https://orcid.org/0009-0004-1285-7826
Figures
Abstract
Sustainability transitions in smallholder agriculture depend on how local communities understand and prioritise different futures, yet most assessment tools rely on externally defined criteria and expert-led evaluations. This study examines how deliberation shapes cocoa farmers’ evaluations of sustainability pathways. In two-day workshops with eight farmer groups (n = 36) in two regions of Ecuador (Amazon and Coast), farmers evaluated four scenarios using intuitive ranking, urgency ratings, and swing-weighting, based on sustainability criteria they had generated in a prior Photovoice process. We compared intuitive and criteria-based rankings of scenarios and calculated indices of individual deliberative reasoning and intersubjective consensus before and after deliberation. In the Amazon, where intuitive scenario preferences diverged, deliberation led all groups to converge on the same scenario rankings and to reweight criteria towards long-term environmental and social concerns. Importance-consensus and individual reasoning indices increased modestly, indicating tighter alignment between what farmers valued and the futures they preferred. In the Coastal region, groups already shared a strong preference for an environmentally restorative scenario. Deliberation primarily reinforced this consensus, substantially increasing agreement on the prioritisation of sustainability criteria while leaving overall scenario rankings unchanged. Across regions, some groups showed closer alignment between intuitive and criteria-based rankings after deliberation, while others diverged, suggesting that deliberation can both stabilise and reconfigure preferences as trade-offs are recognised. The results show that deliberative multicriteria evaluation, when grounded in community-generated criteria and narratives, can articulate and consolidate shared sustainability visions even where prior engagement has already fostered high agreement. The deliberation indices proposed here provide a transferable framework for tracking learning, reasoning, and consensus in participatory planning, supporting the design of more context-sensitive sustainability interventions.
Author summary
How do farming communities decide what a good future looks like? In this study, we worked with 36 cocoa farmers in two regions of Ecuador (one in the Amazon, the other on the Coast) to find out. Over two days, farmers discussed and evaluated four possible futures for their communities, using criteria they themselves had defined in earlier workshops. We measured how these structured discussions changed their preferences and the consistency of their reasoning. In the Amazon, groups that initially disagreed came to a shared vision of the future they wanted; on the Coast, where farmers already agreed, discussion helped clarify “why” they agreed. In both cases, farmers prioritised ecological health and social cohesion over short-term economic gains. Our approach shows that when communities are given appropriate tools and time, they can engage in sophisticated collective deliberation about sustainability. The indices we developed to track changes in reasoning and agreement could be used in other participatory planning contexts.
Citation: Egger M, Curran M, Aubert A, Kearney N, Ventura D, Zambrano Mohauad GA, et al. (2026) Consolidation of shared sustainability visions through deliberative multicriteria evaluation in Ecuadorian cocoa farming communities. PLOS Sustain Transform 5(9): e0000279. https://doi.org/10.1371/journal.pstr.0000279
Editor: Matthew Gibson, London School of Hygiene & Tropical Medicine, UNITED KINGDOM OF GREAT BRITAIN AND NORTHERN IRELAND
Received: February 13, 2026; Accepted: August 3, 2026; Published: September 9, 2026
Copyright: © 2026 Egger et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data and analysis outputs underlying the findings are publicly available from Zenodo (https://doi.org/10.5281/zenodo.18598869), including the criterion weights, scenario performance scores, the ValueDecisions input and output files, and the cross-impact balance matrices and scenario files. Anonymised participant demographic data are provided in the Supporting Information (S3 Table). The audio recordings and full transcripts of the group discussions cannot be shared publicly because they contain potentially identifying information and because participant consent was given for analysis by the research team rather than for public release. De-identified excerpts can be made available to qualified researchers under a data-sharing agreement. Requests may be directed to the corresponding author at egger.moritz@protonmail.com, or to FiBL Research Institute of Organic Agriculture, Department of Food System Sciences, Ackerstrasse 113, CH-5070 Frick, Switzerland (info.suisse@fibl.org), where Dr Michael Curran (michael.curran@fibl.org) acts as the institutional contact for this dataset.
