Mechanisms for enhancing public engagement with citizen science results

1. Citizen science is frequently cited as a successful approach for increasing pub lic engagement with environmental issues, but this requires a purposeful design that is inclusive of, and responsive to, diverse interests. This paper explores the mechanisms for improving participant and public engagement with citizen science results, using the New Zealand Garden Bird Survey (NZGBS) as a case study. It in vestigates how citizen science can apply democratic processes to be more respon -sive, while drawing on insights from behaviour change frameworks to facilitate a purposeful design. 2. By inviting NZGBS participants to select, inform and peer review the design and promotion of new resources, our goal was to embed their values, opinions and per spectives into the developments. This not only empowered 15,844 respondents to contribute directly to the citizen science initiative's governance over 6 years, but also made it more engaging and useful to them and the wider public. 3. New resources were a sense of collective action, making them attractive,


| INTRODUC TI ON
Citizen science is the practice of public participation and collaboration in scientific research to increase scientific knowledge (Bonney et al., 2015). It is frequently cited as a successful approach for raising awareness, knowledge, engagement and action on environmental issues (Bonney et al., 2009Cooper, 2018;Dickson et al., 2012;Schultz, 2011), with the implementation of purposeful design that is inclusive of, and responsive to, diverse interests proposed as an approach to drive such success (Arnstein, 1969;Conrad & Hilchey, 2011;Shirk et al., 2012).
For purposeful design, behaviour change frameworks and tools have been developed that draw on insights from behavioural economics, social psychology and sociology (Behavioural Insights Team, 2014;Darnton & Horne, 2013;Reddy et al., 2017). These recognise that changing behaviour is a complex process influenced not only by the individual but also by their enabling environment (Park, 2020). Their recommendations include targeting not just the individual (taking into consideration, e.g. their motivations, agency and skills) but also their social realm (being cognisant of the influence of social norms and meanings in relation to their social networks, roles and relationships) and their material context (encompassing their access to, and interactions with infrastructure, technologies, time and schedules).
For responsiveness, using democratic processes to drive 'participatory governance' is one approach, enabling citizen scientists to add their values, opinions and perspectives to the decision-making for such endeavours (Reed, 2008). Achieving good outcomes from such processes requires careful consideration of who is participating, how they communicate and make decisions, and the extent of their influence over the resulting decisions and actions (Fung, 2015). Such processes also need to have a clear intent and pathway to satisfying that intent, with its outcomes being meaningful to the participants. If successful, citizen scientists will not only come to support and engage with the institutions and participation practices involved, but also defend any efforts to reduce participation (Irwin, 2006;Rowe & Frewer, 2000).
Recent advances in the development of web-based tools for collecting and visualising citizen science data are widely credited with enabling democratic processes and aiding engagement (e.g. Sullivan et al., 2014). Such tools hold potential to empower people affected by local environmental issues to access local data and make informed decisions , and thus may be a powerful approach for applying democratic processes that enhance broader public engagement in citizen science beyond simply increasing participation (Rowe & Frewer, 2000). However, recognising the challenges of implementing good democratic processes, Irwin (2006) recommends approaches are critically reviewed to evaluate the lesson learnt from operationalising them in different contexts. In addition, the impact of 'democratising' citizen science is still to be assessed (Conrad & Hilchey, 2011;Dickson et al., 2012) while, similar to conservation F I G U R E 1 An overview of our democratic process aiming to empower citizen scientists to influence the governance of the NZ Garden Bird Survey. Specifically, citizen scientists were initially invited (in 2014 and 2015) to inform the design of new brand and results resources, as well as a communication strategy for promoting them (via the pathways indicated by the solid black arrows); they were then invited (from 2016 to 2019) to peer review the new results resources and our communication strategies (via the pathways indicated by the grey dashed arrows) to help us to progressively refine them management and policy (Baylis et al., 2016), citizen science itself still needs to undergo robust impact evaluation to demonstrate the benefits it delivers.
Here we provide such review and assessment, using the New Zealand Garden Bird Survey (NZGBS; Spurr, 2012) as a case study in which web-based tools were designed and employed with the goal of enabling such democratisation processes to increase engagement with the results of the initiative through four key steps (Figure 1): (a) giving citizen scientists participating in the NZGBS a voice in designing and peer reviewing the new resources; (b) designing brand and results resources that create a sense of collective action, are attractive, easy to understand and promote as well as tailored for audiences with different levels of expertise and online communication channels; (c) promoting the new results resources via multiple online channels to reach and engage as wide and diverse an audiences as possible; and (d) evaluating the level, nature and predictors of engagement, aiming to progressively improve our resources and promotional campaigns. In the discussion we critically review our democratic process, highlighting the value added from integrating behaviour change considerations into its design and implementation (Reddy et al., 2017), and assess the impact of the actions employed on engagement with initiative results. Reflecting on the lessons learnt, we highlight the opportunities and challenges for other citizen science initiatives wanting to achieve similar results, while also identifying areas to target for further improvements in NZGBS engagement.

