Terminology and tensions within evidence-informed decision-making in South Africa over a 15-year period

In this article, we examine a key premise underlying evidence-informed decisionmaking (EIDM) – that research is for all, including service users and potential users, service providers and a wide range of decision-makers, from those running local services to national government officials and international agencies. Qualitative data collected on terminology used when writing and talking about EIDM over a period of 15 years during the implementation of a number of capacity development programmes in South Africa were combined with critical reflections in practice. Findings reveal that tensions exist in the titles and terminology used to describe the relationships between academia and government or between research and policy, and that these tensions have shifted over time, but not necessarily diminished. An analysis and critique of this terminology is provided to identify and unpack these tensions, which challenge the central premise of ‘research for all’. The perpetuation of divisive labels that profile people, of job titles and specific terminology that describe agency, as well as the use of technical language, continues to exclude people from the approach. These have the effect of setting up users against producers of evidence. In conclusion, we challenge the advocates of the EIDM approach to review language and terminology to be more inclusive, to enable relationship-building and ease the process of engagement to ensure evidence-informed decision-making is true to its premise that research is for all.


Introduction
The use of research in evidence-informed decision-making (EIDM) is positioned as a public good, which anchors the concept firmly with the research for all premise. Research is perceived as serving to improve the lives of service users, and to inform the decisions that affect us all (Chalmers et al., 2012). In order to use research to inform decisions and to improve lives, research needs to be relevant to all and informed by those with real-life experience. If research is not for all, that is if certain groups are excluded from the EIDM process, this challenges the merit and benefit of research use. Groups can be excluded from research use in multiple ways: not being able to access published research; not being able to make sense of research because it is overly technical; not being able to integrate research results with groups' contexts as the research scope or emphasis is not relevant to these contexts. In each of the examples, users are hindered to consider research when making decisions and thus excluded from the potential insights and benefits of research for their decision-making. If not everybody is able to access research, consider it, and use it to improve lives (where relevant), then research is not for all. Research use and EIDM remain a privilege of those with access to research and the technical skills to make sense of it -a situation that violates its research for all premise.

Evidence-informed decision-making and research for all
Evidence-informed (or evidence-based) decision-making (EIDM/EBDM) encourages the use of rigorous evidence for better decisions. For the purpose of this article, we will use the term EIDM to refer to both the practice and academic discipline of research use. We therefore operationally define EIDM as a process whereby multiple sources of information, including the best available research evidence, are consulted before making a decision to plan, implement and (where relevant) revise policies, programmes and other services. EIDM has a history in evidence-based medicine that 'integrates the best external evidence with individual clinical expertise and patients' choice' (Sackett et al., 1996: 72). In addition to health care, EIDM has an important track record of improving policies and practices ranging across different sectors (for example, social welfare, education, environment change, justice and international development). Public programmes around juvenile crime preventions, saving household electricity or early childhood education pay testimony to the benefits of EIDM (Petrosino et al., 2013;Halpern, 2015;Fiennes and Wulf, 2014).

