Open Research Online Rethinking Research Partnerships: Evidence and the politics of participation in research partnerships for international development

: This article responds to the drive for research partnerships between academics and practitioners, arguing that while potential bene ﬁ ts are clear, these are frequently not actualized resulting in partnerships that are ineffectual or worse, exacerbate damaging or inequitable assumptions and practices. In order to understand/improve partnerships, a systematic analysis of the interrelationship between what counts as evidence and dynamics of participation is proposed. Drawing on data from a seminar series and iterative analysis of seven case studies of partnerships between Higher Education Institutions and International Non-Governmental Organisations, the article concludes by suggesting substantial shifts in the theory and practice of partnerships. © 2019 The Authors Journal of International Development Published by John Wiley & Sons Ltd


INTRODUCTION
In a world increasingly recognized as complex, uncertain and interconnected and facing pressing environmental and social challenges, the contribution of research to understanding and facilitating responses is gaining traction (e.g. Conway & Waage, 2010;Bardsely, 2017). In 2015, the British government launched the £1.5 billion Global Challenges Research Fund (GCRF), marking a significant increase in research funding as part of the official development assistance budget. The fund is framed as a 'solutions focused' and 'challenge-led' initiative that will 'harness the expertise of the UK's world-leading research base to strengthen resilience and response to crisis in developing countries' (Independent Commission for Aid Impact (ICAI), 2017: I). It also emphasizes the importance of 'research partnerships' between UK-based academics and development practitioners. This drive to partnerships is partly fuelled by the British government's mandate to demonstrate 'official development assistance compliance' (Organisation for Economic Co-operation and Development (OECD), 2008) by linking research to development practice and policy in Development Assistance Committee (DAC)-listed countries. It is also rooted in a particular set of ideological assumptions and agendas, including those around 'international development' (specifically as 'the promotion of the economic development and welfare of developing countries'-Department for International Development (DFID), 2016:ii, original not italicized); 'accountability' (as manifested in DFID's 'results framework'); and 'expertise' (as set out in DFID's research strategy which highlights the role of the UK as 'a global leader in scientific, research and technical expertise' implying a unidirectional export model for research-DFID, 2016). At the same time, the GCRF also taps into a broader contemporary discourse around impact, the logic being that involving development practitioners and policymakers in research processes will lead to better integration of research into policy and practice. This logic is substantiated by a significant body of literature, which has evolved understandings of research into policy from a simplistic linear formulation of knowledge transfer to more complex understandings of research ecosystems that encompass knowledge production, communication, access, uptake, adaptation and use (e.g. Georgalakis, Jessani, Oronje, & Ramalingam, 2017; NESTA/Alliance for Useful Evidence, 2016); the engagement of end users and mediators in research processes as well as with outputs (Fransman, 2018); 'community-based research' (e.g. Hall & Tandon, 2017); the open access movement with related work around citizen science and data activism (e.g. Smith & Reilly, 2013); and partnerships between Higher Education Institutions (HEIs) and International Non-Governmental Organisations (INGOs) (e.g. Aniekwe, Hayman, & Mdee, 2012;Hanley & Vogel, 2012, Stevens, Hayman, & Mdee, 2013. This literature suggests there are clear benefits to research partnerships between academics and development practitioners. In addition to potential instrumental benefits (including greater access to data for academics and to technical expertise and infrastructure for INGOs), the broader social benefits of partnerships might include a shared commitment to responsive research that balances rigour and relevance, promoting 'engaged excellence' (Leach, Gaventa, & Oswald, 2017); a shift to inter/trans-disciplinarity that extends research into 'the real world' beyond the silos of the academy (e.g. Lang et al., 2012); and a redistribution of global research resources and outputs with implications for the integration of alternative epistemologies, ontologies and ethics into theory and practice (e.g. Connell, 2007;De Sousa Santos, 2014;Hall & Tandon, 2017). However, the same literature also identifies significant challenges including divergent priorities, schedules and capacity, hierarchies of knowledge and constraints to participation.
Responding to these challenges, a wave of studies and initiatives has focused on providing resources and guidelines for effective and/or equitable partnerships. 1 While many of these resources have contributed useful insights into the workings of partnerships of different types and in different contexts, they tend to be descriptive rather than analytical, focused at the institutional level (with limited attention to the role of subjectivities, materialities and broader socio-political processes) and adopting a somewhat simplistic idea of 'partnership' as a balanced relationship between two coherent and static entities engaged in a discrete and short-term collaboration framed by specific grants and projects rather than as dynamic/emergent research relationships. Moreover, the studies tend to focus on research implementation and dissemination while neglecting agenda setting and research governance through funding and policy. Finally, they have tended to explore the logistics of partnership working, rather than the more theoretical and political issue of how participation in partnerships interacts with different ideas about knowledge or evidence and the implications for expertise, capacity strengthening and impact: what types of assumptions about evidence frame partnerships and where do they come from? Whose knowledge counts in these framings and how is it represented? How can diverse sources of evidence be communicated, negotiated and usefully combined? How can different configurations of participation help to generate new types of evidence? And what is the relationship between evidence and genuine impact?
This article responds to these questions and to the limitations of the existing literature by proposing an original contribution to the theorization of research partnerships for international development. Drawing on the findings of a seminar series, 2 it argues that understanding the relationship between evidence and participation is paramount for understanding and improving partnerships. Over a 2-year collaborative process, the seminar series developed a participatory methodology to analyse seven case studies of research partnerships and an iterative framework to capture the relationship between the politics of evidence and the distribution of participations. This article describes the approach and explores the series findings.
Despite the diversity in scale and the range of approaches employed across the case studies, a common observation was the importance of understanding the contexts which framed the partnerships (as well as the various contexts in which the research itself was implemented and communicated). Each of the case studies involved a central partnership between a UK-based HEI and a UK-based INGO. This enabled an in-depth analysis of the British research funding and policy landscape and its role in shaping partnershipsbut also introduces a caveat around the generalizability of the findings from the study. Partnerships emerging in or across other national contexts or through transnational consortia may well encounter different challenges to those identified here (although the final event in the series sought to partly mitigate this by integrating perspectives from an additional 15 national contexts). However, by analysing the influence of British policy and funding on the seven case studies, this article highlights the importance of understanding specific contexts of research agenda setting and governance as well as the research processes and products themselves. We therefore start by situating the case studies within a broader analysis of the British context 2014-2017.