| ME THODS
Social ethics approval for our research process was secured from the NZGBS's host organisation, Manaaki Whenua-Landcare Research, which is New Zealand's national terrestrial environmental research institute (Application numbers: 1314-27-007; 1819/22). Participation in our online surveys was voluntary and through informed consent.

| Giving citizen scientists a voice
Citizen scientists were invited to contribute to three online surveys: These surveys, administered through the SurveyMonkey platform, took up to 10 min to complete, with responses to all questions being optional. The logo survey was publicised during the 2015 NZGBS campaign on the NZGBS webpage, on Facebook, in an email sent to previous NZGBS participants and a media release. The other surveys were publicised to any citizen scientists who participated in the NZGBS and entered their data online by transferring them directly to the respective survey's introductory page (except in 2016 when they were directed to a thank you page, which included a survey invitation and link), which stated its intent-to improve their NZGBS experience in the future.
To understand who was taking part, respondents were asked to select which (if any) of a list of five to eight statements about environmental interests and roles applied to them (Question 1, Table 1). For the logo survey, respondents were also asked if they were of Māori (New Zealand's Indigenous people) descent, their age group (<15, 15-20, 21-35, 31-50, 51-65, >65 years old) and gender (male, female or prefer not to answer). For the peer review survey, respondents were also asked their gender, age (<18, 18-19, 20-29, 30-39, 40-49, 50-59, 60-69, ≥70 years old) and ethnicity (European/Pākehā, Māori, Pacific peoples, Asian, Middle Eastern/ Latin American/African, other ethnicity) as well as their NZGBS participation history (with four options: first time, once before, a few times or many times). For the design and peer review surveys, participants were also given the option of providing their email address as a form of identification (on the understanding that this information would not be shared with others).

| Logo survey
Participants were asked to first rank eight candidate logo concepts  in order of preference and then to explain why they selected their most and least preferred options.
An open box was provided for any additional comments. A summary of logo competition results was published on the NZGBS webpage.

| Design survey
Participants were asked to indicate their level of interest (not interested, neutral, interested, very interested) for each of the six metrics listed for each of the following questions (Table 1)

| Peer review survey
In 2016, participants were asked to indicate their level of awareness and/or interest (not aware, neutral, interested, very interested) in the NZGBS data visualisation resources (species factsheets and videos). From 2017 to 2019, participants were asked if they were aware of the latest State of NZ Garden Bird report and, if the response was positive, to indicate their level of interest (aware of it but not looked at it, looked at it but not really interested, looked at it and very interested, looked at it and loved it) for different resources The logo was officially launched in the 2016 NZGBS campaign. The graphics were also loaded to an online datastore to make them readily accessible and secure their legacy (Supporting Information A).

| Results resources
A range of resources were designed and tailored for promoting the NZGBS results to audiences with different levels of expertise and via a range of media channels ( and three spatial scales (giving priority to national and regional scales, but also providing some information for suburbs). This report included: the top 10 infographic, species factsheets, species videos and an interactive atlas (Table 2). From 2017 to 2019, the emphasis shifted to reporting trends from national to local scales for at least 14 common garden bird species; these 'State of NZ Garden Birds' (SOGB) resources included maps and barplots, which were also collated into national and regional summary reports, all designed for audiences with low to medium expertise (Table 2). In some years, interactive maps and technical reports, designed with audiences with medium to high expertise in mind, were also published to provide more detailed statistics for multiple spatial scales.
Six key tenets for growing engagement guided resource design.
First, building the NZGBS profile and creating a sense of collective action (Darnton & Horne, 2013)

| Promoting resources
The annual reports were published on the NZGBS webpage (Supporting Information E). From 2017 onwards, high-quality versions of these 'pre-canned' graphics and reports were also available via an online datastore, where third parties could view, download or directly embed them on their own online channels (Supporting Information F); this was not only to secure the resources' legacy, but also to empower others to actively promote them among their networks and tailor their reporting according to their needs. In parallel with the progressive improvement and diversification of report resources each year (Table 2), the content, presentation and layout of the associated webpages were likewise refined, aiming to make these resources easier to use and more attractive for the user.
The results resources were also promoted each year via three

| Engagement metrics
Engagement metrics, which were used as the indices of engagement, were downloaded from each of the respective communication channels (Table 3