The case of learning styles
The myth of learning styles presents an illustrative case to underline the centrality of the research for all premise in EIDM. Despite rigorous synthesized research showing that there is no scientific evidence indicating a positive impact (or even existence) of teaching according to different learning styles (Pashler et al., 2008;Rohrer and Pashler, 2012;Coffield et al., 2004), and these findings being communicated through multiple channels and by multiple sources, teachers' support for the use of learning styles is almost universal (Dekker et al., 2012;Newton, 2015;Goldhill, 2016). What is more, teaching according to perceived learning styles is also highly popular among parents (Dekker et al., 2012;Newton, 2015;Goldhill, 2016). There is therefore a clear disconnect between the research findings and the preferences and beliefs of educational practitioners and stakeholders. These two groups were not included in the process of synthesizing and assessing the research evidence on learning styles and they do not access the insights and benefits of the produced research (that is, understanding the limitations of teaching according to learning styles and using more effective teaching methods). In this case, research did not influence decision-making beyond the academic context, and its potential positive impact on educational practice was not realized.
Describing evidence use: Inclusivity in terminology?
In the literature, and in practice, research for all is in many ways an ideology, rather than a reality; something that should be aspired to. The journal Research for All itself represents a shift in terminology towards a positive, inclusive environment. We are defining 'terminology' as the words used to describe and label individuals, concepts and activities in EIDM. Where terminology is used to categorize professional levels, we have used the term 'title'. Despite the wide range of efforts to ensure research is for all, in both its production and its incorporation into decision-making, the terminology relating to this field is loaded with division, even when the emphasis is on working beyond or breaking down that division. People may feel excluded because they do not understand the use of technical language, they do not relate to the stakeholders referred to, or simply because there is an over-emphasis on the lack of evidenceinformed decision-making, which feels judgemental or exclusive. For example, evidence use is often referred to as: bridging the evidence-policy gap (Grimshaw et al., 2012;Crewe and Young, 2002); getting research into practice (Bero et al., 1998); knowledge translation or transfer (Davis et al., 2003); or knowledge brokering (Meyer, 2010). The box below lists the many terms applied in the literature to describe the process of using research and the study of research use. It is worth noting that: the verbs used are almost exclusively about 'pushing' or 'transferring', indicating a linear movement from point A to point B; the starting point is invariably the research and not the decision-maker or decision-making process; and lastly, that very few terms describe a system rather than an action. Furthermore, the various stakeholders in the evidence-use process are categorized as being on either side of a divide. On the one side, users are often described as policymakers, decision-makers, practitioners or adopters; on the other side, producers are located as researchers and academics, thereby reinforcing the polarity between generators and users of research. An exception to this two-community model of research use (Caplan, 1979) is the provision for intermediaries such as 'knowledge brokers', but even this terminology for those 'in between' the generators and users of research implies some kind of conflict between two communities that the broker has to bridge and resolve.

Commonly used EIDM terminology
Note that 'research', 'evidence' and 'knowledge' are all used interchangeably in these common terminologies.
Research utilization; research dissemination; research implementation; research diffusion; research uptake; knowledge mobilization; evidence use; knowledge application; knowledge exchange; knowledge translation; knowledge transfer; evidence broker; evidence champion; research adoption; knowledge sharing; evidence to policy; evidence into action; research into practice; research impact; evidence-based policy and practice; evidence-informed policy and practice; evidence-based decision-making; evidence-informed decision-making; evidence ecosystem; evidence-policy gap; evidence system. This article explores how this terminology of research use in evidence-informed decision-making has shifted over time in South Africa, and discusses future directions for use of terminology and inclusivity in this field.

Data sources
South Africa has a growing community of people interested in supporting the use of evidence in decision-making (see Stewart, 2015 andDayal, 2016 for more on this context). Our analysis draws on our own work in South Africa spanning 15 years, during which time we have been adopting and adapting terminology to suit the circumstances in which we have been working. We are specifically drawing on data that we collected during three phases of work, in 2001, 2007 and 2014-16, supporting others to engage with evidence-informed decision-making (Ellison et al., 2001;Stewart, 2001;Stewart, 2015). In each of these three phases, colleagues from the universities of Johannesburg and London (including the authors of this article) have collaborated with a range of decision-makers and researchers within South Africa to encourage EIDM, including health and education practitioners, non-governmental organizations and colleagues in national government. This has included provision of workshops and mentorships in EIDM, as well as conducting joint research. Data on the terminology used have been collected using a variety of methods, including: PowerPoint slides, handouts and worksheets used to support the use of EIDM over the three phases; transcribed audio recordings of workshop discussions; and flip charts on which key issues have been noted. Verbal consent was given in all cases for the use of these materials for future analysis and publications. Written consent was obtained whenever audio recordings were made.