RETHINKING RESEARCH PARTNERSHIPS: THE BRITISH CONTEXT
Recent events in the UK have amplified the crises of identity affecting both the higher education and INGO sectors. Universities are responding to a national 'impact agenda' (Higher Education Funding Council for England (HEFCE), 2011(HEFCE), , 2017Nurse, 2015), which calls for the evidencing of applied uses of university research (Brewer, 2013; 2 Project and funding details omitted for double-blind reviewing.

Rethinking Research Partnerships
communication across these groups was facilitated through use of a range of participatory methodologies (including the use of analytical visualizations, embodied exercises and artsbased methods). These helped to disrupt established ways of working, structure analysis and debate, encourage active participation and learning and recognize and confront power dynamics. Table 1 provides an anoymized overview of the seven case studies highlighting (i) the leading organization (budget holder, lead applicant or principal investigator); (ii) the additional partners involved; (iii) the source, level and duration of funding; (iv) the formality, scale and geographical scope; (v) the disciplines or themes/sectors involved; and (vi) the research approaches implemented.
These case studies were all examples of broadly 'successful' research partnerships: sustaining productive relationships beyond the duration of the research; generating trust (to enable critical reflection); and ultimately producing evidence that was useful to both partners as well as contributing to a broader common goal of social justice and poverty alleviation. Crucially however, the series also drew on the additional experiences of participants to consider partnerships that had 'failed' and the rich potential for learning made possible by recognition and discussion of failure.

CONCEPTUALIZING THE POLITICS OF EVIDENCE AND PARTICIPATION IN RESEARCH PARTNERSHIPS: AN ITERATIVE FRAMEWORK
The seminar series collaboratively developed two iterative conceptual tools to analyse the politics of evidence and participation cross the seven partnership case studies. The first of these identified the specific factors that shape evidence and distribute participation while the second provided a heuristic framework for mapping these factors on to the research partnership process. We discuss each in turn.

Conceptualizing the Relationship between Evidence and Participation
The rationale for the series was informed by two rich bodies of research which explore the politics of participation in research partnerships for international development (e.g. Roper, 2002;Clarke & Oswald, 2010;Aniekwe et al., 2012;Hanley & Vogel, 2012;Holland, 2013); and the politics of evidence in international development (e.g. Cornish & Gillespie, 2009;Shutt, 2009;Beardon & Newman, 2011;Newman, Fisher, & Shaxson, 2012;Hayman & Bartlett, 2013;Eyben, Guijit, Roche, & Shutt, 2015;Hayman et al., 2016;Shucksmith, 2016;Chambers, 2017). This literature encompasses a wide range of thematic and professional knowledge, disciplinary traditions, methodological approaches and, crucially, theories of power. Across these, a number of different analytical lenses might be identified, each of which contributes in a unique way to understanding how evidence is shaped around, within and through partnerships with implications for how participation is distributed. These lenses can be sorted into five distinct (but interrelated) levels.
First, at the level of research institutions, studies have drawn on political science to explore structural influences, including shifts in how knowledge is valued within and produced through INGOs (e.g. Shutt, 2009;Beardon & Newman, 2011;Yanacopulos, 2014) and also the changing notions of 'evidence', 'impact' and 'knowledge transfer' in  Georgalakis et al., 2017). Institutional factors incorporate policies and process and might include elements such as organizational strategy, funding systems, processes for career development and infrastructure (e.g. access to repositories). Second, at the level of research practices, there has been a long history of NGOgenerated tools for communication (e.g. Archer & Newman, 2003) that builds on traditions of participatory/action research, while the academic fields of researcher development and academic practice have inspired research into the new literacies required for collaborative research (e.g. Facer & Enright, 2016;Lea & Street, 2006). Related studies have drawn attention to the politics of NGO knowledge engagement practices (Hayman et al., 2016) and explored strategies for strengthening capacities of a diversity of actors (e.g. Clarke & Oswald, 2010).