SOGB 2018
Top 10 infographic A matrix of bird icons showing changes in the top ten species counted for each year (inspired by a similar one used by the Malaysian Garden Bird Survey); it was annually updated Species videos A light-hearted video to convey (within a minute) an interesting fact about the bird, how many gardens in rural and urban landscapes would need to be visited to find it, whether feeding birds in gardens will influence the chance of seeing the species and the regions in which the bird was most likely found. It finished by encouraging people to visit the NZGBS webpage to learn more. A playlist of the 16 videos was loaded to YouTube and individual videos were also embedded on the NZGBS webpage 16

Species factsheets
Included the species' icon and video, a fun fact about the bird, its names (Latin, common and Māori), conservation status, feeding habitats and food sources, a simple, colourful bar plot showing how the species counts varied nationally in relation to garden type (urban vs. rural) and feeding activities (fed vs. unfed) and a map illustrated variation in counts across 16 regions 16 Interactive applications Enabled the user to explore how garden bird counts (NZGBS atlas) or trends (NZGBS explorer) vary across suburbs within each region, using three tabs to: (a) introduce the NZGBS and the application; (b) display bird maps, using dropdown menus to select the spatial resolution, species and/or location and spatial scale of interest, with a pop-up window to show the 80% confidence intervals for a given location's count or trend estimate and (c) survey effort map showing information for each suburb 1 1 1

Species maps
Designed as standalone resources, these simple, eye-catching graphics aimed to report a high-level summary of national and regional results. See Supporting Information C for refinements over time 16 16 16 246 Barplots Generated for individual species at the regional scale, or for all focal species at a given scale (local, regional, national); with a te reo Māori translation available for the national scale. They provided more detail than the maps, while still being simple, eye-catching and standalone; see Supporting Information D for how graphics were refined over time 16 33 206 425

Summary reports
Designed for a non-specialist audiences and were available for national and, in later years, regional scales. The report collated relevant subsets of existing bar plots and/or species maps (described above), with additional text, illustrations and graphics to outline the report goals, analytical approach (including the criteria used to draw attention to trends of concern or interest), number of garden surveys per year and volunteer effort. As above, presentation was refined over time 1 17 17 Technical reports Designed for audiences with some basic understanding of statistics, these national and regional reports included an outline of their goal, and the analytical approach and criteria used to draw attention to trends of concern or interest. Graphics As a function of the total page views for the respective set of report resources. b Binary variable for link shared at least once via email (Email link) or social media (Post link): true As a function of the total number of page views for the respective report resource.
d As a function of the total number of opened emails.

TA B L E 3 (Continued)
that it was not possible to access all the original articles in 2019 so engagement in this year may be underestimated. For each email, engagement was measured as the cumulative number of clicks for each resource link up to the download date (as it was not feasible to specify the observation period of interest). For online resources (NZGBS webpage and datastore), engagement was measured based only on activity associated with the latest report's webpages within the 4-week period immediately following its respective press release.

| Predictors of engagement
A series of models were fitted to identify key predictors of engagement with social media and online resources, as well as determinants of resource awareness and interest of different citizen scientist audiences (based on the design and peer review survey results; Table 3 Generalised linear mixed effects models were fitted in most cases, using the glmer function from the lme4 package (Bates et al., 2015) in R (R Core Team, 2020), where the response variable was binary or count data specifying as appropriate binomial or Poisson errors respectively (Table 3). Most models controlled for repeated measures associated with the campaign year (specified as a factor), different channels on social media and email identity for individual link clicks. When modelling awareness and interest of reports, respondent identity was specified as a random effect; this factor accounted for multiple responses from a subset of individuals over time (as identified by those who opted to provide their email addresses) as well as multiple answers to different questions from same individual within a campaign year (for the interests models only). For the latter models, convergence was added by specifying the 'bobyqa' optimiser and in some cases with 100,000 iterations.
The full model was fitted in each case, and a stepwise backwards selection process (using the drop1 function specifying a chi-squared test in R) was then applied to identify the minimum adequate model (MAM). Fitted estimates were then derived for the subset of variables retained in each MAM (using the predictorEffects function from the effects package in r; Fox, 2003;Fox & Weisberg, 2018.