Funding: This work was supported by the Swiss National Science Foundation through the Sinergia programme (grant CRSII5_202300, awarded to Christian Schader). ME, MC, NK, DV, and GAZM received salary support from this grant. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
1 Introduction
Sustainability is a widely invoked yet persistently contested concept. What counts as a “sustainable” pathway differs across sectors, regions and social groups, reflecting diverse value systems, knowledge traditions and power relations [1,2]. The criteria for what counts as “sustainable” vary widely among stakeholders and frameworks [3], reflecting diverse values and place-based perspectives often rooted in power relations [4]. Despite decades of global discourse, from the Brundtland report to countless indicator frameworks, there is still no universal consensus on how to measure or achieve sustainability in practice [5–7]. In agricultural and land-use settings, these tensions are particularly visible as global market forces, local cultural values, and ecological constraints intersect [8]. In the cocoa sector, for example, industry certification schemes and government programs promote standardised sustainability metrics, while farmers often define a “good life” on the land in more relational and place-based terms, linked to security, reciprocity, and dignity [4,9]. Such divergences reflect broader conflicts between export-oriented agribusiness models and local subsistence priorities, as well as tensions between top-down narratives and local situated knowledge [10,11]. In short, what sustainability means and whose sustainability counts remains contested and context-dependent. In this article, sustainability is treated as an umbrella concept spanning environmental, social, and economic dimensions; throughout, we name the specific dimension to which a given claim refers rather than invoking sustainability as a single undifferentiated good.
Moreover, sustainability research has often failed to address the subjective and contextual factors that shape communities’ daily practices [12,13]. Many policy-driven sustainability assessments rely on top-down indicators (e.g., yields, carbon footprint) that overlook social and cultural dimensions such as communal labour, equity, or local risk perceptions [14]. Frameworks that ignore these lived realities risk reducing sustainability to simplistic metrics by overlooking critical elements such as land tenure, cultural identity, and collective governance that people feel connected to, thereby hindering the lasting implementation of recommendations [15,16]. Calls for more inclusive, participatory approaches have grown louder, especially in low- and middle-income countries, where historical power imbalances demand that local voices be heard in defining problems and solutions [13,17,18]. However, turning these calls into practice is challenging. All too often, “participation” is superficial or tokenistic, failing to level power or genuinely integrate local knowledge [19,20]. Bridging the gap between universal sustainability frameworks and everyday experiences requires new methodological strategies that can accommodate fundamental uncertainty, value pluralism, and power dynamics [21,22]. In line with this, researchers have highlighted the promise of participatory multicriteria approaches: combining systematic evaluation with stakeholder deliberation provides a framework for integrating multiple interests and types of knowledge into decision-making [23–25]. Such deliberative methods align with the idea that sustainability transitions should involve not only scientific expertise but also the experiential insights of those living with sustainability challenges [22,26]. These are precisely the conditions that Funtowicz and Ravetz [21] describe as post-normal, namely settings in which facts are uncertain, values are in dispute, stakes are high, and decisions are urgent. Their central methodological claim is that the quality and legitimacy of knowledge under such conditions cannot rest on credentialed expertise alone but require an extended peer community that includes those whose livelihoods depend on the outcome. Read in this light, smallholder farmers are not only sources of data but legitimate evaluators whose situated knowledge is constitutive of what a desirable and feasible future can be. Later assessments of post-normal science have tempered its early claims and emphasised that extended participation must be deliberately structured, so that competing knowledge claims are surfaced and weighed rather than presumed to converge [27].
Within this context, Deliberative Multicriteria Evaluation (DMCE) offers a way to operationalise inclusive sustainability assessment [28,29]. DMCE involves stakeholders in articulating criteria, weighting their relative importance, and evaluating alternative scenarios through discussion and consensus-building [30]. These approaches have been applied to issues such as coastal zone management, ecosystem services and agricultural policy, showing that inclusive scoring and weighting exercises can accommodate heterogeneous values and make trade-offs transparent [31–33]. However, most studies focus on the resulting rankings or policy recommendations. Much less is known about how the deliberative process itself affects participants’ reasoning, the alignment between their intuitive judgements and formal evaluations, or the degree of consensus within and between groups [25,34]. Existing conceptual work on deliberative reason suggests that discussion can either stabilise shared understandings or reveal deeper conflicts in values, but empirical evidence from smallholder farming communities remains scarce.
