Exploring attitudes to societal relevance: the effects of reflection on research practices among Swedish environmental scientists^*

* The research was conducted at Linnaeus University, Sweden.

ABSTRACT

Funding agencies and policy-makers have put increasing pressure on scientists to better clarify the usefulness of their research. It has been suggested that this may have led to an increased reflection on the societal relevance of research among the scientists themselves. However, this often is more an assumption than a carefully verified fact. This paper investigates whether reflection on societal relevance actually occurs and has a measurable effect on the choice of research and on dissemination activities performed by scientists. A survey was conducted among researchers in environmental science and technology at Swedish universities. Results show that researchers do frequently reflect upon the societal and environmental relevance of their work. We used path modelling techniques to assess how this influences their professional activities. Results show that reflection was important to explain both the choice of research and dissemination activities. Variables such as individual attitudes, experience and collaboration with external actors also affected these outcomes.

KEYWORDS:

• Mode 2
• relevance of research
• choice of research
• environmental science
• reflection

Introduction

The social contract for science, which concerns the responsibilities science has towards society and vice-versa, has been widely discussed in recent years (e.g. Owen et al. 2013; de Saille 2015). These responsibilities became increasingly debated when the Mode 2 concept was introduced (Gibbons 1994). Mode 1 scientific research is theorised as the traditional way to produce academic knowledge, within a university setting and with an internal quality control made by each discipline. Gibbons’ idea is that science has shifted to a new mode characterised by the increased importance of transdisciplinary and applied research. In Mode 2, the usefulness of research becomes a success evaluation variable in itself and, as a consequence, scientists need to reflect on the societal relevance and accountability of their research. The assumption is that research cannot be limited to scientific and technical dimensions alone when the goal is to address problems concerning important social issues. This said Modes 1 and 2 can be thought to coexist as ideal types representing extremes of a continuum within the real world.

Nowotny and colleagues subsequently improved the theoretical foundations of Mode 2 (Nowotny, Scott, and Gibbons 2001), but the idea is still challenged by other approaches on contemporary knowledge production. These include a criticism of a capitalist trend in science (Slaughter and Leslie 1997) and the post-normal science approach stressing the limits of ‘normal science’ (Funtowicz and Ravetz 1993). In addition, the triple helix model focuses on the role of increased interactions between university and other sectors to promote successful innovation (Etzkowitz and Leydesdorff 2000).

Other concepts that have been used to denote the social or societal relevance of research include impact, outcome, accountability, benefit, usefulness and quality (e.g. Mobjörk and Linnér 2006; Roberts 2009; Dilling and Lemos 2011; de Campos et al. 2017). The product of Mode 2 research has been called socially robust knowledge. Such knowledge should be open to monitoring, testing and adoption by a variety of society representatives, who by doing so can validate the research (Nowotny 2003). With a trend towards more collaboration with partners outside academia, research would as a consequence be increasingly monitored and evaluated on the basis of its usefulness.

In a seminal study, Hessels and van Lente (2008) identified seven different ideas of knowledge production, all recognising the shift towards more socially relevant research and the increased interaction between sectors. Among them, reflexivity, social accountability and the context of application are of particular interest for our purposes. We used them as a framework for our research, which focuses on whether scientists do reflect upon the relevance of their research and if this produces an actual effect on their scientific choices and dissemination activities. More specifically, we explored how environmental scientists reflect on how science is conducted, perceive the pressure to produce research with a measurable impact on society and the environment, and accommodate these factors in their professional activities. This was done through a survey of environmental scientists at Swedish universities, to assess their reflection and attitudes towards science in general and the relevance of their research in particular.

Research background

Science is generally linked to an increasing understanding of the world and achievement of societal goals. These two goals have become especially important today given the environmental challenges faced by the present and future generations. Scientists hold a core responsibility to address these challenges in the ‘century of the environment’ (Lubchenco 1998). Here science and policy often meet (Miller 2001; McNie 2007), as the complexity of environmental problems requires the various stakeholders to interact and collaborate with each other. This results in the crossing of disciplinary and organisational boundaries (Perz et al. 2010), meaning that environmental scientists have to learn to collaborate with other actors. At the same time, the involvement of science in environmental controversies also leads to a politicisation of research and its results (Sarewitz 2004). One study found that senior environmental scientists recognise the need for increased political involvement with society, although they express concerns about the consequences of such participation (Barns and Wilson 1996). In such cases, the scientists’ involvement can become an instrument for politicians who exploit it to support their political position.

