Why do we know what we know about development? Knowledge production in Canadian academic-civil society research partnerships

ABSTRACT This article examines how power operates through processes of knowledge production in the Canadian political economy. I draw from the sociology of knowledge to explore four dimensions of power. The first considers implicit power through values and norms; the second, control of means and resources; the third, the exercise of power in social relations; and the fourth, “power as capacity” – that is, how participants engage with research partnerships differently from other forms of development practice. My case studies illustrate the variety of ways in which research partnerships in Canada both reproduce and challenge knowledge hierarchies domestically and globally.


Introduction
Canada as a development co-operation partner thrives on relationships and collaboration. Often, such relationships are fortified through networks, shared histories, and sound project management and governance. What distinguishes research partnerships from other kinds of relationships in the field, however, is the central role of knowledge production. Yet, the processes of knowledge production are intimately influenced by the power dynamics of those who control the means of research financing, agenda-setting and dissemination. These power dynamics are not separable from what in fact we come to know from these partnerships. Responses to questions about what we know about development, who knows it, how and why and for whose benefit, then, are all closely informed by the power relations that shape research partnerships and the practices and policies in Canadian development co-operation sector. 1 Research partnerships have been lauded as a promising means to improve the effectiveness and impact of development, for the opportunities they create to strengthen the connections between a growing evidence base and practice in the field (McGiffin 2017; Martel and Kindornay 2020). Nonetheless, it is impossible to consider the true potential of partnerships to have a substantial effect on Canadian development policy and practice without a critical assessment of knowledge as a form of political and economic power in Canada and globally. Research partnerships, especially those supported by government or private funding, can serve as instruments to reproduce certain kinds of knowledge hierarchies in the context of the Canadian political economy. It is the objective of this research, then, to contribute to a better understanding of the possibilities and limitations of research partnerships' contributions to improving the structures and functions of development processes in Canada.
In this article, I examine how power operates through processes of knowledge production in the Canadian political economy. I begin by reviewing literature on research partnerships in the context of a knowledge-power nexus, examining how global and domestic knowledge economies inform discourses, resources, social relations and the expression of capacity. I then describe the study's methodology, which draws on multiple case study research methods to examine the experiences of seven academic-civil society organisation (CSO) research partnerships in Canada. The findings are summarised along several key questions, including: what knowledge is valuable for academic partners and CSO partners? Which research methods are employed? How is knowledge produced and disseminated? Finally, who has access to collected data? In my analysis, I draw from Zingerli's (2010) sociology of knowledge framework, in which she examines three dimensions of power in research partnerships: first, implicit power through values, norms and discourses; second, control of the means and resources for knowledge production; and third, the exercise of power in shaping social relations. I introduce a fourth dimension to this framework, in exploring the exercise of "power as capacity" (Karlberg 2005;Giddens 1984;Arendt 1969) in how participants engage with research partnerships differently than the way development discourses, structures and relationships tend to operate otherwise. I observe that, in the cases represented in this study, there is a variation as to how the political economy of research partnerships in Canada either reproduces or challenges knowledge hierarchies domestically and internationally.

The problem of evidence and knowledge in development
In development research, policy and practice, evidence faces two central challenges. The first is that while research continues to grow, it does not always have a clear connection to development policy and practice in the field. Scholars, practitioners and policymakers are routinely constrained within their own cultures of operation and institutions that prevent them from engaging with evidence fully and effectively (Carden 2009). Andries Du Toit (2012) suggests that the discourse on evidence-based development policy overemphasises a technocratic and linear understanding of the policy making process. A naïvely empiricist view, he suggests, fails to recognise the significant disjuncture between evidence and policy, with critical consequences for development research and practice: If this disjuncture is not recognised, the implications can be damaging. Policymakers tend to ask academic researchers for clarity they can't provide; and donors may expect that researchers eschew "unnecessary complexity" and provide clear and unambiguous "take home advice". Academics who are eager to please policy "users" may be tempted to oversimplify their findings, or to overstate the certainty and scope of their conclusions in ways that their scholarship, strictly speaking, cannot justify. The result is a state of affairs that serves neither academic research nor good policymaking. (Du Toit 2012, 6) Canadian development practice also has a fraught connection with research and evidence. Tiessen and Smillie (2017) observe trends in Canada such as diminishing funding, a lack of governmental recognition of the partnership potential of universities, a decrease in development-focused research centres in universities and a lack of institutional incentives for academics for practical involvement in development. They argue that these are some of the most significant factors that curtail strong flows between evidence and practice.