Rethinking Research Partnerships
Third, at the level of research identities, accounts of the subjective experience of research tend to draw on psychosocial traditions to produce embodied understandings of partnerships based on personal narratives, biographies and emotions (Griffin, Kalman, & Branstrom Ohman, 2012). However, the series recognized that identities are rarely fixed but also evolve through partnerships in relation to social practices, institutional discourses and the material environment. This resonates with a related body of literature around affect (e.g. Gregg & Seigworth, 2010), which highlights emotional, embodied and interrelational practices of identity building.
Fourth, at the level of research artefacts, research into the 'politics of method' has shown how assumptions about what counts as evidence are embedded in specific tools and templates such as logframes, RCTs and systematic reviews (Cornish & Gillespie, 2009;Eyben et al., 2015). Others have looked at how the material make-up of different research texts and technologies also influences the types of evidence that can be represented (e.g. Fransman, 2014). The rise of digital technologies has combined with a drive towards more available and accessible resources, fuelling a call for 'open development' (Smith & Reilly, 2013). At the same time, digital resources offer new possibilities for representing research through image, audio/video and written language.
Fifth, at the level of research accounts, this final analytical lens adopts a more 'meta' philosophical stance to show how conceptualizations of 'evidence', 'partnership' and 'development' are constructed. Accounts relate to the stories we (as individuals and institutions) tell about our research partnerships. These interact with other narratives or discourses (e.g. those around 'impact', 'co-production' or 'post-truth politics') and are commonly taken for granted although as many have shown, they are produced in and sustained by particular socialcultural contexts of practice (e.g. Sen & Ghani, 2013;De Sousa Santos, 2014;Hall & Tandon, 2017).
Together, these lenses constituted a broad theory of power for the series, serving as prompts to explain the wide range of factors that influence the relationship between evidence and participation in the case studies. To explore these factors in the context of the case study partnerships, we developed a heuristic framework conceptualizing the research partnership process.

Conceptualizing the Research Partnership Process
The series took as a starting point the assumption that research partnerships are not discrete, static entities but emerge from particular contexts with particular histories and

Rethinking Research Partnerships
interests and in interaction with a whole host of political, social, cultural and material processes, practices and products. In order to understand how they work, it is therefore crucial to understand the broader ecosystem of policy and funding mechanisms both for development and research as well as the processes through which partnerships are established, research is designed, implemented and communicated and evidence is accessed, taken up, adapted and ultimately used. To reconcile these elements, a heuristic framework was developed with four interrelated dimensions. First, evidence contexts: understanding the evidence ecosystem(s) in which the research partnership is situated in time/space, through social relations/histories and material environments. Second, research purposes: understanding rationales for different kinds of research and the agendas, interests and framings of the partner institutions/individuals. Third, power within partnerships: understanding ways in which evidence norms are shaped and participation is distributed in partnerships through institutions, practices, identities, artefacts or accounts. And fourth, change through partnerships: understanding the transformative potential of partnerships and determining what changes as a result of the research process. This cycle is set out in Figure 1.

Contexts of Research Partnerships
All research partnerships unfold in specific times and places, and these contexts have geopolitical, historical and material dimensions. The case studies involved organizations, individuals and research subjects from countries including Kenya, Malawi, Mozambique, Nepal, Nigeria, the Philippines, Swaziland, Tanzania and the UK (England and Scotland). Although the primary partners in all the cases were UK-based INGOs and

J. Fransman and K. Newman
HEIs, additional partners included local (national/regional) offices of the INGO; implementing partners (e.g. civil society organizations); communities, governmental bodies and national research institutions; consultants, volunteers and of course research funders-all of whom played different roles at different stages of the research. While most of these networks were transnational-involving participants from two or more countries -one of the case studies was based entirely in the UK with the researchers (INGO and HEI), mediators (e.g. Community-Based Organisation (CBO), government and media) and beneficiaries (e.g. communities) all located in the same region and policy environment. Unsurprisingly, this proximity and shared political context had a positive impact on participation, ownership and learning.
As Table 1 illustrates, the partnerships also varied in terms of scale and duration as well as in their thematic foci and disciplinary underpinnings. Each of the fields of study in which partnerships were grounded has its own history with implications for the types of evidence seen as legitimate. For example, one of the case studies operating within the field of 'sexual and reproductive health' revealed that for an extended period, due to the influence of an ultra-conservative US-Vatican alliance, the research on family planning was disparate and uncoordinated. However, in 2009, the Gates Institute for Population and Reproductive Health took a decision to fund an international conference on family planning, which then become a biennial event, facilitating a merger of NGO and academic research and creating space for joint exploration of more 'controversial' areas. Another case study described how increased attention to bio-medical interventions that perform well according to the criteria of 'evidence-based policy' was placing more pressure on NGOs 'to compete' in a bio-medical arena where only evidence featured in prestigious journals, such as The Lancet, will be considered. Table 2. Influence of evidence contexts on the relationship between evidence and participation in partnerships Institutions Partnerships are situated within complex institutional structures, spanning multiple geographical locations and often incorporating other stakeholder organizations in other locations. The relationship between these different spaces (e.g. national and regional offices) is distributed by organizational structures and processes and also by other factors such as funding conditions (determining spatial conditions for participation) and, more broadly, by geopolitics (e.g. formal classifications, e.g. by UN region or OECD Development Assistance Committee-list). Practices The types of approach and relationships that are valued in partnerships are determined by factors including research cultures, intellectual histories (of disciplines, subjects and fields) and ways of working within different sectors of practice. This includes timeframes, both in terms of the duration of funded projects and the pace and rhythms of working.

