The importance of future generations and conflict management in conservation

The need to scale‐up conservation initiatives is widely accepted, but understanding how to catalyze the adoption of conservation initiatives remains elusive. To address this challenge, we used diffusion of innovation theory and Best‐Worst Scaling experiments to rank the factors that influence the adoption of Locally Managed Marine Areas (LMMAs) by villages in north Madagascar. The most important driver for respondents to adopt LMMAs was the wellbeing of future generations, while the most important barrier was conflict within and between villages that could arise from the adoption of LMMAs. This emphasis on the benefits and costs of adoption is consistent with diffusion of innovation theory. However, our results indicate that people's intrinsic values (e.g., benevolence and peace) were more important to survey respondents in the adoption of LMMAs than is generally reported in the diffusion of innovation literature. Concerns about conflict from LMMAs and the distribution of livestock incentives warrant further consideration to support the adoption of this conservation initiative in the context of the Sustainable Development Goals. Our study can guide future conservation research and practice to identify the “best” and “worst” attributes of LMMAs and other initiatives to increase the adoption of conservation.


| INTRODUCTION
Environmental degradation has led to detrimental impacts on people, ecosystems and wildlife (IPBES, 2019). Conservation initiatives can help mitigate these detrimental impacts by protecting ecosystems and supporting sustainable development (Field et al., 2012). However, despite over US $20 billion of investments annually in conservation (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) (Waldron et al., 2013), and international conservation targets (United Nations, 1992), conservation initiatives are not being adopted at the required rate (O'Brien, 2012). To address this misalignment, we need to better understand what influences the adoption of effective conservation initiatives by governments, communities and individuals.
Research to date finds a variety of factors can influence people's conservation choices, such as their education (Tayouga & Gagné, 2016) and capacity to engage in conservation (Bennett et al., 2018), or their motivations, including a desire to provide public goods (Admiraal et al., 2017;Farmer, Knapp, Meretsky, Chancellor, & Fischer, 2011). Sustaining fisheries over the long-term has been an important motivation for adopting marine conservation initiatives (Govan, 2009;Jupiter, Cohen, Weeks, Tawake, & Govan, 2014;Ruiz-Mallén, Schunko, Corbera, Rös, & Reyes-García, 2015). The local policy environment can also help or hinder engagement in marine and terrestrial conservation initiatives (Gladkikh, Collazo, Torres-Abreu, Reyes, & Molina, 2020;Rocliffe, Peabody, Samoilys, & Hawkins, 2014). Despite this literature base, there remains insufficient information about what drives people to adopt conservation projects (Admiraal et al., 2017) and an integrated analytical framework for conservation is only just emerging (Bennett et al., 2018). However, a framework to understand the factors that influence decisions to adopt or reject innovations is provided by diffusion of innovation theory (Rogers, 2003). This can be usefully applied to conservation initiatives to improve our understanding of the pace and scale of their adoption by individuals, organizations or communities (Mascia & Mills, 2018;Pannell et al., 2006).
Diffusion of innovation theory (hereafter, diffusion theory) is based on seven decades of research to understand, predict and increase the adoption of a range of technological and policy innovations in different disciplines, from agriculture and health to education and sales marketing (Atkinson, 2007;Balas & Chapman, 2018;Dearing, 2009;Rogers, 2003). Diffusion theory proposes that the characteristics of innovations, and how these interact with the characteristics of adopters, and the wider context in which they exist, play an important role in adoption (Dearing & Cox, 2018;Pannell et al., 2006). Five key characteristics of innovations influence their adoption: relative advantage (e.g., status and income); compatibility (e.g., with values and needs); complexity of the innovation; trialability; and observability (Rogers, 2003;Wisdom, Chor, Hoagwood, & Horwitz, 2014). The importance of these characteristics can depend on their interactions with characteristics of the adopter and context (Pannell et al., 2006). For example, the relative advantage of adopting sustainable agricultural practices depends on the compatibility of the practice with adopters' beliefs and values (Pannell et al., 2006) and social norms (Wisdom et al., 2014).
Values drive people's motivations and influence behaviors (Schwartz, 1996). Most values fall within two basic axes; the first ranges from "self-enhancing" (extrinsic) (e.g., personal success, status and wealth) to "self-transcending" (intrinsic) values (e.g., equality, peace, and protecting the environment) and the second, bisecting axis, ranges from "resistant to change," to "open to change" (Schwartz, 1996). Environmental organizations often use extrinsic values (e.g., financial benefits or status) to increase adoption of conservation initiatives (Crompton & Kasser, 2010), although several authors conclude that intrinsic rather than extrinsic values are key in conservation (Admiraal et al., 2017;Crompton & Kasser, 2010;Farmer et al., 2011). Further, extrinsic values, such as payments for conservation, have been found to "crowd-out" intrinsic values (Agrawal, Chhatre, & Gerber, 2015;Crompton & Kasser, 2010) and can reduce support for pro-environmental behaviors (Rode, G omez-Baggethun, & Krause, 2015). In contrast, others (Bennett et al., 2018;Nilsson et al., 2016) find that both intrinsic and extrinsic values play a role in conservation. Therefore, empirical research is required to understand the influence of values, alongside other factors, in the adoption of conservation initiatives.
To understand what characteristics of conservation initiatives influence adoption, we studied Locally Managed Marine Areas (LMMAs), a community-based marine management tool that has spread relatively rapidly in Madagascar (Rocliffe et al., 2014), with 178 LMMAs established since 1998 (MIHARI, 2020). We used a Best-Worst Scaling (BWS) choice experiment (Louviere, Flynn, & Marley, 2015) to rank the characteristics of LMMAs that were most influential in people's decision to establish LMMAs in four villages in north east Madagascar. Our results develop our understanding of what influences the adoption of these conservation initiatives and highlight how diffusion theory can be effectively applied to conservation. Conservation practitioners can use these results to inform the design of future conservation initiatives to improve the likelihood of adoption.