The secondand far more foundationalchallenge lies in the question of the veracity and reliability of truth-claims, centring on crucial epistemological questions about what can even be counted as "knowledge." Hall and Tandon (2017) and Lepore, Hall and Tandon (forthcoming) draw from thinkers from the South to argue to argue that academic research recognises only certain kinds of knowledge systems. As such, "what is generally understood as knowledge in the universities of our world represents a very small proportion of the global treasury of knowledge" (Hall and Tandon 2017, 7). It is essentially that which has been inducted or appropriated into the Western scientific canon. To decolonise knowledge production, they suggest, collaborative research partnerships need to look at those knowledge systems that have been marginalised or excluded. In Indigenous communities, research programmes can be especially precarious given the colonial legacies and continued presence of extraction, exploitation and erasure (Tuck and Wayne Yang 2014;Tuhiwai Smith 2013). Kassam and Tettey (2003) argue for the importance of prioritising communities' own knowledge resurgence and revitalisation when undertaking research partnerships with Canadian Indigenous communities.
Even within a shared epistemological commitment, however, bias can shape what knowledge is prioritised or legitimised. Bias operates implicitly as a form of discursive power that organises social processes and relationships along a power-knowledge nexus (Foucault 1980). The Rethinking Research Collaborative (RCC) (2018) in the United Kingdom analyses the tendency to organise knowledge into hierarchies, by which those who have greater control over the means of knowledge production have greater influence over what knowledge is sought out, how it is analysed and what is shared (Fransman et al., forthcoming). Zingerli's (2010) study on North-South development partnerships further illustrates that research partnerships are not an easy remedy for asymmetries in international development research. In fact, they can greatly exacerbate inequalities in international development, for example, by treating the global south as a laboratory for the North (Zingerli 2010, 222). Political and economic biases can also skew knowledge production, especially when research funding comes from sources with particular interests in the outcomes of research, such as governments or companies. From this vantage point, no evidence can be seen as neutral.

Power and discourse in knowledge production
The operation of power has many expressions and consequences. Political and cultural theorists have well documented the way that power can be wielded by actors to shape subjects' behaviours and thoughts through the circulation and control of ideas, capital and culture. Historically, the analysis of power has been equated with domination, through such influential thinkers as Max Weber (1986) in his work on domination by economic power and authority; Pierre Bourdieu's (1994) examination of structural reproduction of inequality through habitus and symbolic power; and of course, Marx and Engels (1967) analysis of class struggle and the control of labour and production. Among the most influential theorists of power, Foucault (1980) describes power as a relational force that permeates and circulates in society, connecting social groups in interconnected webs of influence by which hierarchies arise through the production of discourse, the imposition of discipline and order, and through shaping desires and subjectivities.
Michael Karlberg (2005) notes that while current analyses often continue to equate power with domination, important critiques have introduced a new perspective of power that he refers to as "power as capacity." He cites Anthony Giddens (1984) notion of power as representing the capacity to transform outcomes. Importantly for the analysis of collaborations, Hannah Arendt (1969) notes that "power corresponds to the human ability not just to act, but to act in concert" (43). Feminist theorists such as Jean Baker Miller (1982) and Nancy Harstock (1974) note the role of power in nurturing and empowering others, recognising the communal dimensions of power and the potential of power as capacity to produce change. Similarly, many systems theorists have outlined similar dimensions of power in the creation of interdependent and complex systems and sub-systems. Kenneth Boulding (1990) describes the notion of integrative power as "the capacity to build organisations, to create families and groups, to inspire loyalty, to bind people together, to develop legitimacy" (25).