In the Swedish context, a trend towards increased applied research has been recognised, partly because of the top-down steering of university–industry interactions operated by research funding agencies (Benner and Sandström 2000). Although the terms ‘basic’ and ‘applied’ science have long been contested (Pielke and Byerly 1998) and are often used to deflect the responsibility to reflect on the social relevance of research (Calvert 2006), we employ the terms here because of their continued use by scientific and science policy actors. A second trend points towards an increased importance of the university ‘third mission’: i.e. that universities should disseminate scientific information and collaborate with the society on important problems. This brings new tensions within academia, as it implies additional responsibilities for science to bring benefits to the society (McCarthy and Kelty 2010; Benneworth, de Boer, and Jongbloed 2015). The interaction between science and society also creates new challenges for individual scientists, as they apparently need to reflect more on the demand to produce socially relevant research.

Previous research helped to define the societal impact of research and how it can be measured (Bornmann 2013; Watermeyer 2014; cf. Donovan 2017). While most studies focused on the economic dimension of this impact, a report from the European Commission also emphasised environmental improvement as an important factor (European Commission 2010). A large-scale survey in Sweden (10,000 respondents) approached the question of impact and dissemination. Dissemination could be seen as one form of relevance (Wahlbin and Wigren 2007), although the activities actually performed outside the university depended on the researcher’s position and research area. The study also highlighted that the attitude of researchers towards collaborations outside the academia was generally positive or very positive.

Scientists’ perceptions of the social impact of science were also explored in another survey, with a majority of the respondents reporting that their latest highly cited paper had significant social and political implications (Small, Kushmerick, and Benson 2007). The response to policies that aimed to increase the societal relevance of science was also considered, with researchers who were generically aware of them but not that some of their main activities could be categorised as having relevant social impacts (de Jong, Smit, and van Drooge 2015).

There are many ways in which scientists can interact more intentionally with society. These include collaborative research, contract research, consulting or commercialisation activities as well as public dialogue and stakeholder engagement (Wilsdon and Willis 2004; Perkmann and Walsh 2008; Perkmann et al. 2013). Collaborative approaches such as socio-technical integration have served to facilitate engagement between researchers and stakeholders (Fisher et al. 2015). Previous studies identified a framework to analyse the effects of academic research including seven elements: research, publishing, education, guidance (advisory roles), commercialisation, infrastructure and networking (Jacobsson and Perez Vico 2010; Jacobsson, Vico, and Hellsmark 2014). A complete list of the usefulness of academic activities also included networking, facilitating dialogue, infrastructure, guidance and public outreach in general. It was also shown that the majority of scientists lay somewhere between an entrepreneurial role and a more traditional ‘researcher’ role, being neither solely negative nor positive to the shifting boundaries between university and industry (Lam 2010). Furthermore, scientists have been shown to be aware of how these ideals of science and the context they are working in are changing (Waterton 2005). At the same time, a counter-reaction to these trends has also been identified and includes scientists who prefer to maintain their disciplinary boundaries and traditional activities (Gieryn 1983; Swan et al. 2010).

The current state of knowledge creation can also be seen as a shift from sponsorship to knowledge partnership in university–industry collaborations. While industry once provided funding and the outline of research and academics could steer its contents, the knowledge-creation process currently focuses on joint problem solving (Jacob et al. 2000). More generally, the involvement of researchers within the industry is well studied in reference to commercialisation activities, reasons to collaborate, success and failure of joint activities and much more (Lee 2000; Barnes, Pashby, and Gibbons 2002; Wigren-Kristoferson, Gabrielsson, and Kitagawa 2011). Reflection on societal relevance, and the effects of such reflection on their research practices, however, has rarely been subjected to systematic study.