| Madagascar and locally managed marine areas (LMMAs)
Madagascar is one of the world's poorest nations, with declining wellbeing fuelled by political unrest and climate shocks (Osborne & Hassine, 2016). Half of all children under five are chronically malnourished, with high levels of stunting and disease susceptibility (Bagcchi, 2020). The population is increasing rapidly, particularly along the coast where many are reliant upon the sea for food security and incomes (Le Manach et al., 2012). Pressure on marine resources led to the establishment of Madagascar's first LMMA in Velondriake, in the south-west, and LMMAs are now established nation-wide (Gardner et al., 2020). LMMAs in Madagascar are agreements within or between villages to sustainably manage shared traditional fishing grounds and include an associated management plan, management body, and defined marine area (MIHARI, 2015). LMMAs can be managed by villages, or government agencies can be involved with permits, legal recognition and enforcement. Most LMMAs in Madagascar have been developed with support from NGOs (MIHARI, 2015;Rocliffe et al., 2014).
In addition to restrictions on gear use and protection of habitats such as coral reefs and mangroves, LMMAs include periodic closures to all harvesting activities ( Figure 1). These closures can be for periods of a few weeks to months (Oliver et al., 2015). Immediately after periodic closures are re-opened, harvests improve with increases in octopus sizes, fish abundance and incomes (Oliver et al., 2015). LMMA rules and periodic closures act in addition to national fisheries management rules and periods of no fishing during the windy season. For the purposes of this study, LMMAs are considered an innovation (something new within this context) that is adopted when an LMMA is established. We selected four villages with LMMAs along the north east coast of Madagascar's Diana Region to study between April and June 2018. An NGO had supported each village to establish an LMMA with training, livestock (hens and goats), and financial support. Each LMMA studied was managed by the villages' fisheries associations.