Research partnerships in a global knowledge economy
In her study of North-South research partnerships in the global political economy, Claudia Zingerli (2010) describes a sociology of knowledge framework that employs three lenses through which power can be examined and expressed in research partnerships: first, power as operating through dominant values, norms, rules and discourses; second, power through the control of material resources, formal structures, institutions and procedures; and third, power operating through social relations to shape experiences and perceptions. In this article, I add a fourth dimension to Zingerli's framework: power to express and develop capacity in oneself and others. Insights from Giddens, Arendt, Miller, Harstock and Boulding show the transformative potential that partnerships hold as they behave in ways that the constraints of norms, structures and relationships could not otherwise predict. It is therefore important to include "power as capacity" as a mode of power's operation in its own right, in complement to the other lenses described above.
The twofold view of "power as control" and "power as capacity" takes into account the effects of both structure and agency in influencing research partnerships. For example, Natalia Yang's (2017) study on South Asian CSOs illustrates their motivation in engaging strategically in collaborations while at the same time being cognisant of the potential obstacles and forces that can prevent collaboration from being equitable and sustainable for all members. Canadian academic-CSO partnerships, in particular, represent this complexity, in which partnerships can be simultaneously a generative stimulus for transformative new directions in development research and practice, while also being a sign of systemic economic and political constraints. In his edited volume Putting Knowledge to Work: Collaborating, Learning and Influencing International Development, Luc J. A. Mougeot (2017) also highlights how increasing emphasis and interest in research collaborations in Canada are not only the results of their general benefits; they are becoming conspicuously propelled by a host of systemic pressures that include continual decreases in net Official Development Assistance (ODA), increased development skepticism and stringency for evidence-driven impacts, and prioritisation of market-driven economic development among other development needs. The questions of "best practices" and "what works" in academic-CSO research collaborations in Canada, then, are continually mediated by the political and economic landscape in which they are embedded at any given time.

Methodology
This study employed a multiple case study research design in which I, with the support of Andréanne Martel and Fraser Reilly-King at the Canadian Council for International Cooperation (CCIC), selected and developed seven cases of academic-CSO partnerships, and derived cross-case findings and conclusions (Yin 2013;Stake 1995). Canadian research collaborations that we identified for this research therefore share a number of characteristics in order to strengthen the ability to observe common dynamics and trends across multiple sites (in other words, literal replication). These characteristics and their rationales included: (1) collaborations that were current or had recently concluded (within the year); (2) partnerships involved at least one Canadian-based academic institution and one Canadian-based CSO; (3) partnerships drew on different sources of funding for the research collaboration, and that funding is not solely sourced from the partner organisations' own resources, unless done so for the specific purpose of the research partnership; and (4) at least two partnersone academic and one CSOagreed to participate in this study. To optimise for representation across the sector, we also sought to identify cases with some variations, including: different collaborative and partnership models; different geographic emphasis (for example, sub-Saharan Africa; South America) and sectoral focuses (for example, health, agriculture, education, human rights, business and so on); different academic programs and disciplines (for example, international development studies programs, literacy, nutrition, health, environment, management and so on); different funding sources (for example, Social Science and Humanities Research Council, International Development Research Council, Global Affairs Canada, Mitacs and others), different provinces and linguistic contexts (including cases in Québec, Western Canada, Ontario and other); and the inclusion of partnerships that involve Indigenous academics and CSO partners.
Taking a snowball sampling approach, potential participants were identified and recruited by key contacts at the participating organisations and universities. Announcements were also sent out to CCIC and Canadian Association for the Study of International Development (CASID) listservs and to subscribers of the Next Generation e-newsletter. Interested participants who met inclusion criteria were provided with an invitation letter and asked to confirm their participation consent in writing or verbally if contacted for an interview. Table 1 summarises the characteristics of the seven cases selected for inclusion in this study.