Deliberative approaches have the potential to reveal how people negotiate trade-offs, such as between economic gains and environmental stewardship, when making collective decisions. However, there is a notable gap in understanding how these deliberative processes influence participants’ perspectives and group outcomes [34]. Do structured discussions and joint evaluations lead to convergence in sustainability priorities? Does the act of deliberation help align intuitive perceptions of what is desirable with more formal, criteria-based judgments? And under what conditions might deliberation fail to resolve differences? These questions remain under-examined in the literature, particularly in smallholder farming communities.
The present study addresses this gap by investigating how cocoa farming communities in Ecuador evaluate sustainable futures through a participatory DMCE process. We build on prior work in these communities, where a transdisciplinary process led to a Photovoice project that enabled farmers to define 60 sustainability criteria in their own terms. That earlier research revealed strong social and cultural priorities in farmer-defined sustainability and showed that regional context (Amazonian vs. Coastal cocoa-farming communities) heavily influenced which themes were emphasised. What remained unclear, however, was how these criteria would be mobilised when farmers jointly assessed alternative future pathways, and how deliberation would influence both their preferences and the coherence of their reasoning.
To explore these questions, we convened a series of two-day workshops in which farmers used a reduced set of 15 Photovoice-derived sustainability criteria to evaluate internally consistent narrative scenarios for the year 2035. Our objectives were fourfold: (1) to analyse how deliberation affected scenario rankings within and across regions; (2) to examine how farmers’ importance weights for sustainability criteria changed through the process; (3) to assess the alignment between intuitive and deliberative evaluations; and (4) to quantify changes in individual reasoning and intersubjective consensus using adapted deliberation indices [34]. By combining visual methods, narrative scenarios, and formal multicriteria evaluation, we aim to demonstrate how an extended peer community of farmers can co-produce context-specific sustainability pathways and how the dynamics of deliberation shape both the content and the coherence of those pathways. The primary contribution of this study is empirical, namely an account of how cocoa farmers in two contrasting Ecuadorian regions mobilise locally generated criteria to evaluate alternative futures. The deliberation indices reported here function as analytical instruments for characterising that process, and their fuller methodological development is reserved for separate work.
2 Methods
A schematic overview of the deliberative multicriteria evaluation (DMCE) process, showing the three phases (workshop preparation, scenario deliberation workshops, post-workshop processing), the activities conducted in each phase, the number of participants and group structure, and the data flows between activities are summarised in Fig 1.
2.1 Study context
Ecuador is the world's largest exporter of fine-flavour cocoa, and cocoa production is a livelihood cornerstone for an estimated 100,000 smallholder families across diverse agro-ecological zones [35,36]. Cocoa production in Ecuador faces a wide range of challenges, including deforestation and climate change-related stress on cocoa trees, as well as contamination of soil and water [36,37]. Furthermore, Ecuador is facing severe impacts from drug-trafficking and resource extraction-related violence, which has eroded the trust in governmental authorities and reshaped many rural areas, including smallholder farmers who unwillingly become targets of criminal operations [38–41].
This study was conducted in two contrasting cocoa-growing regions. The Coastal site, centred on Naranjal in the lowland province of Guayas, represents an intensifying production landscape where conventional monoculture cocoa (predominantly the CCN-51 clone) is dominant, and farmers face persistent pressures from agrochemical dependency, price volatility, and limited state presence. The Amazonian site, centred on Tena in the province of Napo, is characterised by Kichwa-influenced “chakra” agroforestry systems, a traditional practice that integrates cocoa into multi-strata forest gardens, sustaining both livelihoods and cultural identity. These two settings were selected because they encompass much of the socio-ecological heterogeneity of Ecuadorian cocoa farming, the tensions between export-oriented intensification and diversified subsistence, between conventional and agroecological production, and between mestizo and Indigenous knowledge traditions. These contrasting production systems carry distinct ecological signatures. On the Coast, full-sun CCN-51 monoculture depends on high agrochemical inputs and is associated with soil degradation, and cocoa plots adjacent to banana plantations are exposed to pesticide drift that contaminates soil and water [42–44]. In the Amazon, the diversified chakra agroforestry of Nacional fine-flavour cocoa, recognised by the FAO as a Globally Important Agricultural Heritage System, relies on minimal chemical inputs and sustains greater biodiversity, carbon storage, and soil health, while facing localised pressure from expanding alluvial gold mining and the associated mercury pollution of major rivers [37,41,45]. All participating farmers were smallholders cultivating between 0.5 and 15 hectares.