| Survey
The research and survey comprised four elements. First, we used a literature review, discussions with an NGO staff in the United Kingdom and the experience of the research team to generate a list of 33 attributes of innovations that were considered to influence adoption (Table S1). The literature review spanned diffusion theory, including papers from health (Balas & Chapman, 2018), agriculture (Pannell et al., 2006), education (Atkinson, 2007), animal welfare (Hansson & Lagerkvist, 2016), information technology (Moore, 1991) and conservation (Mascia & Mills, 2018). Literature on conservation outcomes, including socioecological impacts (Mascia et al., 2017) and conflict (Bennett et al., 2017), were also reviewed to inform the development of the survey questions and attributes to test.
Second, we designed focus groups to be undertaken in the field to test the list of 33 attributes generated above and identify the most important attributes to research further. We developed the focus group protocol in line with recommendations for low literacy settings (Jakobsen, 2012) and invited male and female community leaders (village representative, council member, vice president or president), and non-leaders, fisher-folk, and non-fisher-folk to attend. Four focus groups were held across two villages, with ten participants in each, with males and females grouped separately to encourage participants to speak freely. Research assistants presented the list of 33 attributes to participants one at a time and asked participants to say whether each attribute was important or not. A simple frequency count was used to identify the attributes considered most important to participants. Two attributes were adjusted based on participant feedback, for example: "The benefit other people or the environment may have from the LMMA" was reworded by focus group participants to: "The benefit to future generations" (Table S1).
The final list of 16 attributes used in the BWS choice experiment (Table 1) included the attributes identified by the focus groups as the most important, ensuring the five key characteristics described in diffusion literature were encompassed (compatibility, complexity, observability, relative advantage, and trialability) because we wanted to test the relevance of this theory (Dearing & Cox, 2018).
Thirdly, we developed a survey instrument to gain information about the villagers and their opinions about the adoption of LMMAs. We then piloted the survey with NGO staff in the field and researchers translated it to F I G U R E 1 Locally Managed Marine Area (LMMA) with a periodic closure to harvesting and fishing (rectangle) marked with buoys (circles). LMMAs are a community-based agreement to establish a management body and plan to manage a defined marine area, such as a village's traditional fishing ground. This may include rules that protect selected species and habitats and ban the use of certain gears, such as small-holed nets and poisons. Typically, LMMAs include areas where all harvesting of any wildlife is prohibited for agreed periods (periodic closures) allowing wildlife to increase in size and number. Periodic closures may cover the whole LMMA and fishing ground, or a portion as depicted by the rectangle in this illustration Malagasy. Ethics approval was secured from Imperial College London prior to the focus groups and piloting of the survey.
The core of the survey was a Case 1 BWS choice experiment (Louviere et al., 2015). In this experiment, we asked respondents: "When your village was deciding whether or not to have an LMMA, which one of these four statements was thought by your village to be the best thing about an LMMA? And which one was thought by your village to be the worst thing about an LMMA?" We provided four attributes in each choice set and asked respondents to select the best and worst attribute from each set ( Figure S1). Each attribute was seen twice across eight choice sets in each survey, and there were eight survey versions, designed with Lighthouse Studio 9.5 (Sawtooth Software, 2017). We used multi-choice and open-box questions to assess respondents' characteristics, including their education, whether they were a village leader, and their views about the LMMA ( Figure S1). For example, we used Likert-style questions to understand respondents' level of involvement in and agreement with the decision to establish the LMMA.
Finally, from April to June 2018, Malagasy researchers undertook key informant interviews with an NGO representative working within the region and village chiefs to understand the context around LMMA establishment and secured permission to undertake surveys from the village chiefs. The researchers used purposive quota sampling in each village with a priori criteria (males, females, leaders, non-leaders, fishers, and nonfishers) to ensure a heterogeneous sample (Robinson, 2013). Snowball sampling was used to identify 30 respondents per village including respondents within each criteria, after Davis et al. (2015) and Peay, Hollin, Fischer, and Bridges (2014). Consenting respondents were interviewed with the questionnaire survey and key informants and focus group participants were not included in the survey sample.

