that projects continued to regularly engage stakeholders in increasingly normalized project rituals through which differences in expectations could be resolved; this in turn generated a sense of shared ownership of the projects beyond the research team to also include stakeholders.

| Working against inclusivity
Inclusivity was not a feature of all the case studies. Although Terwilliger initially had large and inclusive meetings, with some members reporting feeling a positive momentum, not everyone felt comfortable with the large size and expressed confusion over member roles and involvement in project meetings: "[T]here just seemed to be a lot of people that I did not know what they were doing" (Simran). A lack of role clarity and communication continued to be a feature of Terwilliger as more difficulties were encouraged with site and participant recruitment. A key feature was the relationship between core academics and peripheral nonacademics, with a growing sense of disconnect and exclusion in terms of participation in meeting rituals. One member reported that she suggested having wider stakeholder meetings beyond the core research team to generate clinical and commissioning feedback and learning, like Burnside, yet this suggestion was not adopted: "[D]o you ever get that, it's like when you dream, I think we keep saying something and no one is listening. That's what it felt like" (Anita). Similarly, one stakeholder expressed disappointment that feedback was sometimes ignored: "So you felt like you'd given some really great input and it'd been taken on board and then steps seem to go backwards again, just a few months down the line" (Natasha).
It was further observed that documents like information sheets or website templates were always not circulated to stakeholders prior to the meetings making it difficult to participate in group discussion. Even passing out the handouts seemed exclusive in terms of who got their own copy (usually some of the research team) and who had to share (usually the stakeholders). Members also disagreed with another, but without fully discussing or resolving tensions:

Hans (research team) and Jessica (research team) explain they have been working on standard operating
procedures. However, they did not bring enough copies for everyone to view them during the meeting.
Hans says he could circulate them to everyone for comment. Gordon (research team) nods but Manuel (research team) says, "Well who is it for? It depends on who it is for." This is not answered.

Gordon questions one of the steps about GP contact. Hans and Jessica exchange a look and Raymond
(research team) steps in to say it depends on the site and where it is at in the process. Gordon interrupts saying there needs to be mention of GPs. At this, Raymond looks at me, rolls his eyes and shakes his head.
[…] Hans says this is perhaps why it needs to be circulated so that people can comment and discuss these