Analysis
We have conducted a secondary analysis of the data using thematic content analysis of terminology use over time (Cameron, 2001). We followed Guest et al. (2012) and developed descriptive and analytical themes for the data from our three phases of work separately. Therein, we used a mix of inductive and deductive themes (Fereday and Muir-Cochrane, 2006), that is for each data period we identified themes according to a predefined framework drawing on the authors' combined experience of working in the field of EIDM. These themes focused on: (1) who are the perceived actors in the evidence-use process and how are they positioned in relation to each other; (2) how the process of evidence use was described; and (3) who is perceived to drive the evidence-use process. In addition to the exploration of these deductive themes, we further interrogated the data for indicative themes. In this, we used a configurative approach vested in the assumption that language use in psychological terms reflects thoughts and feelings, which can reveal implicit assumptions and power relations (for example, the divisive positioning of research users and producers discussed earlier) (Thomas, 2006). Trends and shifts in terminology over time were noted, and patterns identified through reflection and discussion.

Authors' roles
This paper has been co-authored by university-based researchers and a colleague in a central government department. The authors were involved in the delivery of EIDM projects from which the data for this analysis were collected. The use of 'we' in this article therefore neither represents the research community nor the public service.
In addition, the special working relationship between the authors, and our dual role of providing data and analysing data, deserve attention. First, there is a question of objectivity. Our close working relationship might influence the way we interpret data. We had to take precautions so as not to be influenced by hindsight bias, which might have led us to rearrange data and patterns to lead to a linear and coherent story to fit with our current intellectual positions. This was particularly important in relation to the development of our own joint working relationship, and we actively interrogated ourselves so as to counter the illusion that our current working collaboration as decision-makers and researchers was all but inevitable or primarily of our own making.
Second, the joint reflection on the data set and its analysis led us to discover and challenge a range of assumptions that we thought we mutually shared. For example, we assumed that we had developed a joint understanding and shared views about the titles and terminologies we were using. It was interesting to realize that, in some instances, different authors had simply taken up certain terms because they had heard trusted colleagues using them. Often, we had assumed that terms were adopted because we had a shared understanding about their conceptual advantage in reducing tensions, which was not necessarily the reason for adopting the term. Revisiting the data together for this article led to many interesting discussions tracing back the initial conceptual argument in favour of a term that some of us had simply accepted as a more relevant term through social interaction. Third, by reviewing the data collected at different points in our joint engagement, we realized how our own behaviour and use of terminology had influenced our work. This was a key difficulty in re-analysing the data, as it often made us aware of our own shortcomings and, at times, naivety in trying to support the use of evidence.

Findings
Our analysis revealed five overarching tensions in the terminologies used to describe EIDM: (1) in the titles that are used to address people; (2) in how EIDM is conceptualized; (3) in how EIDM is practised; (4) in how power relationships are constructed; and (5) in the concept of research for all. We discuss each in turn below.

Tensions in how people are addressed
Our data indicate a real and persistent tension in descriptions of people, particularly those working in 'other' environments (what we are referring to as 'titles' for people). These tensions are characterized by uncertainty and frustration with the apparent emphasis on hierarchies and job titles by those working in academia and in government, suggesting a process of adjustment is required while communities of practice that are not used to working together establish their individual status within this multidisciplinary environment. An example of this is the way in which a senior government colleague dismissed academics' obsessions with hierarchy and titles: 'They insist on being called "professor", but they don't call us "director"'. At the same time, colleagues in universities have highlighted the many hierarchies in government and expressed annoyance with the way in which some senior civil servants expect to be treated because of their positions. For example, one senior civil servant at our workshops resented being sent an invitation by email and requested something more formal. Despite noting these current frustrations in our data, we have noted shifts in terminology among some groups, perhaps best illustrated by how we (the authors) refer to one another. Currently we use terms such as 'colleagues' and 'partners' to characterize our relationship, but initially terms such as 'collaborators', 'workshop participants' and 'evidence champions', and the use of titles such as 'professor', shaped our positioning of each other. The initial terms we used were informed by how we came to know each other, in contrast to the current terms we use, which are informed by how we work together.
Tensions also extend to how the research-use process is itself entitled. Our analysis indicated the following four key issues with the terminology that people use that potentially leads to tensions in how we describe the use of research: (1) the supply of evidence is described as a linear push process, indicating few democratic elements of participation, multiple voices and shared control; (2) terms such as 'research communication', 'dissemination', 'diffusion' and 'transfer' all invoke a linear process in which researchers enlighten users about their findings; (3) there is no two-way flow of information or challenge to researchers' status as the bearers of knowledge that is implicit in these terms -evidence is produced and then supplied, and this is as far as the researchers' involvement in the EIDM process goes; and (4) encouragingly, this model of EIDM and its associated terms are now used less, with some proclaiming, for example, that 'dissemination is dead' (Clark, 2016).