Participants in one-on-one or group interviews included academic researchers and principal investigators, executive directors of organisations and directors of research centres, research coordinators and project managers. Interviews were conducted in either of Canada's official languages -English or Frenchaccording to the participant's preference, and all interviews occurred between April and July of 2018. The purpose of one-on-one interviews was to collect information about the experiences and perspectives of individuals during their involvement in the project. Interviews employed a semi-structured protocol and included prompts to collect stories and reflections on the overall outcomes, successes and challenges of the partnership (Bryman 2004;Fontana and Frey 2000). I asked questions related to four areas: structure of the research cycle; knowledge and evidence; finance; and incentives, risks and motivations. Participants were also invited to add and speak to any other important factors that I did not ask about in the research protocol. I transcribed all interviews for analysis and wrote analytical memos and used the qualitative data analysis software Atlas.ti to code the data.

Findings
I asked participants several questions dedicated to knowledge and evidence, such as: what do you and your organisation understand by the terms "research" and "evidence"? What is the usefulness of this research for your sector and for your organisation? What do you understand is the usefulness of this research in your partners' sector or discipline? Where are research dissemination activities taking place? Who has access to the data generated from this research and where is it? Who uses this data now and who will benefit from it in the future? A summary of the seven case studies' responses to questions in section two of the interview protocol is found in Table 2.
What knowledge matters?
The first set of questions (columns 1 and 2 in Table 2) asked what kind of knowledge production is valuable to academic and CSO partners. Some partnerships demonstrated overlap in this category, such as in cases 3 and 7. In case 3, for example, the academic partner was not overly concerned with academic publications, but rather saw publication as a way to continue to build on her commitment to participatory research methods, as well as to connect a graduate student with good professional opportunities. In case 7, the participants shared profound commitments to Indigenous knowledge resurgence and its application to knowledge about climate change. These commitments transcended other priorities or outcomes in determining whether or not the partnership would proceed.
While most partnerships expressed some overlap in their knowledge interests, there were also significant differences. Academics placed far more emphasis on peer-reviewed academic publications in highly ranked journals than did their CSO counterparts. Likewise, scholars were motivated by the ways that partnering with CSOs could make their research findings more impactful and relevantmentioned in cases 1, 4 and 5emphasising empiricism and real-world relevance over theoretical critique. Some academics, such as those in cases 2 and 3 were interested in how the research partnerships provided avenues to socialise and train graduate students in impactful knowledge production. Academics in case 6 were interested in how the research could help validate methodological indicators and models. Academics in cases 4 and 7, on the other hand, were interested in how academically rigorous research could provide a launchpad for stimulating community knowledge and context-based knowledge generation in their Southern partners' communities and countries, beyond the organisations where they worked.
CSOs in cases 1, 2, 5 and 6 were interested in evidence that would evaluate or improve the quality and efficacy of their programming. In some of these cases, monitoring and evaluation were required by their funding institutions; thus, the idea to work with academics did not always initiate with the CSOs. In cases 3 and 4, CSOs were interested in generating new knowledge in their fields that could help position them as thoughtleaders domestically or internationally. In cases 3 and 7 CSOs were especially interested in catalysing knowledge generation and systematisation in the global South and in the Indigenous communities in which they were involved.
Which methods generate "good evidence"?
The second question (column 3 in Table 2) inquired into the kinds of methods employed in research partnerships to generate evidence and knowledge. Research methodologies, while often seen as falling within the purview of the academic partners, were also of 280 E. TOUKAN  vital concern to the activities of the development organisation whose project would be affected. Methodologies can align with or depart from the programmatic approach of the project's practitioners. Examples of these kinds of methodological tensions emerged most noticeably in cases 2, 4 and 6, which incorporated experimental methods and randomised control groups. Among their concerns, CSO members expressed anxieties about the ethical risks of creating situations among the beneficiary communities who may be unaware of the premise of their participation in a blinded randomised research trial, but cannot be prevented from discovering the experimental nature of the project while communicating with members of other communities. One participant expressed this point: For example, if you do a randomised group of beneficiaries, even if you are in a region where you are clustered, and you are trying to keep the intervention participants away from one another, it's tricky. The channels of communication are incredible and sometimes not clear to us. I have seen these types of designs create problems between communities, and sometimes people are allocated randomly to an intervention group but not others, and that generates issues that perhaps give us a stronger design for the conclusions that we are hypothesising but might even create some negative impacts in the people.