Identities
Personal research biographies are informed by different contexts through formal roles and job descriptions and informal experiences and sense of self. Different subjectivities are granted different status in different cultural contexts, distributed, for example, along lines of age, gender, sexuality, ethnicity, language, nationality, (dis)ability and socio-economic background. Artefacts Material environments (including the layout and content of buildings and virtual spaces) as well as objects (such as texts and technologies) and infrastructure (including databases and virtual systems) can have a symbolic and instrumental influence on what counts as valid evidence and who can access and contribute to existing sources of evidence. Accounts The idea of what constitutes a valid research partnership is rooted in social discourses that evolve through and across institutional and disciplinary narratives in interaction with social practices and global events. These include trends around, for example, 'impact', 'public engagement', 'official development assistance-compliance' and 'capacity building' and contribute to the ways in which partnerships are configured, the value given to different types of evidence in the partnership and, of course, the purpose of the partnership.
Finally, context is channelled not only through geopolitics and social discourse but also through material environments, which make up the spaces and places of the partnerships. Our seminars took place in a mixture of HEIs, INGOs and public buildings, and each of these locations (as well as the layout of chairs and tables within them) contributed to the (in)formality of the tone of the events and assumptions about the types of knowledge we were interacting with and generating. This was also observed in the case studies themselves. For example, one of the partnerships had been struggling to resolve a series of tensions related to the perceived status of the partners along the lines of age, job title, gender and nationality. This was exacerbated by the location of the face-to-face interactions, which privileged the position of the host institution. However, when some of the partners met up later at an international conference in a 'neutral' venue, they found their relations were freer and that their values and perspectives were far closer than they had previously thought. Table 2 draws on the iterative framework to illustrate how context is mediated by institutions, practice, identities, artefacts and accounts to influence what counts as valid evidence with implications for participation in research partnerships.

Purposes of Research Partnerships
Purposes for research emerge from evidence contexts. They might be set by specific funding calls (in turn dictated by policy agendas) or relate to the institutional agendas and/or personal interests of the different partners. The seven case studies unearthed a wide range of motivations, both for undertaking research and for initiating a partnership including documenting 'what is happening' in a programme or intervention (monitoring); providing evidence of 'what works' (summative or formative evaluation); discovering more about a topic or concept by engaging with existing evidence and/or generating new data (exploratory research); solving a specific practice-based problem (explanatory research); trialling or testing a new method, tool or approach (innovation/design-based/ developmental research); improving capacity of individuals, organizations or networks (capacity-building); and changing practice through collective critical reflection linked to action (transformative research).
Participants agreed that in reality, most partnerships integrated a number of different purposes. For instance, one of the INGO's desire to work with a particular HEI stemmed from their interest in developing skills around a particular approach in which the HEI had international expertise. They saw the research not only as an opportunity to strengthen their evidence base but also as a chance to practice different approaches to designing, implementing and reviewing programmes.
At the same time, motivations are not necessarily consistent across an organization. Within the INGOs, for example, there were marked differences in understandings of research between those working in a monitoring, evaluation and learning capacity on programmes (and therefore largely interested in improving their interventions) and those working in an advocacy capacity on policy campaigns (and therefore more interested in strengthening their arguments for greater influence). Within HEIs, there were also divergent priorities. Institutional agendas around securing impact, obtaining funding and scoring highly in the research excellence framework were often at odds with the interests and values of individuals. For instance, several of the academics considered themselves 'scholar-activists' and were passionate about contributing to practice-based knowledge J. Fransman and K. Newman and developing relationships with social movements. At the same time, motivations around job security and career progression were felt by those working in both sectorsalthough the pressures to develop an individual research profile and the likelihood of a precarious position were greater for the academics. This led to the motivation for partnerships frequently being grounded in individual interests on the part of the academic and organizational interests on the part of the INGO.
An example of how some of these motivations sit in tension came from an academic participant. Clarifying her own interests in 'advancing current thinking … ultimately in a way that would be publishable in a peer reviewed journal', she explained that 'to be publishable, we often look for a topic that is 'new' or a research design that is neat or striking in some way'. When an INGO colleague proposed researching a new health initiative that was not working as planned, she expected the study to tick all the boxes: ('a novel treatment … AND, things not going as expected: a perfect rationale for a qualitative study to investigate why people were not behaving as the guidelines presumed'). However, further investigation revealed that the NGO was actually becoming more successful in achieving uptake of the treatment ('which meant our framing in terms of a 'puzzle' that research could solve lost traction'). Ultimately, the academic took the decision not to continue with the research. However, the experience served to drive home an important point of learning: 'just as policy can idealize and simplify the field, so can research … The real world is messy and interventions evolveas they shouldin response to local needs, opportunities and constraints'.
Conversely, another INGO partner explained how its ultimate purpose (providing critical services) contrasted with that of its academic partner (improving health through the medium of research and its translation into policy/practice). This discrepancy led to some conflict around which institutional agenda were the driver for their joint research: 'Triggered by a difficult budget decision there was a rather painful discussion about whether "saving lives" or "quality academic research to inform how best to save those lives" should be paramount'. In this case, it was the INGO that held the purse strings and therefore had the final say.
While such negotiations can be uncomfortable and even lead to the discontinuation of a project or partnership, they can also create a dedicated space for understanding each other, building trust by confronting power and exposing agendas and assumptions. Strategies to disclose and negotiate difference included the use of formal artefacts such as ToRs, MoUs and 'Codes of Conduct'. However, difference across partnerships is not always negative provided that it is addressed with explicit attention to exclusions and hierarchies of knowledge. The participants therefore found it helpful to distinguish between 'productive' and 'problematic' tensions. While productive tensions may not be resolvable, they can often set conditions for creative emergence and trigger critical thinking and action.
Finally, the case studies also revealed that research purposes and motivations for partnerships are not necessarily static but can evolve over time. Examples of this included partnerships adapting to encompass teaching and learning as well as research, INGOs developing their own research portfolios and individuals embarking on further study. As one partnerships agreed, 'sometimes you do not know what is of value until you see it first-hand. This brought about surprises over how much one particular output ended up being applied and the wider impact it was having. Over time we are better at understanding what is of value for both partners'. Table 3 illustrates how 'research purpose' is mediated by institutions, practice, identities, artefacts and accounts.