Hypotheses

Recent literature on the societal relevance of research poses the empirical question of whether university scientists are increasingly reflexive, especially in the sense of being more aware of potential societal effects of their research, and take these into account in their choices about the research topic, methods and approaches (Hessels and van Lente 2008). In addition, it has been argued that, although the contract between science and society has been longstanding, the usefulness of research has become an increasingly important criterion today (Hessels, van Lente, and Smits 2009; cf. Guston 2007). This led to the question of how environmental scientists in Sweden perceive the increasing pressures to do ‘useful’ research. In addition to this, we were interested in how these pressures influence research choices and practices. We hence propose the following hypotheses:

• H₁: Environmental scientists do reflect upon the social and environmental relevance of their research.

• H₂: The reflection upon this relevance influences the scientists’ choice of research.

H₁ is a compounded hypothesis where it is assumed that scientists reflect both on the social and environmental relevance of their research, although these dimensions were split into different questions to capture eventual differences between them.

In H₂, ‘influences’ refers to the existence of significant paths between the variables included in the proposed models (see below). In practice, it is likely that reflection and choice of research interact in a positive feedback cycle. In other words, doing research on environmental issues affects reflection on this topic just as reflection affects the researchers’ choices. Moreover, both variables may affect the deeper environmental attitude of the scientists. This said, for analytical purposes we decided to frame our hypotheses in a more testable manner, remaining aware of the complexity of the actual relationship.

A related topic concerns the factors influencing the dissemination activity of scientists. Dissemination activities are one form of ‘usefulness’ explicitly considered by research funding agencies and policy-makers. On the light of existing literature, it seems likely that reflection and pressure to collaborate with the world outside academia also affects dissemination activities. We hence propose one further hypothesis, where dissemination includes informing media, advising the public or private sectors and taking part in commercialisation activities:

• H₃: The degree of dissemination activities is affected by the extent of the scientists’ reflection on the social and environmental relevance of their research.

Methods

Procedure

The survey was administered online using a prevalent platform (Survey&Report, http://www.artologik.com/en/SurveyAndReport.aspx). The target population included scientists, at all academic levels, affiliated to 79 research groups and centres in environmental science and technology in 19 Swedish universities. More specifically, the target population included (i) research groups in environmental disciplines and (ii) interdisciplinary research groups having an environmental focus. Email addresses were collected from university homepages. In addition, we contacted the group/centre leaders in those cases where websites did not provide information about the involved scientists.

After conducting a pilot test of the questionnaire on a small group of scientists at our university (LNU), a revised version was sent to the larger sample between the middle of October and the middle of November, 2015. A link to a web-based questionnaire was sent to all researchers included in the sample, who had the opportunity to fill the questionnaire anonymously at any time during that period. We sent reminders to the individuals who did not answer after two and four weeks, respectively. The platform we used guaranteed the anonymity of the respondents.

The questionnaire was designed to check our three hypotheses and included seven parts. Questions about background information (discipline, position, etc.) were included in part A and covered variables identified in earlier research (Watermeyer 2014). In part B, respondents ranked five definitions of usefulness derived from previous studies (Stephan and Levin 1992, 2001; Jacobsson and Perez Vico 2010). Parts C–E were based on 5-point Likert-scale questions. In part C, we asked about the respondents’ attitudes towards science and research in general. In part D, the respondents indicated to which degree they agreed or disagreed with several statements about the societal and environmental relevance of research. Part E was designed to measure how reflection on that influences the scientist’s choice of research. Part F measured the amount of dissemination activities performed by the respondents. Part G consisted of two optional open questions focusing on the respondents’ opinions about science, in general, and the questionnaire itself. The full questionnaire is posted as supplementary material.

Limits

We identified some limitations and risks linked to the sampling process. While reaching the entire population was the initial goal, we cannot be certain that we identified all possible research groups and corresponding email addresses. For instance, it is likely that some email addresses were outdated and others were simply not available from the websites. A risk of bias was due to the over-representation of universities having large environmental research groups. Furthermore, response rate of web-based surveys is about 10% lower compared to mail or telephone surveys (Fan and Yan 2010). Finally, the possibility that a higher rate of responses came from scientists interested in the topic of the survey cannot be excluded, which may have artificially increased the recorded interest in making science relevant and the amount of reflection reported.