Similar concerns were voiced by several other participants, as well as points about the importance of maintaining bonds of trust between CSOs and community partners. However, the majority of participants found that these potential concerns were alleviated in most cases in the way that academics take on these concerns as their own and design the research cycle accordingly in a way that is acceptable to all partners. Successful partnerships in these cases often involved periods in which partners take time to gain a thorough understanding not only of the activities of the project, but the underlying values, assumptions and orientations of the project's approach. This did not always change the methodology chosen for the study but it helped academics take organisations' concerns into account in their research design to a greater degree. It was important to many partnersacademics and CSOs alikethat the resulting design would be mutually adopted by all partners as a shared standard of investigation, both for the quality of the study and its ethical implications.
In cases 1, 3 and 5, academic research methods involved literature reviews, case studies, interviews and surveys, and these methods were not of significant concern to CSO partners. Indeed; organisations found many of the findings generated by these methods to be easily translated to staff, donors and Southern partners. Partners in case 7 placed priority and emphasis on research methods that were community-based, culturally appropriate and non-exploitative to Indigenous peoples. Importantly, they critique the fact that academic research in their communities has historically been extractive. Instead, the partners emphasise a perspective of "research as relationship," involving trust, humility and openness to revising one's own assumptions. The academic partner notes that these conceptions of research and relationship do not exclude the use of big data and research tools, but rather imply that the relationship with the community is given priority among the researcher's responsibilities. Researchers pay close attention to community-specific protocols and the contextual nature of Indigenous knowledge. This has a reciprocal benefit as communities reach out to trusted researchers in collaboration and participation.

How are knowledge and evidence disseminated?
Data about research dissemination practices and outcomes were only partially available or not available for all cases, as some research partnerships were still engaged in the processes of data collection and analysis, such as cases 1, 4 and 6. Most research dissemination occurred first and foremost in the form of reports created by academic researchers for partners and funders on the initial findings of the research. After this, the responsibility for how and with whom the research would be disseminated fell on each respective partner, according to their needs and interests.
Case 1, the largest in scale and complexity among all the cases in this study, mobilised its research agenda to have as significant an impact on the field of international criminal justice. Knowledge mobilisation has been multifacetedacademic and practical, as well as public-facing through engagement with mass media platforms including radio interviews and newspaper editorials. Knowledge production and dissemination operate to some degree as a self-reinforcing ecosystem, in which research findings are shared through summer workshops and colloquia that give rise to further research initiatives, both within and beyond the scope of the research partnership.
Academic publications have emerged from collaborations in cases 1, 2 and 5, with further publications anticipated in all other cases, once research is complete. CSOs have shared research findings through blog posts, website content, promotional materials and evaluation reports to funders and other organisational partners. Both academics and CSOs have presented research findings verbally at conferences, symposia and meetings. Partners in case 7 also hope to mobilise knowledge for educational purposes, such as by designing school curricula for elementary students that explore the impacts of climate change on Indigenous populations.

Who has access to what data and how?
The fourth question (column 5 in Table 2) inquired into the question of data access and the spatial implications of where data are housed and managed. In cases 2, 4, 5 and 6, data are initially collected and managed by academic researchers. In cases 2, 4 and 5, access is also managed by academic researchers and housed on their universities' servers, and in all three cases they are willing to grant access to anonymised portions of the data set by request from the organisations. Only in case 6 has it been made explicit in their memorandum of understanding that the copyright for the data remains with the CSO and not the academics, and that academics must seek permission and give credit before publishing findings based on the data from this research.