Divisive terminology in the EIDM concept and process
The terminology used to conceptualize EIDM is divisive. The importance of demand for evidence and the agency of evidence users to shape EIDM is not reflected in most terminology. We identified within our data the use of terms that assume that the evidence produced is the finished product that has only to be accessed and used by decision-makers, including 'knowledge transfer', 'research uptake' and 'knowledge application'. The implication is that evidence will automatically be taken up and applied (implemented, mobilized, utilized, adopted). This terminology does not cater for a change of the evidence by the users, who are semantically presented as a passive entity. We observed a few terms that allow for a more active involvement of decisionmakers in constructing evidence and an increased agency to determine its usage. These refer to knowledge 'sharing' and 'exchange', which are less one-sided and less one-directional, recognizing that all stakeholders have knowledge to share and exchange. However, these terms seem to miss the evidence use and decision-making aspect of EIDM and do not move beyond the supply of evidence.
Furthermore, the findings demonstrate that the EIDM process was frequently presented as something initiated by researchers. The few terms that picture EIDM as a process (for example, 'research into action', 'knowledge into practice', 'evidence to policy') all position the process as starting with the research. This reinforces the push element of EIDM, does not see EIDM actors on the same level and gives EIDM a research-dominated feel. The assumption within the terminology we observed, that EIDM is initiated by researchers who push knowledge into action, suggests that the users of research should merely receive knowledge in a passive way. We saw repeated use of research-driven process terms that construct a false dichotomy between research and action, knowledge and practice, evidence and policy.
Despite the conceptualization of decision-makers as the recipients and not the initiators of EIDM, recent innovations in EIDM terminology present a more organic and systemic understanding of evidence use. We recorded the use of terms such as 'research-policy landscape', 'evidence-policy interface', 'evidence discourse' and 'evidence ecosystem', all of which seem to indicate a more democratic understanding of evidence and decision-making. The process of use in these examples from our data is more interactive and less linear; actors involved in EIDM are seen on equal footing and possess the same power to shape the process. Such systemic terms are far from the norm and can likewise fall into a more divisive conception of EIDM (as is the case of talking about 'evidence-policy gaps' and 'divides to bridge'). This raises the question of what to prioritize: research or policy? In South Africa, for example, Dayal (2016), writing as a public official in government, uses the term 'policy-research interface', while Choge et al. (2014), who are writing from an academic perspective, refer to the 'research-policy landscape'. Clearly, even the use of these compound terms reveals institutional biases and one's inherent conception of EIDM.