Power in Research Partnerships
The allocation of leadership in a partnership is often dictated by funding. Power can be connected to control of funds with relationships distributed-explicitly or implicitly-by the specifics of funding calls. For instance, in one of the case studies although funding was sought for a collaborative project, the role of 'principal investigator' was restricted to the academic partner, and caps were placed to restrict the expenditure on non-academic and non-UK-based institutions to a third of the total budget. This had implications for the allocation of roles and resources as well as a democratic sense of ownership across the partners. The case studies also tended to confirm that the official budget holder was usually the primary decision-maker within the partnership. Interestingly, this position was occupied by a relatively even spread of INGOs and HEIs across the cases. In three examples, the ongoing 'partnership relationship' involved different sources of funding formally held by first one and then the other partner-with some interesting implications for the reprioritization of agendas.
However, control of funds was not the only indicator of power in the partnerships. Ideas about 'capacity' and 'research expertise' contributed in many of the cases to allocation of research design and decision-making primarily to the academic partner. Together, the seven case study partnerships employed a wide range of research practices, including more formal academic methodologies such as survey, systematic review, longitudinal studies and archival/documentary research, and also more practice-based approaches including action research, participatory mapping, participant observation and the use of discussion forums. Table 3. Influence of research purpose on the relationship between evidence and participation in partnerships Institutions All institutions have their own agendas whether ideological or strategic. These interact with sectorwider interests such as the 'impact agenda' in the HE sector or DFID's 'results-based' funding, which can be drivers for partnerships and can also influence the generation of particular types of evidence and contribute to particular configurations of participation (e.g. casting academics/ practitioners in specific roles or privileging certain methods of analysis or types of output). Practices Different (methodological and theoretical) research approaches lend themselves better to different research purposes, for example, participatory processes are well suited to context-based learning where as more quantitative approaches are better for predicative modelling at a more macro scale. However, innovative models of participation that bring together different knowledge can also contribute to new understandings and frameworks, generating new research agendas and interests. Identities Personal motivations can be linked to job security/progression, professional integrity or recognition, ideological agendas or emotional factors such as creative excitement, intellectual curiosity and political passion (as well as more negative emotions such as pressure, insecurity and competitiveness.) These motivations frequently evolve over the course of the partnerships as new research roles and identities emerge. Artefacts Different types of research artefacts privilege different research purposes which implications for what is considered to be valid evidence. For example, peer-reviewed publications respond to particular interpretations of 'academic impact' within different subject areas while tools such as logframes support particular types of data. At the same time, certain artefacts can be used to try to recognize and/or bridge different motivations and agendas (e.g. MoUs/TORs). Accounts Narratives such as those around 'impact' and 'best practice' influence understandings of research partnerships and therefore incentives for different actors to enter into partnerships. They also influence the way partnerships represent themselves, for example, through the tendency to demonstrate instrumental success (as opposed to relationship-building or critical learning from failure).
DFID, Department for International Development; HE, Higher Education.

J. Fransman and K. Newman
This discrepancy between the formal roles allocated to institutions and the actual practices of (and relationships between) individuals was a recurring theme across the case studies. While the strength of individual relationships was paramount to the success of partnerships, individuals inhabited different identities and levels of status with access to different types of power within their organizations. For instance, in a number of the case studies, the academic partner was relatively unsupported, which caused problems around navigating bureaucratic obstacles and limited administrative assistance. This was despite the exceptionally high overheads for academic partners, which lead some of the INGOs to question whether alternative partnerships with consultants might be preferable. In many of the partnerships, the research was therefore dependent on the INGO partner as an institution and the HEI partner as an individual. Staff turnover and restructuring in the INGOs were generally manageable (although changes in individual relationships had both positive and negative affects). However, when academics from three of the partnerships took maternity leave, the research struggled to continue without them, because of the lack of institutional ownership of the research.
Linked to the tensions between individuals and their institutions is the related issue of status-based hierarchies and assumptions about what that means for expertise and leadership within partnerships. This emerged in one of the case studies as tensions between a team of female early career researchers from the UK and a team of senior male professors from a university in Africa. Although the tensions were initially conceived around conflicting conceptualizations, they soon manifested as implicit and explicit accusations: ' "You are being imperialist;" "No, you are being patriarchal!" '. However, a similar distribution of participants in another project country-with a longer history of engagement and where 'relationships had been given the time and space to breathe'played out quite differently with 'difference forming the basis for some exciting and indepth discussions'.
Once again, trust is key as well as creating spaces for listening, reflecting and imagining. The participants observed that this is also crucial for taking full advantage of 'productive tensions'. But even where good relationships exist, there are challenges, as one academic partner observed: 'we are so anxious to be friendly and mutually supportive. Critique can be difficult in such situations, even if it is valued. I was invited to offer critical feedback to the INGO, but tended to critique the actions of other stakeholders, and be positive about their own work. Yet, I have a niggling feeling that the aid system of which the INGO is a part is a major part of the problem'. Such a position is hard to communicate as it calls into question the very raison d'être of organizations; however, a genuine commitment to understanding power and knowledge necessitates these uncomfortable questions being raised.
This type of tension was frequently carried over into artefacts such as research outputs. While some of the academic partners experienced the INGO as a 'red pen censor', other INGOs responded that academics were concerned with advancing their individual profiles and did not appreciate the complexities of institutional risk. Some of the partners worked around this by framing certain outputs as 'academic arguments' and publishing them not as an official partnership resource but 'under the "auspices" of the partnership'. Another issue relates to the 'affordances' of different types of research outputs. Findings that are frozen in time through publication can create challenges for an approach that is emergent and grounded in learning and critical reflection. Participants therefore distinguished between dialogical outputs such as blogs, events and workshops and more-authoritative-but-lessdynamic-and-interactive publications such as peer-reviewed journal articles, policy briefs