In case 1, data predominantly come from the CSO's own case files and information, which they often have little time to analyse to the degree that they would like. Other data in case 1, as well as that for the landscape studies in case 3, come from reviews of academic literature, which is primarily available to academics through university journal subscriptions. Most organisations find the hefty fees for journal subscriptions to be prohibitive and thus rely on their academic partners to provide access to academic knowledge on their fields.
Partners in case 7 emphasised that data and knowledge ownership must remain clearly with the Indigenous communities that they engage with. They assert that Indigenous protocols and the rules of relationships must govern the appropriate communication of some kinds of knowledge, while other forms of knowledge should be protected and preserved by Indigenous communities as knowledge holders.

Discussion: four lenses on power in knowledge production
Having described findings from the seven case studies, in this section I will analyse them in the context of the sociology of knowledge (Zingerli 2010) according to four lenses on power. It is important to note, however, that in a given research partnership, power can operate in many ways simultaneously. For example, one Program Director shared that his organisation received only a small amount of funds from the research partnership in comparison to the amount of effort that they were contributing, yet still found it overwhelmingly worthwhile: for us, it's not funding that we're looking for, but knowledge that we're looking for. At the same time, being able to answer some of these questions is very important for our ability to generate further funding for our work.
This comment illustrates the tension between wanting to build their own capacity to better development (the fourth lens), but also needing to think about future appeasement of funding agencies who control the means and resources for their organisation's continued operations (the second lens).
Lens one: values, discourse, norms and rules for knowledge production through research partnerships The research partnerships in this study exemplified how partners are often operating under the constraints and incentives of values, discourses, norms and rules that differ according to their sectors, disciplines or institutions. The question of methodological worth, for example, is an interesting example of how evidence hierarchies are manifested in practice. In the academic social sciences, randomised control trials are seen as among the most objective methods for determining the effects of an intervention, as researchers can try to control for the influence of other social, economic, cultural and demographic factors that could play into the efficacy of an intervention. Fields that rely on statistical models, such as economics, skew heavily towards preferring models that can be generalised and replicated. As one non-tenured economics professor highlighted, without strict adherence to their experimental protocol, the data would be "tainted and thus unpublishable" in his field, affecting the future of his career at his institution.
Such extreme methodological rigour in the research design, however, did not hold the same priority for the practitioner partners, for whom timely, practical and applicable research findings were of greater importance, with as little interruption to their core programming as possible. As mentioned earlier in the findings, CSOs identified other risks associated with methodologies that required them to deploy different interventions to different populations without their clear awareness or consent. Practitioners emphasised the great importance of building and maintaining trust with the communities that they worked with. While academics generally strove to respect this value, it did not hold the same immediate import or consequences for them as it did for their counterparts. At the same time, CSOs still do want high-quality data produced on their programs, particularly when such data could be leveraged for future funding. As such, the norms governing the respective research discourses came into tension with each other in certain partnerships in this study.
Lens two: control of means and resources in knowledge production through research partnerships The means of research production are controlled in these cases primarily through funding structures, incentive structures for promotion and tenure, and journal access paywalls. A lack of financial support has consequences for the extent to which partners can pursue a given research agenda. It also raises questions about what kind of research is recognised by funding agencies in Canada. Non-mainstream and Indigenous research especially risk being overlooked or marginalised by mainstream funding structures. A participant in case 7 described: I think the funding is always a sore point because I think the kind of work we're doing is seen as on the edge or on the fringe. It's not seen as valuable to the extent we see it. So, our job is to convince funders that it is an important topic and we need to engage with Indigenous communities on these topics.
The partners in case 7 describe being unwilling to compromise their values as having a cost in terms of the time, resources and labour that they put into funding proposals that are ultimately not accepted in the current Canadian funding landscape.
Several academics described the influence of the "publish or perish mentality" prevalent in their institutional and disciplinary contextsespecially for early-career and untenured scholars. Because of the importance of peer-reviewed academic publications for tenure evaluation and promotion, some academics describe being selective about partnerships that they expect will have good publishable outcomes, avoiding collaboration with CSOs with less interesting research questions. Others feel compelled to take on more of the research design and execution to ensure that the study will be publishable, rather than taking on some of this work more collaboratively with CSO partners.