| Citizen scientist numbers and composition
The logo survey was completed by 695 people (with most probably being previous NZGBS participants, as there were 659 survey-link clicks directly from an email sent to them). Three-quarters were >30 years old and female; 10% was under 15 years old and 6% was of Maori descent (Supporting Information J). Compared to the design survey, respondents were more likely to be students, and work in environmental management or biological sciences (Supporting Information K).
The design survey received 4,418 responses, equivalent to c.

| Citizen scientist design preferences
Interest was high for all six of the listed bird metrics (i.e. >90% of respondents were interested or very interested in each metric), with the highest number of responses and interest levels (98%) being for the bird trend and the most common or rare species options (Supporting Information M). The spatial areas of most interest (≥89%) were national, regional and town/city, with lower levels of interest (68%-82%) for respondents' own neighbourhoods or gardens.
Infographics (≥76%) were preferred over tables, written summaries or reports as the format for presenting NZGBS results

| Online resources
Individual webpages were more likely to be viewed if they occupied a shallow tier within the nested website and promoted via at least one email link, with a similar pattern detected for webpage entrances but for which the impact of email or social media promotion was more evident ( Figure 5; Supporting Information Q). Resource type was the strongest predictor of webpage engagement overall (Supporting Information Q). Compared to technical reports, the 'Story so far' resources were much more likely to be viewed, particularly the top 10 infographic, but less likely to be an entry point; for the SOGB resources, the national summary was the most popular (for both views and entrances), closely followed (in descending order) by the report subpages, regional reports, barplots and, to a lesser extent, maps, all of which were also regular entry points.
Email-link entrance probabilities for the webpages were highest for the technical report (c. 70%), closely followed by regional reports, report homepage and species factsheet ones, but only a quarter or less of those for the national summary or report subpages ( type, with the most popular (in descending order) being the national summaries, report directories (homepage or subpages) and regional reports ( Figure 6); link type, where text and image hyperlinks were at least twice as likely as subtitle ones to be clicked; and an accompanying image, which reduced the likelihood of a link click.

| Peer review survey
Awareness of results resources was most strongly predicted by participation history, but also varied over time (increasing by c. 25% over [2016][2017][2018][2019] and in relation to the citizen scientists' age group, roles and interests (Supporting Information S). Awareness was positively associated with previous NZGBS experience (almost quadrupling for those who had participated many times compared to first timers) and, to a lesser extent, age (Figure 7). Participants who identified themselves as environmental society members, working in the biological sciences or environmental management or as citizen scientists were also more likely (in increasing order) to be aware of the resources.
For participants aware of the results resources, interest probabilities were most strongly predicted by resource type and campaign year, with a threefold gain from 2016 to 2019 (Figure 8; Supporting  Interest was higher for participants who identified themselves as citizen scientists, teachers or working in environmental management. Following the release of the new results resources, interest in barplots and maps increased (compared to interest levels stated in the earlier design survey; Figure 9, Supporting Information T) but decreased slightly for the summary reports (cf. the original 'summaries with text and graphics' category). Interest also increased for garden society and, to a lesser extent, environmental society members (but dropped slightly for students) as well as those who did not select these three role or interest categories. Interest levels were unchanged for citizen scientists, those working in environmental management or biological sciences or who did not select any of the role or interest categories.

| D ISCUSS I ON
Our goal was to embed citizen scientists' values, opinions and perspectives into a citizen science initiative through a democratic process (Cooper, 2018); this was achieved by inviting NZGBS participants to help select a logo and to inform and peer review the design of new results resources and an associated promotional strategy ( Figure 1). This not only empowered participants to contribute directly to the initiative's governance (Shirk et al., 2012), but also made it more engaging and useful to them and the wider public (Dickson et al., 2012), while creating a sense of collective action (Darnton & Horne, 2013). Over 6 years, NZGBS participants contributed to our process via 15,844 responses to our online surveys (Figure 2).
In Table 4, we summarise how our approach for involving NZGBS participants met nine evaluation criteria for a robust participation process (Rowe & Frewer, 2000). Below, we highlight the opportunities and challenges in implementing our democratic process, and the value added from integrating key behaviour change insights into that process, to deliver good outcomes (legitimacy, effective governance and social justice; Fung, 2015).