things. Manuel replies again that it depends on who it is for […] It is not resolved and the conversation moves on. (Meeting)
It was found that the core Terwilliger research team discussed many aspects of project management and progress away from full team meetings and did not always update stakeholders about these conversations. Thus, it was sometimes unclear who was doing what: Talk then moves to one of the research sites. Anita (research team) says that the letter of approval to start recruitment is coming through soon. There will be two therapists at this site. Manuel (research team) explains there are questions over who is going to provide backfill. […] There has been some work with the medical director and it is getting the balance between recruiters and therapist capacity. But it is unclear who is doing this work, and more importantly, who will help resolve the issue of backfill. (Meeting) As the project unfolded, problems continued, particularly around low recruitment numbers. Managing these challenges was difficult and time consuming: "By trying to satisfy so many different masters all at once, it just takes time to consult with all of those teams and there are so many different needs" (Nathan). The wider organization that oversaw the project and its budget ultimately decided to close Terwilliger prematurely. Although members could not identify a single cause for the project's failure, the majority expressed displeasure at how the difficulties were managed and a lack of communication, including by the wider organization: An interview participant reflected he had not seen "that level" of "unprofessionalism" in the 10 years that he has been in research. By the time the project had their funding stopped prematurely, the core team expressed frustration, anger, and tension, and at times, members reported heated internal arguments. One of the final meetings was ). Our research highlights that although these factors are important in the early stages, over time, they are not necessarily enough to sustain these inclusion of stakeholders, particularly when challenges occur. Like other communities of practice (Lave and Wenger, 1991), research groups must maintain and extend inclusion of participants through situated everyday routines and rituals, especially where positive emotional encounters help to reinforce their shared sense of identity and solidarity within the community. Our study illustrates this process of inclusivity in two ways that have the potential to be mutually reinforcing in both positive and negative ways. The first is through relational inclusivity whereby the unique insights, capabilities, and resources of individuals are (or are not) actively recognized, valued, and included in group routines and rituals. The second is through emotional inclusivity, whereby individuals experience (or do not experience) positive emotions through participation in group meetings, which reinforce their sense of belonging and encourage their continued involvement. The existing literature often emphasizes the importance of relational inclusivity, that is, where individual stakeholders are actively engaged and facilitate to participate in group activities (Brown et al., 2016;Kislov et al., 2014;Rycroft-Malone et al., 2016), but these are seldom analysed as rituals of coproduction that have an affective quality. Together, these aspects of inclusion enable coproduction initiatives to sustain situated practices in the face of challenges and disagreements.
In terms of relational inclusivity, all of the case study research initiatives began with structures and processes to facilitate the inclusion of diverse individuals, organizations, and resources. None of them were "user-led," but the majority used facilitative methods to engage and include nonacademic stakeholders throughout the phases of research. Projects had to operate with an internal level of cohesion that remained consistent over time in order to withstand routine research pressures. Additionally, they had to manage pressures on stakeholders' time, varying degrees of stakeholders' research experience, and navigate the power imbalances between academic and nonacademic members. Our research suggests that sustaining inclusivity in the face of these challenges does not mean an absence of disagreement, rather it affords individuals the opportunity to discuss their disagreements, ask questions, and find resolution collectively (Vertesi, 2012). Importantly, this occurs through routine, rather than once-off or extraordinary, situated practices. Similar findings have been made elsewhere. Pestoff (2014), applying Ostrom's (2009) structural variables to coproduction, notes that sustainable groups are more likely to be those that have opportunities to meet regularly, to observe each other, and to build shared expectations and goals and are of a size where individuals can get to know each other.
Emotional inclusivity builds upon relational inclusivity by generating a sense of belonging within the group that provides the motivation for individuals to continue their participation (Collins, 2004). Over time, the majority of projects continued to build consistent group expectations, whereby individuals were clear in their own and others' roles and knew what to expect from the group. Shared ownership occurred through transparency of difficulties and acting upon stakeholder feedback. For the majority of the cases, everyday rituals such as asking questions and generating discussion were "hallowed" as they were full of EE (Waring & Bishop, 2010). These rituals were not mere rote; they were full of symbolic meaning and positive expectations, particularly as these rituals solidified over time. We suggest that hallowed rituals are indicative of this second component of inclusivity, namely, the emotional state of feeling included, of belonging, and of feeling valued within a group. These feelings of inclusivity are necessary for successful, and sustainable, coproduction (Brown et al., 2016;Pestoff, 2014;Rycroft-Malone et al., 2013;Rycroft-Malone et al., 2016).
We suggest that relational inclusivity does not automatically lead to emotional inclusivity. Simply bringing diverse individuals together in interaction rituals, such as meetings, does not mean that interactions will generate a shared sense of identity and solidarity. The Terwilliger project, for example, had similar rituals to the other projects, but these did not generate positive emotions through stakeholder participation. When Terwilliger encountered challenges, the "core" project teams restricted their meetings to the exclusion of more "peripheral" stakeholders. One interpretation of Terwilliger's approach is that coproduction is strategic, only inviting stakeholder involvement at set times, which exacerbated the potential "expectation gap" amongst participants (Flinders et al., 2016). The consequence of varying expectations and interrupting the relational inclusivity of the group meant that the emotional inclusivity and momentum was eroded. When they did resume regular meetings, some stakeholders stopped attending and reported that they had moved on to other priorities. By the time the project was in its final days, interactions had fractured and lacked any kind of symbolic meaning. Their rituals could be described as "hollow" as "the activities no longer had relevance" to the "activity" (Waring & Bishop, 2010). Thus, it is not what the ritual is that facilitates emotional inclusivity, but how individuals interact within them.
Establishing and maintaining inclusivity to coproduce knowledge is therefore an ongoing process involving both relational and emotional inclusivity. Returning to the "community of practice" literature (Lave & Wenger, 1991), our findings suggest that sustained peripheral participation and the gradual progression towards more central positions within a community not only relies upon relational inclusion in activities, but also on emotional inclusion through which "feelings" of belonging reinforce the "practices" of belonging. The wider literature on student "drop-out" highlights an interplay between relational and emotional isolation as shaping student commitment (Bennett, 2003). We elaborate this in Figure 1 as a recursive or circular process whereby relational inclusive can drive emotional inclusion, which in turn foster greater relational inclusion, and so on. For some communities, positive interaction rituals promote a "virtuous cycle" of belonging and inclusion, but as we also show where these rituals are experienced more negatively, they can lead to a "vicious cycle" of gradual exclusion and marginalization.
Elaborating our model further, we propose that "relational inclusivity" in the form of repeated everyday rituals is a necessary first step in generating and sustaining inclusivity in coproduction processes. As the literature suggests above (Pestoff, 2014;Rycroft-Malone et al., 2016) ances. Together, and over time, these everyday interaction rituals can combine to generate interpersonal momentum leading to "emotional inclusivity." This momentum can help research initiatives mitigate risks to wider issues of sustainability, such as changes in priorities or power imbalances, as they are already working well together.
This momentum can produce a sense of belonging, feeling valued, shared learning, and ultimately, for coproduction work, knowledge grounded in practice (Collins, 2004). As membership within research initiatives can frequently fluctuate, successful initiatives will be able to incorporate newer members easily, so there is not a break in momentum. Interactional histories also matter as all interactions are connected through chains (Collins, 2004). If interactions within a group are consistently generating feelings of unity and belonging, researchers and stakeholders will be motivated to continue to coproduce knowledge, despite the challenges and the real differences in terms of power and hierarchy. Done well, the everyday interactions will produce virtuous, rather than vicious, cycles of interpersonal dynamics. That is, consistency of relational inclusivity through everyday encounters enables coproduced initiatives to develop emotional inclusivity and interpersonal momentum, which better equips groups to navigate through changes and challenges.