Results

A total of 307 out of 1037 people completed the questionnaire, with a response rate of 29% covering 18 over the 19 targeted universities. The distribution of academic levels of the respondents included PhD students (39%), professors (16%), researchers and postdocs (16%) and other teaching positions and administrative personnel (29%). In general, the survey results indicated that most researchers have planned or conducted research with partners outside academia and consider this collaboration as an important part of their work. A majority of researchers also reported that they reflect on their role as scientists, deem that science and research should benefit society, and are overall sceptical towards the notion of producing knowledge that solely advances their own discipline. A descriptive analysis for all variables is posted as supplementary material.

First test of the hypotheses

The frequency distribution of the four questions on reflection and choice, included in part E of the questionnaire, is shown in Table 1. In accordance with H₁, a majority of researchers did frequently reflect upon the impact of their research on society and the environment. The distribution of the four variables is similar, with all medians at the ‘frequently’ level. More specifically, 69% of the researchers reflected ‘frequently’ or ‘very frequently’ on the impact of their research on society, while 59% similarly reflected on its impact on the environment. Moreover, they declared that their choice of the research was affected by its impact on society (66%) or the environment (58%).

Table 1. Frequency distribution (%) of selected variables on reflection and choice of research.

	Never	Very rarely	Rarely	Occasionally	Frequently	Very frequently
Reflect on the impact of research on society	0.7	1.0	2.9	25.8	42.8	26.8
Choice of research affected by its impact on society	1.7	3.0	7.6	25.9	40.2	21.6
Think about the environmental impact of research	1.0	1.0	3.0	17.1	41.8	36.2
Choice of research affected by impact on environment	1.7	4.3	7.0	18.5	39.4	29.1

Initial tests of H₂ ascertained whether a correlation exists among the four questions between reflection and choice of the research. All tests showed that these variables were significantly and positively correlated. If researchers considered the impact of their work on society, they also considered its impact on the environment (Spearman’s σ = 0.47, p < .001 two tailed). Notably, reflection on social impact was more strongly correlated with the choice of research leading to a potential societal impact (σ = 0.51, p < .001 two tailed) than with research having an environmental impact (σ = 0.38, p < .001 two tailed). Vice-versa, reflection on environmental impact had a higher correlation with the choice of environmentally relevant research (σ = 0.65, p < .001 two tailed) than with that of research having social importance (σ = 0.51, p < .001 two tailed). In addition, respondents ranked five definitions of research usefulness. The top-ranked definition for most respondents (49.5%) was that useful research should provide findings that can influence policy-makers and industry actors in making decisions that are beneficial for society. These results are consistent with our second hypothesis.

Consistently with the third hypothesis, the reflection on the social impact of research positively and significantly correlated with dissemination activities. The correlation was stronger with the use of social media (σ = 0.24, p < .001 two tailed) and advising activities to the public or the private sector (σ = 0.33, p < .001 two tailed).

Extended models

Although the above correlations provided overall support for all our hypotheses, further analyses were performed to better understand the mechanisms through which different factors affected reflection, choice and dissemination, taking into account the actual complexity of the paths linking the researcher and his or her attitudes to reflection and choice. We estimated two structural equation models using a partial least squares approach, a technique also known as partial least squares path modelling (PLSPM) (Esposito Vinzi et al. 2010; Lohmöller 2013). This technique allows models encompassing complex cause–effect relations and latent constructs, i.e. variables that are not directly observable but can be inferred using the questionnaire items. PLSPM includes two linked parts. First, latent constructs are built through principal component analysis using the manifest variables included in the questionnaire. Each construct is thought to represent a single ‘dimension’ underlying the questionnaire items. Then, a network of relations among these constructs is defined, where links are assumed to represent cause–effects processes. The network is formed by one or more starting nodes (‘independent’ constructs only affecting other nodes), one or more intermediate nodes (construct both affecting and being affected by other nodes) and one or more terminal nodes (constructs affected but not affecting other nodes). Finally, the resulting ‘path coefficients’ are quantitatively estimated by considering the overall network as a system of multiple interconnected linear regressions.

Model 1: choice of the research

The first estimation model tested H₂ by placing the ‘Choice of research’ construct as a terminal node and reflection as intermediate 1. Following our hypotheses, we expected reflection upon social and environmental relevance to have a strong influence on choice. Reflection was assumed to be influenced by attitudes to societal relevance, collaboration and experience. Finally, the attitude to societal relevance was assumed to be influenced by pressure to reach applied results and collaboration. The resulting model is presented in Figure 1.