Tensions in the practice of EIDM
Our analysis indicated that the inherent tensions of EIDM terminology had important implications for the practice of EIDM. We found multiple instances where EIDM terms coined by academics (examples include 'systematic review' and 'critical appraisal') did not resonate with decision-makers, resulting in a reluctance to engage with EIDM and even, at times, hostility towards the approach. For example, one decisionmaker branded the EIDM approach as too critical, as it seemed to imply that current government decisions are evidence-uninformed. Another public servant recalled being talked down to by researchers who knew little of how government works or how policy is made. A strong theme in the data was a researcher-driven construction or implicit suggestion of a deficit model of decision-making in practice and policy that, unsurprisingly, increased reservations toward EIDM. For example, civil servants felt that academics made unhelpful assumptions about people working in government, labelling them collectively as policymakers and assuming that they need to have their capacity built. We have noted repeated challenges from colleagues working for government that university academics, and other outsiders, know little about social policy or the processes of government. In line with this theme was our observation that when EIDM is presented to decision-makers by others within government, it is better received than when presented by outsiders. On one level, this internal promotion of EIDM within government reduces or even removes the tensions discussed above, where individuals are busy emphasizing their own status through the use of hierarchical titles, and pushing an approach from 'outside' with poor understanding of internal government language and practice. On another level, it seems that if EIDM is presented by government colleagues or encouraged by government institutions, the use of evidence is demandled and is deemed as a more legitimate process, and potential best practice, in the public service sector.
Not only did we find assumptions that civil servants have a skills deficit that needs to be addressed, we also found no acknowledgement in the terminology used that researchers have a deficit in their own skills and understanding. For example, civil servants repeatedly spoke of frustration at the assumption that they need their 'capacity built', and expressed annoyance at how poorly informed researchers are about how government works. Taking this point further, we also found limited EIDM capacities among the researchers who were themselves promoting the approach. For example, few EIDM capacity-building programmes with which we engaged were informed by rigorous systematic review evidence of which capacity-building models work best (Stewart, 2015;. Our analysis also showed that when promoting and supporting systematic reviews (often positioned as the gold standard for evidence within the EIDM paradigm), most researchers who are unfamiliar with EIDM themselves feel threatened and become defensive. They are not always comfortable with the ideas of critical appraisal and of excluding research based on predefined criteria. By virtue of making decision-making explicit and requiring criteria for what information is relevant and trustworthy to inform this process, EIDM can be perceived as challenging and threatening (Stewart, 2007). The defensive responses observed by decision-makers and researchers alike suggest that some tension is inherent within the approach, which presents a conceptual challenge given EIDM's embrace of the research for all premise.
Our data indicate a progressive shift over the 15-year period towards a more inclusive practice of EIDM. This included changes from using terms such as 'training', to 'capacity building' and 'capacity sharing'. It also included an increasing emphasis on the need for broader, arguably non-research, capacities to complement the research skills that were promoted in the earlier stages of our work. The EIDM paradigm calls for research skills and/or evidence-use skills to be built (searching for, appraising and synthesizing evidence), and these were very much the emphasis of our workshops in 2001 and 2007. In 2015/16, we saw a shift towards the inclusion of programme planning and programme evaluation within this EIDM capacity-building package. Given the strong role that government has taken in South Africa in building the evidence system, there is also an increasing shift to conceiving evidence use as a process that starts with the user rather than the producer of evidence (Langer and Stewart, 2016;Dayal, 2016).

Prevalent but shifting power dynamics
The practice of EIDM also reveals a tension in power relations. This has changed over time, but still remains today. An early theme in our data was the unease among nonacademics when talking about research and researchers. We observed nervousness, laughter and sometimes silence when non-academics were invited to comment on a piece of research. However, this has shifted over time, with a qualified and experienced cohort of people working within government who are confident in their engagement with academia and with consultancies. There is less dependency on researchers to come up with answers alone, allowing increasing numbers of research users to assume agency in setting the research agenda as well as the terms of research usage. Also, more researchers are entering, and have taken up a career path in, the public service sector, which may be contributing to the shift in terminology and practice. In 2001, data show unease with research terminology by decision-makers. Non-research participants in our workshops repeatedly asked for terms to be explained and expressed frustration with researchers, feeling that they were being deliberately excluded by researchers' use of technical language. In 2007, the unease about research terminology was noticeably reduced among participants, who were more comfortable discussing pieces of research and research methods more generally. It is impossible to know whether this shift has taken place across the board in South Africa, but it was noticeable from our data that people felt more comfortable with the concept of EIDM and the terminology used.
Finally, in 2015/16 we have experienced discussions in which government colleagues are confident in talking about research and the use of technical language around EIDM. On the other side, non-government colleagues are beginning to recognize how much they have to learn about how government works, and we have come across very few examples of academics with a good understanding of policymaking and procedures in government.
This shift in power extends to meta-discussions about what is research. Now we are observing tensions within government in how people understand 'research'; questions about how research informs planning, monitoring and evaluation; and whether 'evidence' is a better term to use. It is important to note that these are discussions we are having with colleagues in government, and not with researchers -a clear indication of how the agency in EIDM has shifted. Government colleagues are starting to play a greater role in agenda-setting for EIDM. For example, they demand 'policy-relevant' research, which entails tailoring research methods and production to take decision-makers' needs into consideration. Government is no longer a passive consumer of research, but is developing an active voice about what it understands as relevant research and how it can be supported to use this research for decision-making.