| Analysis
Initial cleaning of the data removed respondents who were not present during the adoption decision. We also undertook a sensitivity analysis to determine the effects of incomplete responses in the choice experiment (Table S2), and eliminated responses with ≤60% of best or worst responses completed.
Initially, we assessed the frequency with which each attribute was selected as best or worst to rank the attributes (Louviere et al., 2015). In contrast to Likert-scale questions, BWS experiments provide two sets of data, best and worst choices, indicating the drivers and barriers to adoption of LMMAs along the same perceptual preference scale (Louviere et al., 2015). The frequency with which an attribute is selected and the consistency with which respondents rate an attribute as either best or worst influences the attribute's ranking on the Best-Worst scale. To assess the significance of any differences in the rankings of attributes along the derived preference scale, we first analyzed best and worst responses separately and then combined, using conditional logit models (Hoffman & Duncan, 1988) in Stata 16. The results provide the statistical difference between each attribute and the attribute selected as the reference case. Across the sample, no respondent selected conflict as "best" in the choice experiment. To facilitate model estimation, we therefore included two best observations for this attribute to two randomly selected choices in the sample. We then assessed combined best and worst responses using a heteroscedastic conditional logit model (HOLE, 2006), which allows for variance between the best and worst choice types. We tested whether best and worst responses could be T A B L E 1 Attributes used in the Best-Worst Scaling choice experiment and abbreviation codes

Attribute used in the BWS choice experiment Code
The amount of money we could earn Money The food we could catch to eat Food

Connections to people beyond the village Connections
The control we would have over our marine resources

Control
The level of conflict in or beyond the village Conflict The monitoring and enforcement of the rules Monitoring How complicated or simple it seemed to establish an LMMA Complexity Someone in the village had seen or heard of LMMAs and their impacts Observability Whether LMMAs can be tested or changed to meet our needs

Trialability
The restriction of certain activities in an LMMA

Restrictions
New habits that might be needed Habits The health of the environment (non-target species)

Conservation
The benefit to future generations Future generations The pressure on the resources due to more people fishing

Resource pressure
The incentives for villages to establish and run an LMMA

Incentives
The funding and training provided by an NGO Funding combined through two log likelihood tests of the best and worst models compared to the combined conditional logit and heteroscedastic conditional logit models. The attributes were mapped across values that motivate behaviors (Schwartz, 1996). For example, attributes relating to conservation and conflict were considered to reflect intrinsic values of "universalism"; while control over resources and financial attributes indicated more extrinsic values of "influence and wealth." Attributes relating to changing habits and connections to others within and beyond the village were considered to relate to respondents' values of "openness to change" as described by Schwartz (1996).

| Survey and sample
Local researchers undertook 25-30 surveys in each of four villages studied (n = 110), and each village had one LMMA. We report results for 88 respondents who met our criteria of being present during the decision-making process about establishing an LMMA and who completed >60% of both the best and worst responses in the BWS questions. Two-thirds of respondents were male (68.2%) and approximately half (52.27%) were leaders in the community (Table S4).