2.2 DMCE workshop preparation
2.2.1 Photovoice-based criteria development.
The initial set of sustainability criteria was developed during a prior one-month Photovoice process involving 43 cocoa farmers from both the Amazonian and Coastal regions of Ecuador. In a series of facilitated workshops, farmers photographed and discussed sustainability challenges and aspirations in their everyday lives, following a classic Photovoice cycle of image production, collective dialogue, and theme identification [46]. Through discussions grounded in their own words, participating farmers identified, defined, and rated 60 distinct evaluation criteria spanning the environmental, social, and economic dimensions of sustainability. These ranged from tangible issues, such as access to technology and forest protection, to more intangible aspirations, such as belonging and identity, or the reduction of corruption.
At the end of the Photovoice workshops, farmers individually prioritised these criteria. Each participating farmer received five voting stickers to distribute across the full set of criteria according to personal preference. This ranking produced quantitative preference scores indicating the relative salience of each criterion within and across regions. These scores were analysed alongside information collected in the Photovoice process regarding the criteria's perceived current status and past development trajectory. This combination of salience, current state and perceived trend provided the empirical basis for identifying the most pressing and widely shared concerns.
To design a manageable yet representative evaluation framework for the Cross-Impact Balance (CIB) and scenario phase, the research team synthesised this information into a reduced set of 15 criteria. The selection was based on farmers’ recorded preferences, qualitative definitions, and regional distributions, with an emphasis on criteria that received high preference scores, were perceived as urgent, or were cross-regionally important. An exploratory Principal Component Analysis (PCA) was also conducted across all criteria based on current status and 5-year development scores given to each criterion (on a 5-point scale from very bad/rapidly worsening to very good/rapidly improving) to check for correlations between criteria to help in selection. Each selected criterion was named and defined using wording as close as possible to farmers’ original expressions, drawing on the Photovoice criteria glossary and group discussion transcripts. For instance, the responsible use of agrochemicals explicitly reflects farmers’ concerns about human and ecological health, while the guarantee of civil security captures anxieties about theft, violence, and weak institutional presence. These definitions were retained throughout scenario development and evaluation to maintain continuity with farmers’ lived experience. This bottom-up framework was cross-checked against external sustainability indicator sets, such as the national cocoa sustainability index by Heredia-R et al. [36], to ensure we were not omitting any critical domain; however, where differences arose, local relevance was prioritised over external completeness. Ultimately, the development of evaluation criteria resulted in a grounded yet structured set that balanced breadth (covering all sustainability dimensions) with focus (limiting the scope to the issues most salient to participating farmers).
2.2.2 Cross-impact balance (CIB) and scenario creation.
To develop internally consistent scenarios, we used Cross-Impact Balance (CIB) analysis [47,48]. From the 15 Photovoice-derived evaluation criteria, we identified 23 driving variables (descriptors) of change in the cocoa sector, each characterised by three qualitatively distinct states (improving, status quo, deteriorating). Two members of the research team (ME and GZ) independently interpreted the Photovoice discussion transcripts to construct an initial cross-impact matrix reflecting the causal relationships articulated by farmers during the group discussions. The two matrices were reconciled into a consensus version, which was then independently evaluated and adapted by six external experts with knowledge of the Ecuadorian cocoa sector. The resulting set of matrices (one farmer-derived and six expert-validated versions) was analysed using the Prophesy software [49], with inter-matrix variation serving as an implicit uncertainty term. See S1 Text for full methodological detail and Zenodo [50] for all cross-impact balance matrices and scenario output files.