From tensions towards research for all
This analysis has shown how the way in which EIDM is presented is directly challenging for decision-makers and provokes an unhelpful tension or polarity between those who produce research and those who use it, challenging the premise of research for all. We have, however, observed a marked shift in how the approach is presented in response to these tensions and in an attempt to diminish them. EIDM is often introduced using spectra and cycles to illustrate the process of integrating evidence into policy decisions. These tools are generally developed by those outside of the policy space and promoted to decision-makers. In South Africa, we have clear indications of how this has progressed in our capacity-building activities. In 2001, we used a linear model and imposed it from outside with 'unsubstantiated certainty' at one end of the spectrum and 'evidence-based certainty' at the other end. In 2007, we changed our linear illustration of EIDM into a cyclical diagram to describe the EIDM decision-making process. This had less emphasis on the strengths or weaknesses of an individual's position, and more focus on the cycle of reviewing the evidence, making a decision, implementing it and evaluating again. However, this cycle continues to be developed by the researchers and presented to decision-makers as a way of improving their work. In 2015, we saw a marked shift as we are now being encouraged to promote an evidence-based policymaking cycle developed by colleagues inside of government (DPME, 2014). This shift has implications for capacity-building curricula for EIDM, as well as for the general approach to research use.
In addition, challenges to the dichotomy between research users and producers have increased, now taking the form of strong opposition rather than merely concerns or challenges. There has been an increase in calls for co-production and co-use. In 2001, we had calls for co-authorship with non-university attendees, who felt that they were left out of publications that were about them. Non-researchers voiced criticisms of researchers for conducting research with the decision-makers as research subjects, and then disappearing and not reporting back, giving those same decision-makers no chance to learn from the research. By 2015, there had been several calls for the breaking down of the system structure that suggests that some people produce and others use research. In many cases, calls for co-production have been conceptual arguments for a shift in the terminology used and the ideas explored. In other cases, we have started to see co-production where those employed in academia are working with those employed in government to co-produce evidence maps to support EIDM.
These welcome shifts towards a more inclusive conception and practice of EIDM, reducing tensions and changing roles and power relations to move closer towards the premise of research for all can be observed in a number of key terminology changes as well. EIDM itself -while still imperfect -presents a shift away from a topdown research-driven conception of evidence use, as in evidence-based practice and policymaking. 'Evidence-informed' acknowledges the many factors other than evidence that influence decision-making, while the use of 'decision-maker' rather than 'policymaker' acknowledges the diversity of public sector professionals and their roles. In South Africa, the term 'evidence' is increasingly substituted for 'research' to incorporate monitoring and evaluation data, and government internal evaluations and reports.