| Best-worst scaling (BWS)
Attributes ranked frequently as best were not ranked frequently as worst, and vice-versa ( Figure S2), permitting the use of BWS frequency analysis to derive a preference scale (Louviere et al., 2015). Conflict and resource pressure were selected most frequently (89.8% and 65.3%, respectively), and consistently ranked as worst when selected (100% and 93.9%, respectively), resulting in highly negative rankings for these attributes. Future generations and conservation were selected respectively 60.0% and 36.9% of the times they occurred, and consistently as best (93.4 and 92.3%, respectively) (Figure 2), resulting in strong positive rankings. Conversely, trialability and complexity were infrequently selected when they occurred in the choice experiment (31.8% and 32.4%, respectively), and were selected as best or worst in similar measure, resulting in rankings close to zero on the preference scale (Tables 2 and 3, and Figures 2 and 3).
Log likelihood ratio tests of: (1) the conditional logit models of best, worst, and combined choices; and (2) the heteroscedastic conditional logit models of combined best and worst choices rejected the combined models. Therefore, we report the results of the conditional logit model for best choices here with the most centrally ranked attribute, trialability, as the reference case (Table 3) as this provides a clear scale of preference of LMMA attributes. The conditional logit model of best responses ranks the benefits to future generations as the best attribute of LMMAs, followed by funding, then resource control, conservation, food, money and enforcement, with significant positive rankings in relation to the reference case, trialability (p < .01). Conflict was ranked significantly worse than any other attribute (p < .01), followed by the pressure on the resource (p < .01). The remaining six attributes were ranked with no significant difference to the central reference attribute. Figure 3 illustrates the positions of the attributes on the preference scale relative to trialability, and the rankings are provided in detail in Table S2. Consistent with previous research (Davis, Burton, & Kragt, 2016), the ranking between the simple frequency count analysis and conditional logit model is similar.

| Adoption-decision
As indicated by their response to Likert-style questions, most respondents felt very involved or somewhat involved in the adoption-decision (71.76%). However, leaders felt more involved in the adoption decision than non-leaders, on a scale of one (I was unaware of the decision) to five (I was very involved in the decision). Males T A B L E 2 Preference scale of attributes; derived from frequency data, weighted by the occurrence of each attribute across all surveys, such that 1.00 would indicate an attribute was chosen on every occurrence (completion rate > 60%, n = 88) Note: Italicization indicates attributes with a significant difference (p < .01) to trialability (in bold). Two best selections were randomly included for the attribute "conflict" to facilitate model estimation, as this attribute was never selected as the best attribute in the choice experiment. and females in leadership roles scored their involvement level as four or five out of five more often (males, 85.71% and females, 90.91%) than did male or female non-leaders (males 47.83% and females 68.75%) ( Figure S6). Overall agreement with the decision to adopt an LMMA was high (91.95%) and most respondents (91.43%) reported that a particular individual or NGO influenced this decision.

| Personal perspectives
When asked in open-ended questions about the decision to establish an LMMA, many respondents said that livestock incentives (chickens and goats) were important in the decision and mentioned that they would like more goats to be provided in the future. Livestock were provided by the NGO to villages establishing LMMAs. However, several people, including fishers, reported disquiet about the way the livestock were distributed and asked that everyone in the village receive livestock incentives, not just fishers. The distribution of livestock incentives was reported as one source of conflict in villages. Participants also reported issues associated with villagers agreeing with and respecting the fishing rules, and negative impacts on village traditions caused by LMMA adoption. Pressure on the resource, which ranked as the second worst aspect of LMMAs in the conditional logit model (Table 3), may have reflected concerns expressed during interviews that the fishing resources had been deteriorating before the LMMA was established, which motivated them to protect the resource.

| DISCUSSION
We found the strongest positive driver for establishing Locally Managed Marine Areas (LMMAs) in our sample was the belief that LMMAs would bring benefits to future generations. Other positive influences included, financial factors, control over resources, conservation and food availability. The strongest opinions overall however consistently related to a negative attribute associated with the adoption of LMMAs: the level of conflict within or between the villages.