Rethinking Research Partnerships
and press releases. Participants agreed that these choices are both practical and political. They are also informed by the schedules of the partner organizations as well as by the timings of national or international events, which might provide opportunistic spaces for dissemination. Such considerations often involve tensions based on conflicting timetables and the pace and rhythms of working. For example, the pressure for one INGO partner to communicate findings before the analysis was complete raised concerns for the academic partner for fear of broadcasting premature messages and compromising future academic publications (on which careers depended).
Finally, participants discussed the influence of accounts (or the way partnerships are represented) on the dynamics of participation and evidence mobilization. All case studies were conscious of the additional partners omitted from the process of representing the case studies in the series itself-many of whom were based in the Global South. While two of the case studies integrated comments and video interviews from some of these partners into their presentations, participants agreed it was important to acknowledge partiality. Table 4. Mediation of the relationship between evidence and participation in partnerships

Institutions
Power is partly distributed not only through institutional control of the partnership (e.g. to the budget holder and/or coordinator and/or principle investigator of the project) but also through the conditions of the funding (e.g. budgetary limitations for certain categories of partner) and the focus of the scheme which might prioritize research and also other aims such as intervention, product development, engagement and/or learning. Practices What counts as valid expertise is distributed in partnerships in different ways by different research approaches. Some of these require specialist training around data collection, analysis and communication. Others facilitate collaborative sense-making between different groups and knowledge. Still, others incorporate an explicit learning dimension so that part of the research process involves partners learning from each other and developing new skills. Participatory processes can unearth new types of evidence and enhance learning but are time consuming. Conversely, more rigid, top-down processes can face challenges in communication (including interpretation of research concepts, methods and findings) and buy-in to the process.

Identities
Institutional roles (e.g. titles, job descriptions and expert profiles), identity markers (e.g. age, gender, sexuality, ethnicity, language, nationality, (dis)ability and socio-economic background) and subjective sense of self as a researcher all contribute to how relationships are formed, trust is built and common understandings and ways of working are negotiated in partnerships. While friendship and emotional ties can sustain partnerships and facilitate collective reflection, they can also make some types of critique uncomfortable, especially when personal values are at odds with institutional agendas. Artefacts Research outputs can be dialogical (facilitating ongoing interaction and evolving over time, e.g. through seminar series, events or blogs) or fixed, immutable resources (e.g. publications) with more authoritative and symbolic power but less interactive potential. The ways in which research is represented also affects accessibility of the research and its potential to influence different audiences (e.g. with statistics often carrying more weight in policy and personal stories affecting publics), as do dissemination strategies linked to accessibility, affordability and adaptability of research. Tensions related to the schedules and pace of the different partners also exist around the timing of dissemination. Accounts Given the myriad of actors involved in partnerships, there are always exclusions in representations of the partnership which privilege authors of outputs and those responsible for presenting findings. While collaboration about the development of outputs can mitigate this to some extent, the pressures of time, lack of funding to cover dissemination and inaccessible reporting systems that privilege certain types of outputs (e.g. the Research Council UK's 'Researchfish' system) contribute to particular partial accounts. Representations of partnerships are also coloured by ideas about what counts as legitimate research processes, outputs and impacts with certain people, activities and products excluded as a result. Table 4 sets out the influence of institutions, practice, identities, artefacts and accounts on the workings of power within partnerships.

Change through Partnerships
All of the case study partnerships were brought together by a shared commitment to actionable research that responds to the many facets of poverty and ultimately aims to inform more just and equitable societies. Many of the partnerships were also committed to using research to give voice to excluded or disaffected groups; to 'research as challenge'-incorporating alternative knowledge into established debates. This included, for example, evidence from low paid workers to understand the notion of 'decent work' and from excluded communities to inform policy debates around austerity, health and welfare reform.
However, this broad vision of affecting social justice included different and sometimes conflicting agendas for change with many of the INGO participants seeing research as a catalyst for change and many academic participants seeing research as a means of exploring the types of change that are needed. In some of the case studies, this meant that INGOs tended to view change as about working with and within their existing processes and structures, while academics tended to view change as a critical engagement with these processes and structures. 3 For example, in one case study, the INGO had engaged in research around supporting housing cooperatives to manage savings funds for acquisition of private housing. However, the academic partner determined that this approach was accessible to only a small proportion of households, and likely to exacerbate processes of housing inflation and economic displacement in communities where the vast majority are renters. Such 'productive tensions', however, can also have a positive effect, with INGOs challenged to think more critically about their structural change processes, and academics challenged to be more pragmatic and link critique to viable action.
Methodologically, the different understandings of change led to two distinct approaches: the theory of change, which involves determining the desired impact from the offset; and the research question/hypothesis, which involves establishing a field for investigation. Academic artefacts, such as 'Pathways to Impact', are attempting to bridge this divide, and likewise in the INGO sector, some funders are starting to support 'adaptive management' and 'strategy testing' although the results-based approach still dominates. At the same time, many of the participants felt that in order for a proposal to be accepted, the likely impacts had to be emphasized to the extent to which it shut down flexibility and room for emerging outcomes. This was coupled with the emphasis placed by funders on innovation and new or even controversial findings. In contrast, some of the participants argued that there is also value in more mundane evidence. One of the cases, for example, 'set out quite consciously to provide rigorous evidence to prove what anyone who knew anything about the matter would already have suspectedthat low paid workers are concerned about wages, job security, conditions, health and safety, and a having a good line manager'. The aim of this type of research 3 Although this critical approach is arguably a bias of social science research.