FIGURE 1 Inclusivity in coproduction
Our study also contributes to the growing interest in, and application of, experience-based codesign in the redesign of health services (Bate and Robert, 2006). The experiences of applying experience-based codesign, as a form of coproduction, suggest that it needs to be valued as a creative process informed by the lived experiences of different stakeholders (Greenhalgh, Jackson, Shaw, & Janamian, 2016). This calls for forms of facilitation that ensures not only stakeholder engagement and inclusion to manage knowledge or power imbalances, for example (Crompton, Waring, Roe, & O'Connor, 2017), but also strategies to more fully recognize and manage the emotional qualities and implications of codesign activities.
We also offer a distinct decentered analysis of coproduction (Bevir, 2013) that looks beyond the elite narratives of policymakers or even PPI advocates, to show how stakeholder inclusion is realized through the situated and meaningful practices of local actors. The affective dimensions of the decentered perspective are not always recognized, but we would argue that the recreation and elaboration of traditions is realized through ritualized social practices within which meanings are reinforced and recreated, which are sustained through actors' feelings of inclusion and belonging. By this, we mean that traditions not only have to be meaningful but also have to be affective, and it is through interaction rituals that this is made possible.
There are some limitations to this study and areas for future research. Members within a research initiative are connected to wider groups and networks, and further studies exploring these groups would aid our understanding of what changes as a result of coproduction. It is also not known what happens to the relationships between members in the long term, particularly once there is no further funding to help support coproduction activity or members move on to other projects. Moreover, Kislov, Walshe, and Harvey (2012) have identified that where individuals have successfully worked together before, further collaborations are more likely to succeed. Future research could examine the relational and interactional history of members as there may be factors that better position individuals to engage in coproduction.

| CONCLUSION
If successful, coproduced research initiatives generate a shared interpersonal momentum between members that continues to motivate individuals to spend their time, energy, and resources. The task of establishing inclusivity is not something that is ever completed, as inclusivity is continuously negotiated through every day and mundane situated practices. These encounters add up over time, creating expectations, setting the "order" of how the group operates, and providing members with implicit social guidelines for what can be discussed, questioned, and actioned. Our paper has implications for both health research and policy. Efforts at an authentic sense of coproduction must generate and sustain inclusivity over time (Rycroft-Malone et al., 2016). Attention to how coproduction is formed and performed can greatly facilitate meaningful and successful collaborations between diverse members.