| Participation composition
As the composition of participants for the primary NZGBS event is unknown, and participation in our online surveys was determined by self-selection, it is not possible to determine whether respondents to our online surveys were in fact representative. However, the response rate for the design and peer review surveys was high (≥60% in all but one year), with participant numbers increasing and diversifying over time, encompassing a broad range of roles, interests, ages, genders, ethnicities and NZGBS participation histories (Figure 2; Representative criterion, Table 4).
The high response rate to these surveys was likely driven, at least in part, by their presentation, setting and framing. This TA B L E 4 An evaluation of our democratic process in relation to five criteria determining whether it was acceptable to the public and four criteria assessing whether it took place in an effective manner (Rowe & Frewer, 2000) Component Criterion How our process met the criterion Acceptance Representativeness: Public participants should comprise a broadly representative sample of the population of the affected public Numbers of people taking part in our online surveys increased and diversified over time, encompassing a broad range of roles, interests, ages, genders, ethnicities and NZGBS participation histories (Figure 2). The response rate for the design and peer review surveys, which were open to anyone submitting their NZGBS data online, was high (≥60% in all but one year). However, despite making an open-call for the logo survey, responses were likely biased towards previous NZGBS participants, as there were a high number of survey-link clicks from the email sent to them. In all cases, however, they were self-selected Independence: The participation process should be conducted in an independent, unbiased way The online surveys were designed by the NZGBS's host organisation, so in that respect were not independent. However, as this paper aims to show these data were reviewed using rigorous methods designed to be objective (including inviting peer review of the surveys designs and seeking social ethics approval for their application) Early involvement: The public should be involved as early as possible in the process as soon as value judgements become salient Participant input was invited from the outset to inform the development of new brand and results resources for the NZGBS. The form and structure of this scientific citizenship was shaped by our survey design, which enabled dual engagement pathways: structured responses allowed for a fast, standardised evaluation process, which were then cross-checked in relation to the unstructured responses (via open box comments), thus providing a richer understanding of the impact of the engagement process and its value Influence: The output of the procedure should have genuine impact on policy The results derived from initial design and logo surveys directed the new resource and media campaign developments, with subsequent ones being used to progressively improve Transparency: The process should be transparent so that the public can see what is going on and how decisions are being made A high-level summary of the results from the first design survey and the logo one were reported via the NZGBS website, with subsequent research report summarising the findings of the design and peer review surveys for 2014-2018 published online in 2018, but not actively promoted. Instead, priority was given to delivering to the new brand and results resources and the associated communication strategy and its progressive improvements. Addressing this shortcoming is one motivation for this paper

Process
Resource accessibility: Public participants should have access to the appropriate resources to enable them to successfully fulfil their brief NZGBS participants were invited to take part in the design and peer review surveys immediately after submitting the NZGBS data online, ensuring that they had access to the resources they needed to take part; the anticipated time taken to complete the online survey was stated upfront and all questions were optional, allowing the user to adjust their input in relation to the time they had available. In 2018 and 2019, we added a link to the latest NZGBS report to direct respondents to the resources in case they had missed them Task definition: The nature and scope of the participation task should be clearly defined The purpose of the online surveys was stated on the front page, making it clear that its aim was to improve their NZGBS experience in the future Structured decision-making: The participation exercise should use/provide appropriate mechanisms for structuring and displaying the decision-making process This paper aims to summarise the process used to develop the report resources and evaluate and improve engagement with them by as wide and diverse an audiences as possible Cost-effectiveness: The procedure should in some sense be cost-effective The online surveys were designed to minimise the time investment required by participants but at the same trying to ensure that the process was also as inclusive as possible 'choice architecture' design purposefully aimed to minimise the time and effort required to participate (Behavioural Insights Team, 2014;Darnton & Horne, 2013;Park, 2020), applying a number of 'nudges' to increase the likelihood of participation (Reddy et al., 2017): giving participants a timely prompt and making it easy by seamlessly transferring them to the surveys once they entered their NZGBS data online, while still giving them the choice of whether to take part or not. The seamless transfer, for example, removed a small friction cost (or seemingly trivial point of hassle; Park, 2020) to increase the participation rate by at least 30% (as evidenced by the drop in participation in 2016 when a separate survey invitation and link was presented on an interim thank you page).
Despite making an open-call for the logo survey, responses were likely biased towards previous NZGBS participants, as there were a high number of survey-link clicks from the email sent to them, suggesting that link also acted as a nudge.