Rethinking Research Partnerships
was to strategically project the evidence into the public domain at an important moment in the development of policy.
As several of the case studies found, timing of research outputs is critical to influence policy, as is designing the right types of output for the right audience. One partnership explained how they made use of government consultations and inquiries to contribute to parliamentary motions. Another invited policymakers to facilitate workshops around the findings in order to mobilize their ownership of the research and create a space to explore how it might contribute to policy.
As well as influencing change in policy and practice, many of the case study partnerships also contributed to changes in perspective. For example, by supporting community-based research, one partnership identified a change in the community's sense of self-both as providers of legitimate knowledge and as researchers in their own right. Others generated critiques of influential policy concepts such as 'capacity building' and 'social capital', although participants insisted that such conceptual work be linked to changes in practice.
While all the case studies sought impact in the world beyond the partnerships, change was also identified within the partnerships. This included changes in individuals through the development of new skills, languages, sensibilities and confidence and through changes in role or occupation. For example, in just one of the case studies, an INGO project manager and volunteer researcher both embarked on PhDs while three other participants took on new roles as INGO researchers, and a fourth continued the research in her own country context.
Change also occurred at the institutional level. For example, one INGO integrated the participatory research tools into their monitoring, evaluation and learning systems and drew on learning from the partnership process to develop more adaptive programming. For another INGO, the development of a theory of change (as an unplanned by-product of the research) informed the redevelopment of their results framework used to monitor and evaluate the work of 2000 NGOs around the world.
In many of the case studies, the partnerships also engaged critically with tools, technologies and texts, adapting existing artefacts and generating new ones. In one of the partnerships, a systematic review was chosen as an approach favoured by policymakers. However, when the findings proved largely 'inconclusive' and therefore of little use to the practitioners, the partners decided to conduct a 'scoping review' to explore contexts of implementation, examples of good practice and lessons learned. The academic partner explained that 'while the 'scoping review' was more useful it was not publishable, because it was a rather unsystematic 'pick and mix' of interesting studies. But the 'systematic review' was publishable, because it followed all the strict guidelines even if those guidelines did not result in a very useful answeran example of the methodological tail wagging the dog!' A solution was finally reached through the development of a paper that presented both the review findings and a critique of systematic review as an approach.
Finally, the series participants considered how their partnership experiences had influenced the ways in which they understood 'research partnerships' and the language they used to describe them. As one participant observed, 'we often call it a partnership when we start working together, but actually it is the process of working together that creates a partnership'. Table 5 illustrates how change at the level of institutions, practice, identities, artefacts and accounts can be effected by partnerships.

CONCLUSIONS: IMPLICATIONS FOR UNDERSTANDING AND IMPROVING PARTNERSHIPS
Across the different dimensions of the framework addressed earlier, a number of themes can be identified.
First, despite the assumption that evidence is a commodity that can be unearthed and channelled into policy and practice, our analysis has shown that evidence is never neutral. What counts as legitimate evidence is shaped by the institutional contexts that frame partnerships; funding protocols; systems of career progression; research approaches, tools and infrastructures; and language use. This does not mean that all evidence is of equal worth or quality. All research should be evaluated against its aims and objectives and within the standards of its organizing frameworks. For academics, these might include quality control mechanisms such as peer review, while for INGOs, they might relate to theories of change. What is vital is that research is explicit about its purpose and approach and that the design and implementation of research are internally coherent and fit for purpose. Partnerships can influence change in institutional structures, processes and agendas (e.g. internally, by channelling learning from the partnership process back into the partnership institutions, or externally, by generating evidence to inform wider policy). However, institutional agendas also influence the types of change prioritized in partnerships (e.g. academic impact through specific contributions to academic knowledge in the HE sector versus results to support or inform existing practice in the INGO sector) results/practice. Practices Partnerships can engage critically with existing research practices and contribute to new practices (e.g. concepts, methods, languages and ways of working) of the partners themselves (through critical reflection and learning from each other); their broader institutions (by using learning to inform culture change and researcher development); and beyond (e.g. contributing to changes in sector-wide practices or in disciplines or fields of inquiry Partnerships can contribute to the development of new ontological products including actual outputs and changes to the perceived legitimacy of different types of outputs. At the same time, different artefacts (e.g. 'theories of change' and 'pathways to impact') have different implications for how change is understood and planned. While there has been pressure from both HEI and INGO sectors to map out anticipated changes from the start of projects, the case studies showed that some flexibility allows for these transformative goals to evolve in response to evolving understandings and for unanticipated changes to be recognized and built on. Accounts Working in partnership can contribute to new understandings of research partnerships in terms of their potential and also the challenges they face. This might also involve critique of the language of partnerships (which implies equity, symmetry and discrete entities when the reality is far messier, unbalanced and more political). Finally, the case studies showed that the meaning of partnership is often emergent-unfolding through the process of partnership as opposed to providing a blueprint for research relationships and processes.