Figure 1. Overview of the choice of the research model.

The ‘Pressure to produce relevant research’ construct (henceforth pressure) was based on the variables C3 (‘when applying for funding you have to convince funding agencies that you can reach applied results’) and E5 (‘you have promised more applied results than you have provided’). The ‘Attitude to collaboration’ construct (collaboration) was based on the A7 (planned/conducted research in collaboration with partners outside academia) and D5 (importance of a network of collaborators outside the university setting) variables. The ‘Research experience’ construct (experience) was based on the A2 (academic position) and A4 (years of working within research) variables. The ‘Attitude to societal relevance’ construct (soc. relevance) was based on the D1 (importance of results that can be applied in practice), D2 (importance of results that benefits society), D3 (importance of results that can improve the environment) and B1 (research is most useful when it provides findings that can influence policy-makers and industry actors in making decisions that will be beneficial for society) variables. The ‘Reflection’ construct (reflection) was based on the E1 (reflect on the impact of research on society) and E3 (think about environmental impact research) variables. Finally, the ‘Choice of research’ construct (choice) was based on E2 (choice of research affected by potential impact on society) and E4 (choice of research affected by impact on environment) variables. Further details on the analysis are included in the supplementary material.

The PLSPM estimations showed that all paths were significant at the 5% level except the one between experience and choice (p = .84). The strongest total (direct + indirect) effects on choice were those by reflection (0.60), pressure (0.24), soc. relevance (0.24) and collaboration (0.19), the latter construct having only indirect effects. Overall, reflection was an important mediator for most factors affecting choice. The model goodness of fit – i.e. the geometric mean of the average communality of the constructs and the average R² value of the model paths – was 0.38, a value that, although not very high in absolute terms, is in line with what is found in research linking environmental attitudes and behaviour (Susilo et al. 2012; Franzen and Vogl 2013).

Model 2: dissemination

The second model utilised the ‘Dissemination activity’ construct (dissemination) as the terminal node to test H₃. Dissemination was assumed to be affected by three other constructs: reflection, collaboration and experience (Figure 2). Note that, in a previous version of the model, choice was also included but turned out to be not significant and was removed in subsequent analyses. The collaboration, experience and reflection constructs were built as in Model 1. The dissemination construct was based on the F1 (informed media about research), F3 (advised or consulted the public or the private sector) and F4 (took part in commercialisation activities) variables.

Figure 2. Overview of the dissemination model.

All path coefficients in the dissemination model were significant at the 1% level except between experience and reflection (p = .18). The experience construct had the stronger total effect on dissemination (0.53) followed by collaboration (0.32). Unlike the previous model, reflection, although significant in itself, did not seem to act as an important mediator for factors explaining dissemination activities, which were instead directly affected by experience and collaboration. The goodness of fit for the model was 0.41, suggesting the appropriateness of the model construction.

Open-ended questions

An open-ended question at the end of the questionnaire offered the opportunity to freely express comments about ‘how science is conducted and what role scientific research has in society’. Respondents overall showed a strong interest to express their opinions on this: 95 over the 307 respondents provided answers ranging from 71 to 995 characters of length. In the text analysis, we extracted the terms having at least 5 occurrences (37 in total). The top 3 were, as expected, ‘research’ (94 occurrences), ‘science’ (77 occurrences) and ‘society’ (77 occurrences), but the answers also included several terms linked to reflection – ‘responsibility’, ‘important role’, ‘future’ and ‘opinion’ – which passed the 5-occurrences threshold. All terms were mapped into a network following a visualisation of similarities (VOS) algorithm approach using VOSviewer 1.6.3 (http://www.vosviewer.com/). VOS techniques applied to textual analysis select the most relevant words form a ‘corpus’ (in this case the open answers in the questionnaire) and build a network of co-occurrences by linking couples of terms that appear together in the same text. In a second phase, network clustering techniques are used to form larger groups of co-occurring terms (van Eck and Waltman 2009). This resulted in the creation of the four clusters identified by different colours in Figure 3.