| Motivations for adoption of LMMAs
Our findings are broadly consistent with previous research into the motivations of individuals and communities to engage in conservation, and the importance of relative advantages and disadvantages in the decision to adopt within diffusion theory (Dearing & Cox, 2018;Pannell et al., 2006;Rogers, 2003;Wisdom et al., 2014). However, our results emphasize the importance of intrinsic in influencing the decision to adopt LMMAs. The importance to adopters of avoiding conflict, of future generations, and conservation in our study indicates that self-transcendent (intrinsic) values of benevolence (i.e., preservation and enhancement of family, friends and community) and universalism (i.e., tolerance and protection of all people and for nature) (Schwartz, 1996) were of primary importance to respondents. Adopting temporary closures, primarily for the benefit of future generations seems particularly benevolent as fisheries and other harvests for food and income can be constrained during periodic closures (Hall, 2002). Keller (2009) reports that increasing future generations (the offspring of oneself and others) and providing natural resources for them is the essence of a meaningful and moral life in rural Madagascar. Bequeathing benefits to future generations was also a key motivation for establishing LMMAs in Pacific islands (Govan, 2009;Jupiter et al., 2014). Intrinsic values, The preference scale of LMMA attributes, derived from the rankings (coefficients) of each attribute relative to the most centrally ranked attribute, Trialability (in bold), using a conditional logit (CLogit) model of attributes selected as Best. Italics and red shading indicates attributes with a significantly more negative coefficient relative to Trialability. Grey shading indicates attributes with no significant difference in their ranking in relation to Trialability. Green shading indicates attributes with a significantly more positive coefficient relative to Trialability (p < 0.01). Two Best selections were randomly included for the attribute "conflict" to facilitate model estimation, as this attribute was never selected as the Best attribute in the choice experiment. Conflict was the strongest barrier to adoption of LMMAs and Future generations was the strongest driver of adoption of LMMAs in this study including place attachment and providing a public good (Farmer et al., 2011) and having a meaningful life and moral attitudes towards others (Admiraal et al., 2017;Bennett et al., 2018), are key drivers in influencing adoption of conservation initiatives more broadly.
In contrast, the motivations for adoption discussed within diffusion theory, for example, financial rewards, risk reduction, and status (Kuehne et al., 2017;Pannell et al., 2006;Rogers, 2003), are underpinned by selfenhancing (extrinsic) values (Bennett et al., 2018;Crompton & Kasser, 2010;Schwartz, 1996). Self-enhancing values are also found to be significant in this study, through the importance attributed to financial factors, resource control, and food, but of significantly less importance than two of the factors reflecting self-transcending (intrinsic) values of "benevolence" (Tables 2 and 3) (Bennett et al., 2018;Crompton & Kasser, 2010;Schwartz, 1996). Motivations for autonomy and control over local resources were also found to be important in the adoption of LMMAs in the Pacific (Jupiter et al., 2014) and in community-based conservation in Latin America (Ruiz-Mallén et al., 2015). There is less consistency, however, amongst the literature about the relative importance of financial benefits and incentives in the adoption of conservation initiatives. These have been found to be important in Payments for Ecosystem Services (Ruiz-Mallén et al., 2015) and conservation agriculture (Pannell et al., 2006), but were shown to be the least important factor influencing the adoption of conservation easements (Farmer et al., 2011). Therefore, we find that motivations that are consistent with intrinsic values played a greater role in the adoption of this conservation initiative than is typically described in diffusion theory (Kuehne et al., 2017;Pannell et al., 2006;Rogers, 2003;Wisdom et al., 2014).