| Communication and decision-making
Citizen scientist input was invited early in the development process (thus meeting the Early Involvement Criterion; Table 4). However, as the form and structure of this scientific citizenship was shaped by our survey designs, it is important to acknowledge that we held the professional power to set the structural constraints on the engagement possibilities and consider its implications (Fung, 2015;Irwin, 2006 Actively involving the public and NZGBS participants in selecting a logo, for example, gave them ownership from the outset (Darnton & Horne, 2013). Other benefits of this process were ensuring that the winning logo would have broad appeal, while also meeting our design requirements for a simple, clean, memorable, timeless graphic (Behavioural Insights team, 2014) that leveraged the links between the public and an iconic, charismatic garden bird (Smith et al., 2020). This was validated through a review of participants' comments, which expressed a range of values, emotions, tastes and attitudes (Darnton & Horne, 2013); for example, some (7%) emphasised their appreciation for the artists and their logo concepts, while others (2%) contemplated the importance of having an accurate depiction of bird in the logo (vs. a cartoon) or specifically signalled (5%) their support, delight, pride, excitement and gratitude in being part of the selection process and the NZGBS. This feedback was valuable not only for cross-checking that the structured approach met participants needs and interests (e.g. the winning logo was in fact an accurate depiction of the bird), but also indirectly confirming that this survey successfully initiated a sense of collective action (e.g. participants valued being part of the logo selection process, having their voices heard and being part of the wider NZGBS initiative).
Overall, our research developments ( which are expected to be more closely aligned to those of the general public (Fung, 2015), and thus offset the influence of institutional powers which historically would have had a more dominant voice (i.e. institutions directly or indirectly involved in the NZGBS governance by way of their NZGBS hosting or partnership roles; Irwin, 2006). In effect, this has enhanced the legitimacy of the NZGBS's governance (Fung, 2015), starting to shift it from a contributory model (where a project is designed by scientists, with the public contributing data) to a more collaborative one (also designed by scientists but with the public contributing data, refining the project, analysing data and disseminating findings; Shirk et al., 2012).
Our decision process for prioritising the results' resource developments could have been strengthened overall if we had asked participants to provide more information about themselves in the initial design survey (e.g. age, gender and previous NZGBS experience). However, at the outset, we deliberately minimised the length of the design survey to avoid overburdening participants, being cognisant of the fact that they had already contributed their bird count data, while also being uncertain about their appetite for engaging with our democratic process. However, having access to that more detailed information would have enabled us to understand participant subgroups' interests and needs from the start (Smith et al., 2020) and thus weight priorities and tailor more effective engagement strategies to reach, communicate and motivate key groups (Donovan & Henley, 2010;Scheufele, 2018). In effect, this would have allowed for a more robust evaluation of whether our project also delivered social justice-that is ensuring that our governance mechanism did not advantage some groups over others (Fung, 2015).