Discussion
In our analysis of the titles and terminology used in EIDM in South Africa over a 15-year period, we have noted several areas of tension, while also observing some marked shifts towards research for all. We are nonetheless left with some important questions, which are outlined below.
One issue is whether the word 'research' is actually accurate or whether 'evidence' is a preferable term, particularly if different people use 'research' broadly or narrowly, without necessarily providing a definition. It is not clear which term is more accurate or inclusive. Both words exist in lay terminology but mean slightly different things in different contexts: a child may research a topic for homework, a lawyer may bring evidence to a court room.
A second issue is whether, if research is for all, the terminology around research use should be adapted to overcome the two-dimensional conception of research and decision-making, and perhaps be described instead as an organic process within an evidence ecosystem. By avoiding a two-sided conception, this also avoids the necessity to name one side first: the policy-research versus research-policy tension. This overlaps with some related discussions about 'evidence literacies' (Newman, 2014) and 'evidence thinking' (Dayal, 2016). Both authors wonder whether evidence use is a behaviour that should be inherent in the decision-making of all actors in society. This view positions the use of evidence as a behaviour emerging from the bottom up in society, and that shares and reinforces the skills associated with it at all levels of decision-making. Individuals shaped by society to be receptive towards the use of evidence, then merely advance and apply this behaviour of using evidence when assuming decision-making positions.
A third issue that we are deliberating is whether or not it matters if practice is keeping up with terminology. Should we be proposing and using more inclusive terminology to show the direction in which we hope to move; or should we be using the terminology that stakeholders prefer in their contexts, whatever those terms might be? This line of thought equally applies to the term 'research for all'. Would practitioners, decision-makers, and the general public actually agree that research should be for all, or is this an assumption made by academics on behalf of others?
This question leads us to our last issue, which some might view as the solution to our tensions, issues in finding common terminology and practice of supporting the use of evidence: co-production. Many argue that institutionalizing co-production will diminish tensions in terminology and practice of EIDM, as researchers and decision-makers become more familiar with one other, develop common practice, and discuss relationships and language, resulting in evidence that is useful and will be used. This might be a desirable but somewhat romantic notion of co-production. Our reflections and experience indicate that co-production can be a powerful tool to establish trusted relationships and support evidence use. We have also experienced the practicalities of making co-production work and the politics associated with this method. Co-production enters both parties into a situation of interdependency, which, if successful, can emphasize each other's respective strengths and expertise. It can also reduce stress and tensions, for example users' concerns about whether the final product will be useful to them or producers' worries about whether there is actually anybody waiting to use their research. However, we have also observed that making co-production work requires an investment in financial and non-financial resources, including time and perseverance by both parties, and that there is not a single method of co-production. Co-producing with government officials, highly skilled and equipped with postgraduate degrees, differs from co-producing with others who may work in different spheres. If one does not tailor co-production carefully, there is again a danger of placing users into EIDM boxes in which they do not fit.
The user-involvement movement in health promoted as much user engagement as possible, with 'user-led' research promoted as a gold standard by some (Faulkner and Thomas, 2002). Is this the ultimate means to achieve research for all, that is removing the researcher as an institution in the research process, and if so, what does this mean for EIDM? We are cautious about advocating such a complete blurring of professions, in which the profession of researcher and its associated skill set become obsolete as this function is primarily exercised by users. Stewart and Liabo (2012) argue for contributions from all (research users and producers), but taking care not to diminish the technical skills required to conduct research. In 2017, we reflect on this and extend the argument that co-production equally runs the risk of diminishing the technical skills of civil servants. We are wondering whether research for all can best be seen as a spectrum of engagement that varies with the particular context of research producers and users. This would position co-production not as a prescriptive end goal, but instead as a spectrum that recognizes the different elements required in EIDM and how people need to work together with mutual respect to achieve the multidisciplinary, participatory and problem-based processes required. The context in which evidence is to be used will determine what expertise is required, as well as how much to blur the boundaries between different types of expertise.

Conclusion
Terminology can both reflect and create barriers, and the kinds of partnership working that breaks down those barriers. If EIDM is to reflect the principle of research for all on which it is based, there is a need to continue the process that we have begun, of reflecting on the terminology we use, critiquing our own positions and looking to a future where our terminology facilitates the production and use of evidence.

Notes on the contributors
Ruth Stewart is Director of the Africa Centre for Evidence, University of Johannesburg and chair of the Africa Evidence Network. She has worked for nearly 20 years on the production of synthesized evidence to inform decision-making, including systematic reviews, rapid evidence assessments and evidence maps. She is committed to making a difference through her work by ensuring that research evidence is useful and used, most recently through leading the three-year University of Johannesburg Building Capacity to Use Research Evidence programme (UJ-BCURE).
Harsha Dayal is Director: Research Management at the National Department for Planning, Monitoring and Evaluation in South Africa. Harsha has worked extensively on the nexus of science and government, including positions at the University of Witwatersrand and the South African Human Science Research Council, before joining the public sector. She led the 20-year review in government and recently published a report on using evidence to reflect on South Africa's 20 years of democracy.