| Removing barriers to adoption of LMMAs
Our study highlights the need to address the conflict that can be generated by conservation, not only because it was perceived as the most negative characteristic of LMMAs and could therefore reduce adoption of conservation initiatives, but also for ethical reasons (Bennett et al., 2017). The cause of conflict in the villages studied was not always clear, but the distribution of livestock incentives was cited as one cause of conflict during LMMA establishment. Accounting for local perceptions of "fairness" and local power dynamics that may lead to elite capture when distributing incentives are known to be important in influencing attitudes towards conservation initiatives (Randrianarison, Ramiaramanana, & Wätzold, 2017).
In line with a "do no harm" approach, the United Nations (UN) encourages the use of social protection measures to protect food security during initiatives that affect food availability (CFS, 2012), and providing livestock for villages with fisheries closures in LMMAs can be considered as fulfilling this responsibility. While social protection measures (such as providing livestock) are important in conservation, the risk of causing conflict in each sociocultural context needs to be considered (HLPE, 2012). The potential for conservation to cause conflict can often be predicted a priori (Ban et al., 2019;Matulis & Moyer, 2017), and the onus is on practitioners to minimize and manage conflict, with particular attention to protecting vulnerable smallscale fishers (UNFAO, 2015). This can be achieved, for example, by providing stakeholders with access to conflict management (Ostrom, 2009). UN Food and Agricultural Organisation recommend that in support of the Sustainable Development Goals (SDGs), community-based conservation "pay particular attention to issues of gender equality, participation, transparency and accountability in decision-making and the progressive realization of the right to adequate food for all" (UNFAO, 2017).
In addition to livestock incentives being cited as a cause of conflict in our study, incentives can also prime extrinsic values and crowd-out intrinsic motivations for undertaking conservation (Agrawal et al., 2015;Crompton & Kasser, 2010;Randrianarison et al., 2017;Rode et al., 2015). Given that intrinsic values were important to the adoption of LMMAs in our study, it is worth considering whether livestock can be provided in ways that prime intrinsic values and are perceived more positively by adopters, for example as a social protection measure that is implemented equitably with emphasis on protecting vulnerable community members (UNFAO, 2015; UNFAO, 2017).

| Limitations and recommendations for future research
Our study was limited to a relatively small sample size of 88 respondents across four villages in Madagascar; however, an advantage of BWS is that preference estimation is robust to small sample sizes (Lancsar, Louviere, Donaldson, Currie, & Burgess, 2013). As with Davis et al. (2015) and Peay et al. (2014), we used quota and snowball sampling (Robinson, 2013) due to field constraints. This limited our interpretation of results to the sample group rather than permitting generalization to the whole population. Additionally, data collection was undertaken by Malagasy research assistants to minimize the influence of the presence of foreign researchers on respondents, however respondents may have associated the researchers with environmental NGOs and responded accordingly. Fortunately, BWS has been shown to reduce biases such as acquiescence bias, and errors due to cultural or language obstacles commonly found in other survey methods (Lee, Soutar, & Louviere, 2008).
Future research should investigate whether the patterns found here are relevant for LMMAs across Madagascar and other cultural settings internationally, and delve deeper into understanding responses associated with future generations and conflict. Although we identified that "benefits to future generations" influences adoption decisions, our study does not differentiate which benefits for future generations were most important. The causes of conflict in the adoption of conservation initiatives, mechanisms to mitigate conflict, and the most effective management of livestock or other social protection measures, warrant further research. It would also be interesting to investigate whether people considering conservation initiatives linked to their livelihoods, such as in farming and fisheries, respond differently to financial motivations and incentives compared to those considering conservation initiatives that are unrelated to their livelihoods.

| Implications for conservation science and practice
In conclusion, our research found that the most important driver for respondents when adopting LMMAs was the wellbeing of future generations, which relates to intrinsic values. Direct benefits to adopters that relate to extrinsic values, such as food and financial security were of less importance in the decision to adopt. The strongest barrier was conflict that arose within and between villages, at least in part relating to elite capture of livestock provided to support villages adopting LMMAs. This might be addressed through the delivery of UNFAO and High Level Panel of Experts recommendations for conservation to avoid conflict, support food security, and protect vulnerable communities.
We found that while diffusion theory can be applied to conservation initiatives it should be refined to better fit conservation initiatives as people's intrinsic values and motivations may be particularly important in applying diffusion theory to this context.
Methodologically, a benefit of our approach is that a simple BWS count analysis can effectively approximate preference scales without reliance on more computationally complex estimation methods. Conservation practitioners can use BWS to easily identify the factors that are most important to potential adopters and design conservation initiatives accordingly to increase adoption.