Lens three: shaping social relations of knowledge production through research partnerships
Trust is a well-documented requirement for effective research collaborations (Zingerli 2010, 227). Partners identified networking as one of the primary motivators for partners to engage in collaboration; indeed, networking and building strong relationships are understood to be in themselves a pivotal mechanism for knowledge generation. Social relations characterised by reciprocity and adaptability are also conducive to research partnerships, as one CSO's program director described: "I wanted researchers who are generous, willing to learn and adapt. We put this risk on the table right away and made it clear we didn't want people who would come and take while leaving nothing behind." Having identified this approach from the outset has helped the relationship to evolve smoothly over the course of the partnership. In other cases, in which researchers took on a more authoritarian or even a transactional role, organisations identified discrepancies between their reported benefits.
Six of the seven collaborations in this multiple case study research were also shaped by a funded project. It was possible for either academic or CSO partners to be involved in multiple partnerships at once, creating overlapping umbrellas of research collaboration. Partners described benefits in terms of providing co-collaborators with a wider range of experience with effective relationships, as well as greater impact of research on their sectors. The power to leverage and benefit from research collaboration and other opportunities moves swiftly through these networks, facilitated by trust and shared histories of collaboration. On the downside, however, it can be difficult for newer organisations and academics to enter into tight networks, regardless of their merit. The social capital awarded to those who know the language, norms and social expectations of relationships may also receive greater opportunities, further marginalising those less connected or familiar with the norms of the field, as a measure of symbolic power. As Bourdieu (1994) might read it, power thus reproduces the very stratification and hierarchical tendencies that development researchers purport to overcome, and such a culture can prevent the most well-connected actors from being aware of those who are excluded.
Lens four: capacity for knowledge production through research partnerships Participants in this study also illustrated instances in which research partnerships leveraged the power to transform outcomes away from the social, economic and political reproduction that many political economists would expect to see (Giddens 1984). In case 4, for example, the academic research lab at a prominent business school challenged a fundamental assumption in the business fieldessentially, that serious knowledge production only takes place in the private sector. In case 3, CSO and academic partners are seeking to transform the hierarchies of power around knowledge production about literacy in Africa, by shifting the locus of such knowledge production away from the Global North. Workshops for Southern partners in case 6 were a measure to help change the means by which food security was measured and documented, through capacity building of local members.
It should be noted, however, that the exercise of "power as capacity" in these cases will need a much longer time frame in order to assess whether or not the capacity-building efforts in fact lead to lasting change. Intentions can easily be interrupted or co-opted, and the integrative power "to inspire loyalty, to bind people together, to develop legitimacy" (Boulding 1990, 25) can be rerouted as symbolic gestures of empowerment without lasting transformation. Power as capacity, then, is observed both descriptively as well as potentially in this study. A challenge facing many partnerships, bound within the time frames of a funded project, is that a few years is not long enough to ensure the sustainability of lasting change. The partners in case 7 seemed most explicit in their awareness of this constraint, continuing their collaborative efforts to promote knowledge dissemination on the topics of social justice and climate change, regardless of whether or not they are bound by a funded project.

Analysis and critique across cases
Through this research, I have inquired into the evidence priorities and practices of academics and CSOs in research collaborations. Future research could interview policy makers and funding institution members to understand their preferred evidence practices, methodologies and outcomes, and examine the political and economic interests that likewise inform them. Additional studies could inquire into the question of whether research conducted by Canadian CSO-academic partners is helpful or harmful for partners in the Global South. This issue was raised a few times by participants, some of whom felt that more could potentially be done in research partnerships to include added benefits for Southern partners, such as capacity-building opportunities and engagement with southern academics. The global knowledge economy's stark hierarchies centre academic knowledge production primarily in the global North, regardless of where their data is coming from. Partnerships between Canadian CSO and academic actors in the development sector are only part of the question of making the development sector better informed by evidence. The questions of who is in the position to produce evidence and what evidence is seen as valid and valued are equally significant.