Figure 3. Term network clusters. The node size is proportional to the term frequency, the edge size to the number of co-occurrences of the linked nodes.

Answers stressing the role of knowledge and basic science for society concentrated in the first cluster (red in Figure 3), which grouped 14 items with ‘society’ and ‘knowledge’ as most prominent terms. The second cluster (green) grouped 12 items mainly linked to the practical issues of doing research, with ‘research’ and ‘science’ as most prominent terms but also references to research management (‘funding’, ‘time’) and external actors (‘industry’). The third cluster (blue) included 7 items linked with the reflection on the role of researchers, research results and academia. The last cluster (yellow) included 4 items, possibly linked with the societal role of scientists and scientific results.

We also investigated whether ‘basic’ and ‘applied’ science-oriented researchers used different terms in their comments. To do so, we built an ‘applied science orientation’ (ASO) construct using non-linear iterative partial least squares principal component analysis (Wold 1975), the same technique used in the first step of PLSPM estimations (estimation details included as supporting material). The resulting ASO score had mean zero (by construction) and standard deviation 1.5. The higher the score the more applied science-oriented was the corresponding scientists. We then mapped the new variable onto the term network (Figure 4). Applied science-oriented researchers tended to use terms included in cluster 2 and 4, while basic science-oriented researchers used more cluster 1 and 3 terms. In other words, applied scientists seemed to be more interested in discussing issues linked to the cooperation between research and the external world, while basic science researchers stressed the role of knowledge for the sake of society. It is also worth noting that, while ‘science’ was a neutral term (i.e. it was used about as much by basic and applied science-oriented scientists), ‘knowledge’ was mainly used by basic science-oriented scientists and ‘research’ mainly by applied science ones.

Figure 4. Term network with node colour shades proportional to average orientation of the researcher who used the corresponding term: blue = basic science-oriented, red = applied science-oriented.

As shown in Figure 4, there were differences regarding how basic-oriented compared to applied-oriented scientists reflected upon the societal relevance of science. This suggested that the environmental scientists were distributed on a spectrum between basic and applied and that their position on the spectrum affected how they reflected upon the state of science.

It is also interesting to qualitatively consider the open-ended questions in relation to the PLSPM models estimated above, as they clearly illustrate the divide between applied and basic scientists and how both groups reflect on the role of science in society. The basic-oriented scientist quoted below, who had a value of −3.4 in the ASO score described above, recognised the importance of developing science for the sake of the society, not of economy:

Scientific research, first and foremost, should be actionable and beneficial to society. The economic growth associated with it is secondary, in my opinion.

On the other extreme, an applied-oriented scientist (ASO score 1.8) reflected on the challenge of transferring research results to practice:

A problem for both research and society is that it takes (today) too long to get research results into practice. The time must be shortened. The reason for the long time is that scientific journals take too long for publication, and before publication the researchers do not process the result into practice.

Other scientists, like the one below (ASO score −1.0), represented a more moderate view, showing the complexity and balance between these two ideal types of research:

My view is that science has two purposes. First of all it should follow lines of research independent of any influence from society, which is basic research for its own sake. Second, it should provide relevant solutions for societal problems. As a scientist I can of course do both types but not necessarily within the same project. Sometimes the two purposes coincide, but it is probably not the general case.

These quotes serve as illustrations of the way in which scientists manage the two main tasks identified by the respondents: to conduct independent research and at the same time to produce benefits for society and the environment. These quotes indicate that the categories of ‘basic’ and ‘applied’ science do affect how scientists reflect upon the relevance of their research, although basic and applied scientists both reflect upon the social relevance of their work.

Discussion

This study focused on a single discipline, environmental science and technology, which served an interesting case due to the increasing emphasis that research funding organizations appear to be placing on pressing environmental problems. Reflection on the environmental relevance of research seemed here to be connected to a broader reflection on social relevance, and we can hypothesise that different issues may play a similar role in other disciplines. Future studies should hence consider the inclusion of more disciplines, or even consider a large-scale study transcending disciplinary borders. More qualitative research would also be beneficial to better understand the mechanisms through which reflection occurs and translates into actual research choices.