Rethinking Research Partnerships
Second, the ways that evidence is valued in partnerships have implications for participation, determining whose expertise counts and to what extent. This can have a detrimental effect on research that aims to be actionable-excluding or misrepresenting the voices of those deemed to be less qualified or skilled. However, the series has shown that hierarchies of knowledge might be unsettled or renegotiated: by incorporating civil society into agenda setting and governance systems; by disrupting norms; and by building movements, which extend beyond the boundaries of institutions and are based on shared values and agendas.
Third, research partnerships are complex and emergent. Research is seldom a straightforward linear process. The messiness of personal circumstance, social relations, conflicting agendas, institutional restructuring and unanticipated geopolitical events is compounded when the goals of producing knowledge and informing action are brought together in a collaborative process. This grows messier still when partnerships involve multiple institutions spanning diverse contexts. The most successful partnerships tend to be those able to accommodate uncertainty. This might involve the design of flexible approaches and clear communication strategies. In this way, 'partnership' might be understood as a verb rather than a noun; an emergent assemblage of people, things and ideas as well as institutions.
Fourth, many funding schemes incentivise the rapid development of short-term projectbased partnerships. This can have some benefits (e.g. mobilizing new networks and focusing attention on specific time-bound research challenges). However, our analysis suggests it may be helpful to understand partnerships as grounded in pre-existing (although shifting) networks and relationships that continue beyond the duration of the research. There are therefore benefits to investing in long-term collaborations, which are based on evolving but shared understandings, values and agendas and enable learning to be channelled back into institutions. This involves allocating sufficient time to develop and manage relationships, responsive context-analysis, collaborative (re)design, critical reflection and more formal learning/training and concluding research in a way that will maximize evidence-use and ensure sustained collaboration.
Fifth, the spatial dimension of partnerships should also be recognized. There is a tendency in social science research to consider only the empirical context of 'implementation' without adequately acknowledging the 'home' context of the research institution and funders. Another spatial consideration is the effect of scale. While there are undeniable complications to ensuring genuine participation in large transnational and multi-stakeholder partnerships, there is also potential for creating new collectively negotiated (and possibly virtual) spaces for research. Related to this, it is important to recognize that while all research partnerships are grounded in specific contexts, they also have the potential to transform and create new contexts.
Sixth, research partnerships have the power to affect transformation. This may be internal to the partnership (through changes in practices/knowledge/skills; development of relationships and resources or infrastructure); external in the short term (through institutional policy change within partnering organizations, establishment of networks and specific resources/infrastructures external to the partnership); and external in the longer term (through the rocky road to impact from access of outputs to uptake, adaptation and use) and through changes in discourse around the meaning of evidence and partnerships. This links to the challenge of attribution and the implication of perverse incentives to attribute change to a single project, institution or individual rather than to broader networks and/or movements.
Seventh, learning is key for productive partnerships. As well as developing traditional academic skills/knowledge, the series also identified other types of 'research engagement literacies', which might support aspects of partnership development from commissioning J. Fransman and K. Newman research and understanding context to communication, management, brokering and networking. Participatory approaches also provide informal training, allowing diverse knowledge to be brought together and partners to learn from each other. Another key consideration was how to capture learning and channel it back into institutions and the wider sector. To support this, research 'brokers' were seen as key. These actors commonly have experience with different institutional cultures and often play a training role across multiple institutions. Finally, learning from failure-and therefore being willing to acknowledge and discuss failed partnerships and research-was seen as a rich but largely untapped opportunity.
Eighth, the representation of research is a crucial consideration, and the series addressed the significant pressure on academics to produce peer-reviewed publications, against recognition that these types of text are not as timely, accessible or useful as other outputs. Conversely, documents produced by INGOs, such as policy briefs, while accessible are sometimes seen as less credible with advocacy agendas undermining the rigour of content. This contributes to an argument for more co-authored research, provided both partners are involved in and recognized for outputs. The choice of text is important, too. The format of academic journals, for example, tends to privilege a particular type of evidence, usually communicated although written language (most commonly in English). However, advances in digital media are increasingly enabling new types of representation in traditional spaces (e.g. multimedia embedded in online journals). Alternative spaces/systems of evidence (e.g. exhibitions, films and performance) are also increasingly recognized.
Ninth, good communication was a recurring theme in the series, and central to this is the importance of interpersonal relations and recognition of emotion. The series unearthed a range of positive and negative subjective experiences (including feelings of empowerment, friendship, sense of belonging, excitement, alienation, anxiety, jealously, guilt and exploitation). Articulating the harder-to-name or more uncomfortable emotions can be a helpful indicator of the deeper power relations embedded in processes and practices.
And finally, the series called into question some of our language practices with participants questioning the value of the term 'partnership'. Drawing on our analysis of partnerships as emergent networks rather than bounded projects, do we need a new terminology and to what extent are terms such as 'engagement' more helpful in accounting for our work in research collaborations?
Together, these lessons inform the ways that we understand and implement research partnerships. They imply six shifts that might improve our thought and practice: (i) moving beyond individuals and institutions (embracing relationships, networks and movements); (ii) moving beyond instrumentalism (embracing criticality and affect); (iii) moving beyond linear, short-term projects (embracing long-term agendas, complexity and flexibility); (iv) moving beyond participation in the production and communication of research (embracing participation in agenda setting, research governance and evidence use); (v) moving beyond traditional written outputs (embracing alternative modes of representation); and (vi) moving beyond notions of 'success' (embracing learning and unknowing and destigmatizing failure). These movements resonate with a more systems-based, adaptive, emergent and learning-oriented approach to development (Burns & Worsley, 2015;Chambers, 2017) with different implications for INGOs, HEIs and research funders. 4 Looking forward, these institutions must work together to draw on this learning and ensure that research collaborations produce better evidence that is relevant and actionable in the fight to eradicate poverty and achieve social justice.