Most noteworthy is who is not and cannot be funded under current practicespartnerships that place too much of the knowledge production process on the Southern partners who themselves are most impacted by Canadian development intervention or study. Canadian ODA requires measures to rely less on Canadian institutions and personnel to be directly involved on the ground in circumstances where grassroots or international actors would be best placed to research and read the reality of their own communities and circumstances. In this way, reconsidering the nature of "partnership" with Southern, community-based and Indigenous communities and institutions would be a decisive shift in knowledge-power hierarchies away from those that are currently upheld by the ODA policy.
As seen in the analysis of how knowledge and power are produced in these case studies, research partnerships are not simply one thingneither a panacea for solving the problems of the Canadian development sector, nor merely reproductive of the status quo. Cases in this research demonstrate evidence of altering the systems they are involved in in small and incremental ways through research partnerships. Research in Case 3, for example, is striving to create new centres for knowledge production on literacy in sub-Saharan Africa and thus de-centre the North. It has yet to be seen how this potential for research partnerships to challenge entrenched sectoral and disciplinary lines, to redefine what it means to be an "engaged academic" or what it means to be a "thoughtleader" as an organisation, or to reveal the limitations of funding structures and the premises that underlie them, could have more radical effects in the Canadian development sector.
At the same time, they also represent a response to the neo-liberalisation of academia and to the corporatisation of Canada's development agenda. Research partnerships that receive funding from Canada's major channels, in one way or another, are seen to "fit" with Canada's trade-centric agenda for foreign aid and development. Informants also found the Canadian government's mechanisms (or lack thereof) for drawing from the development research it funds to be ad-hoc at best. If, as Tiessen and Smillie (2017) suggest, Canadian development practice is not as evidence-informed as it should be, does this gap reflect the absence of collaborative spaces alone or the lack of clear channels through which development research is not only financially supported, but actively drawn from? Participants in this research wondered, along with me: how can we hope to claim that development is evidence-informed if we don't know what evidence Canadian public funding mechanisms are even seeing? Despite the best intentions of individuals within Canadian funding systems for development, the frenzied pace with which development funding is awarded, projects implemented and results generated does not always allow for meaningful reflection about how learning informs action. Programme officers are under pressure to deliver "results" and thus often have little time or incentive to consult with the development research community, even at its highest federal levels. This systemic "time poverty" to attend to research has direct consequences for development effectiveness. Whatever we can claim to know about development as a sector, then, is not nearly as much as it could be.

Conclusion
This article has reflected the insights of a diverse range of recent academic-CSO research partnerships in Canada that are operating within the current constraints and possibilities of the Canadian political economy of international development. The four lenses of power reflect the ways that values and norms, financial means and incentives, and social relations of knowledge production are produced and reproduced, and that what comes to be known about international development in Canada is heavily mediated by its political economy. At the same time, the fourth lens of power that data in this research indicatespower as capacitywarrants further exploration, in the ways that agency and structure interplay, even within constraints. Willingness to incorporate evidence into policy and practicenot only incrementally but into a review of the essential structures, relationships and hierarchies of development assistance, especially at the federal levelwould be a decisive step towards seeing the promise of research partnerships better realised. Note 1. In this article, use of the terms "development" and "development research" reflects the framing of this special issue as whole and their meaning in the Aims & Scope statement of the Canadian Journal of Development Studies: aiming to provide a "forum for critical research and reflection on the complex problems of international development theory, policy and practice". https://www.tandfonline.com/action/journalInformation?show= aimsScope&journalCode=rcjd20 Notes on contributor Elena Toukan is a PhD candidate in Curriculum, Teaching and Learning at Ontario Institute for Studies in Education (OISE)/University of Toronto. Her research interests include the interplay between global education policy and local agency, as well as the acquisition, generation and mobilisation of knowledge through education, particularly as it relates to community development. Elena is an associate editor at Curriculum Inquiry, and her published research appears in such journals as Education, Citizenship & Social Justice, The Journal of Peace Education, Education as Change, and Compare.