A related question is whether the demand for potential societal benefits restricts academic freedom. Although direct questions on this topic were not included in the survey, the textual analysis showed that scientists did also consider it, although with some differences between applied and basic scientists. Applied scientists reflected more on the meaning of research in cooperation with other actors in society and about the end-product of their research, while basic scientists viewed knowledge in itself as important for society as a whole and appeared more worried about potential limits to the academic freedom.

A surprising result was that, for a majority of the respondents, research is useful because of its societal benefits rather than simply on the basis of the intra-disciplinary advancement of science (part B results). Although it may have been somewhat influenced by the response bias (Section 4.2), this result highlighted how a large share of scientists did feel a strong responsibility towards the rest of the society, thus fulfilling the social contract of science. On the other hand, the answers provided in the open-ended question also illustrated the fact that many researchers consider usefulness too narrowly defined in society today. Pressure from universities and funding agencies that forcefully endorsed ‘useful’ activities (Mobjörk and Linnér 2006) were overall criticised in this study by the same scientists that were prone to reflect on the societal relevance of their research. This may be a lesson for policy-makers, as the growing focus on the application of scientific results and on measuring their impacts risks to backfire if not backed by that part of the scientific community that is already open to non-academic activities and collaborations.

A related issue concerns quality control and how it reflects the current trends in science. The difficulty to anticipate the future societal implications of results should be noted by decision-makers (Nordmann 2014). This also means that measurements of publication and citations need to be complemented with other indicators (Ernø-Kjølhede and Hansson 2011). On the other hand, as highlighted by the answers in our survey, the use of external grants as a measure of success may leave basic science-oriented researchers at a disadvantage, as grants often prioritise applied research and collaboration with non-academic partners.

Conclusion

On the basis of existing literature on the relevance of research (Small, Kushmerick, and Benson 2007; Hessels and van Lente 2008; de Jong, Smit, and van Drooge 2015), we advanced the hypotheses that scientists do reflect upon the societal relevance of their research (H₁) and that this influenced their research choices (H₂) and dissemination activities (H₃). The analysis of the survey data supports all three hypotheses. This said, it is worth repeating that our linearly framed hypotheses may actually conceal the presence of relevant feedbacks between the various constructs.

Part of this complexity was analysed through the PLSPM models, which indeed showed that several individual and contextual factors interact to produce reflection, research choices and dissemination activities. A key finding was that reflection was crucial to explain choices. Both basic and applied scientists reflect upon the relevance of their research and this has become part of their everyday life and impacts what type of research they chose to conduct. This supports the idea of responsible innovation (Owen et al. 2013; Taebi et al. 2014) which appears to have entered the decision-making process of scientists, insofar as societal desirability is accounted for when choosing research topics.

Research experience, in turn, affected reflection but did not have a significant impact on the choice of the research. Collaboration with partners outside academia was instead important for the latter purpose, possibly through its indirect effect on reflection and on the researchers’ attitudes to societal and environmental relevance. Reflection, experience and collaboration were also important in explaining dissemination activities of scientists. Collaboration seemed to spark reflection and to influence attitudes towards societal relevance. A possible mechanism explaining this is that interaction with other sectors and being part of a research network pushes scientists towards reflection over the social impact of their research. This is consistent with the findings that those who collaborate and reflect more also put more effort on dissemination activities and, overall, shows the importance of fostering links between the academia and the rest of the society.

Acknowledgements

We would like to thank the participants in the LNU sociology seminar for their input. Further, we would like to thank Yahya Jani and Flaminio Squazzoni for valuable comments on the paper. Lastly, we extend our gratitude to Michelle Wille for proofreading the paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Joacim Rosenlund is a senior lecturer at Linnaeus University and has a PhD in environmental science. Previous and current research concerns cross-sector collaboration in environmental science. He has a background in sociology from Lund and Gothenburg universities.

Peter Notini worked as a project assistant in the environmental science and engineering group at Linnaeus University. He has a background in both sociology and psychology.

Giangiacomo Bravo is a sociology professor at Linnaeus University, Sweden. His main interests include environmental and computational social sciences. He has published in high-ranked disciplinary and interdisciplinary journals and has participated in international research projects.

ORCID

Giangiacomo Bravo http://orcid.org/0000-0003-2837-0137

Notes

* The research was conducted at Linnaeus University, Sweden.