| Influence and empowerment
The NZGBS development met its intent-to deliver resources and promotional campaigns that incorporated and reflected citizen scientist values, interests and needs to successfully build public and participant engagement with results resources (Figure 3)-and in doing so empowered participants to contribute to the governance of a citizen science initiative. Engagement levels increased for social media posts, media outlets and online resources (webpage, datastore and email). The depth of that engagement was also enhanced, with more people actively investigating, discussing and promoting the results resources. Overall, this signals a willing, mutual and beneficial exchange with NZGBS participants and the public, thus meeting the requirements of effective social marketing campaigns (Smith et al., 2020). In effect, our project therefore delivered effective governance (Fung, 2015).
Overall, the key elements shown by analysis to improve engagement were those hypothesised by the behaviour change framework, a combination of individual, social and material factors (Figures 4-9).
By making our materials attractive, easy to understand and readily accessible (Behavioural Insights Team, 2014), we motivated and empowered others to promote them. For example, we successfully nudged media outlets to tailor their reporting for local scales and promote our key motivational message (e.g. 'birds as backyard barometers') and high-quality 'pre-canned' graphics (by including these or links to them in a media release). This is significant because media outlets can act as key influencers, helping build engagement with citizen science results as a social norm, thereby encouraging others to adhere to it (Darnton & Horne, 2013;Park, 2020). Similarly, social media posts were shared by more numerous and diverse institutes, who can also act as key influencers, helping to increase resource visibility and salience, thus shaping engagement as a social norm (Smith et al., 2020). However, as individuals are most likely to influence other people within their own social networks or peer groups, their increased sharing of social media posts was a key achievement (Clayton et al., 2013;Park, 2020).
Valuable insights were gained from the peer review surveys; these signalled that, despite our progress in building public engagement with the results resources (Figure 2), there is still plenty of scope for improving the NZGBS campaign strategies to reach and engage wider and more diverse audiences. Roughly only half of NZGBS participants were aware of the results resources and 80% of them were interested (Supporting Information S), with different suites of factors (relating to audiences composition) predicting resource awareness and interest (Figures 7-8). This highlights the complexities associated with implementing successful behavioural interventions (Reddy et al., 2017), with further refinements required to better tailor and target campaigns to understand, reach and engage key groups (Smith et al., 2020) and ensure that some are not advantaged at the expense of others (to facilitate the good governance goal of social justice; Fung, 2015). Encouragingly, however, as we adapted our campaigns in response to our engagement evaluations and the peer review surveys, awareness of and interest in results resources also improved between 2016 and 2019.
Furthermore, compared to the original design surveys, interest levels were overall either unchanged or improved in the peer review ones (Figure 9).

| CON CLUS IONS
Citizen science is considered a mechanism for democratising sciencefacilitating the public and science to work more closely together, thereby empowering the public to engage actively in dialogue and decisionmaking around environmental issues (Bonney et al., 2015;Irwin, 1995).
Citizen scientists are thus 'people exercising their rights and responsibilities to participate in collective scientific endeavours' (Cooper, 2018); as citizen science stakeholders, they can be affected by or can directly affect decisions about the governance of a citizen science initiative (Reed, 2008). However, the degree to which citizen scientists participate in the governance of such endeavours will influence the extent to which they can add their values, opinions and perspectives to decision-making (Reed, 2008).
Substantial gains in logic, innovation and invention can be achieved from collective intelligence, where people with diverse backgrounds and skills work together towards shared goals (Woolley et al., 2010). In our case, applying a democratic process to drive participatory governance demonstrated how citizen scientist perspectives can be elevated, to offset the influence of institutional powers which would otherwise have a dominant voice, while also confronting the challenges of mitigating its influence in shaping the engagement process (Fung, 2015;Irwin, 2006). Working together to better meet the needs and interests of citizen scientists and the public successfully built and diversified the audiences engaging with citizen science results.
Integrating behaviour change insights in our purposeful design was a critical factor contributing the success of our initiative, helping us to understand and address the influence of individual, social and material contextual factors on engagement behaviour (Behavioural Insights Team, 2014; Darnton & Horne, 2013;Park, 2020;Reddy et al., 2017). While the democratic process was vital for identifying the diverse needs and interest of individuals, the behaviour change frameworks were valuable in understanding the role of choice architecture, social networks and key influencers, and their complex interactions, in creating the enabling environment required to influence engagement. Adding to this complexity is the different timescales that these factors operate.
Overall, this highlights four key challenges for other citizen science initiatives wanting to make similar gains. The first challenge is the magnitude of effort and resource required to make a real-world difference in terms of engagement, not just with previous participants but also the wider public. Second is the strategic budgeting of the available effort and resourcing that is needed to have biggest impact, a complexity that is generally under-appreciation.
Third is recognising that the data visualisation and communication processes needed to make real difference requires progressively building a broad range of skills and capability for learning how to build engagement. Fourth is acknowledging that climbing the social engagement ladder takes time, initially aiming to build awareness of the citizen science results and their importance, then using those resources to build user confidence and capability to engage and eventually actively participate in the citizen science initiative itself (Schultz, 2011), and going beyond that to inform, grow awareness and engage the wider public.

ACK N OWLED G EM ENTS
We thank Eric Spurr (Manaaki Whenua), who founded and organ- and an anonymous reviewer for their constructive comments.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

AUTH O R S ' CO NTR I B UTI O N S
C.J.M. and K.S. conceived the ideas, designed the methodology and collected the data; C.J.M. analysed the data and led the writing of the manuscript. All the authors contributed critically to the drafts and gave final approval for publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data presented or analysed in this paper are available via the