From the impact-balanced configurations identified by the CIB, we selected four divergent scenarios representing distinct sustainability trajectories: continuation of the current situation (Scenario A, status quo), an economically focused pathway with environmental degradation (Scenario B), a socially focused pathway with economic trade-offs (Scenario C), and an environmentally focused pathway with social trade-offs (Scenario D). For each scenario, we derived performance profiles on the 15 criteria by averaging the current status score (1–5) with the CIB-predicted future state (1–5). To ensure comparability, total performance for Scenarios B–D was constrained to 60 points across the 15 criteria (mean = 4.00), so that scenarios differed only in how performance was distributed across criteria, not in overall level. The status quo (Scenario A) was unconstrained and reflected current conditions (Coast mean = 2.89; Amazon mean = 3.11), ensuring it would consistently rank below the alternative scenarios. Full performance scores are reported in S2 Table. The 15 farmer-defined criteria and the four scenarios are summarised in Table 1.
2.2.3 Scenario narratives and storyboard visualisation.
To translate the CIB-derived scenario logics into formats meaningful to farmers, we adopted a narrative storytelling approach rooted in local language and lifeworlds. Drawing directly on the earlier Photovoice study and its transcripts, we constructed scenario texts that echoed farmers’ tone, concerns and idioms. Each scenario was written as a first-person account by a central character living in a cocoa-growing community around 2035, reflecting on changes in farming practices, household well-being, community relations and the surrounding environment. The narratives were thus not abstract policy summaries but situated accounts of possible futures in coastal and Amazonian Ecuador.
The visual dimension of the scenarios was built from the corpus of over 3,000 Photovoice images generated in previous project cycles. Rather than using generic stock imagery, we selected and collaged elements from participating farmers’ own photographs to create large storyboard posters for Scenarios B, C and D (Scenario A, the status quo, was not depicted, as farmers already knew this situation intimately). Each collage visually encoded the defining features of its scenario by drawing on recurring motifs and contextual details captured by farmers. For example, Scenario B contained imagery associated with economic aspiration and environmental degradation (e.g., stacked cocoa bags, motorbikes, pesticide bottles), Scenario C emphasised communal gatherings, seed sharing and diverse agroforestry plots, and Scenario D combined images of forest regrowth and mixed farming systems with motifs signalling the erosion of social practices. Images of the scenario storyboard collages are shown in S1 Fig.
The storyboards were printed in large format and displayed during workshops. Farmers were invited to extend on the storyboards after the initial presentation by adding further photo cut-outs or writing comments on coloured sticky notes, thereby co-constructing and personalising the meanings of each scenario. This co-engagement with visual material created a reflective space that connected abstract futures to embodied knowledge. Throughout, we ensured that narratives and visuals remained consistent with the underlying CIB structure and with the relative scenario performances on the 15 criteria. In this way, the narrative-collage approach created an interface between formal scenario modelling and community-based knowledge practices, enabling farmers to engage with complex futures through both analytical reasoning and emotional resonance.
Several features of this procedure limited the risk that the collages biased farmers towards a particular scenario. All storyboards were assembled from the same corpus of farmer-generated images and followed an identical visual grammar, so that no scenario was rendered more professionally or more attractively than the others. The status quo (Scenario A) was deliberately left unillustrated, which avoided privileging any of the change scenarios through visual contrast with a baseline. Farmers were invited to extend and annotate the posters, so the final visual material reflected their own additions rather than the research team’s framing alone. Finally, the intuitive desirability ratings on Day 1 were elicited with the collages present, and the same collages were present again on Day 2; any anchoring effect of the visual material therefore applied equally before and after deliberation and is bracketed by the pre-post design. The illustrative backgrounds of the storyboards were hand-painted for this study, and the photographs collaged onto them were taken by participating farmers. All data-bearing content in the figures consists of participant photographs, and the authors retain full responsibility for the figures.
2.3 Scenario deliberation workshops
Scenario evaluation workshops took place in September and October 2025 with 20 participating farmers in the Amazon and 16 on the Coast. All participants had participated in the prior Photovoice process and were reinvited through local facilitators and cooperative contacts. Participation was voluntary, and dropouts from the original cohort of 43 were due to scheduling, relocation, or other conflicting responsibilities. All farmers were familiar with the co-developed sustainability concepts. Anonymised participant demographic and descriptive data are shown in S1 Table. Most activities were organised in the same small subgroups used during Photovoice, stratified by sex and production system (conventional, organic or chakra agroforestry). The workshops aimed to explore alternative sustainability pathways by evaluating the four contextually relevant scenarios using both intuitive and structured decision-making tools.
On Day 1, each workshop began with a collective reading of the four narratives and familiarisation with the corresponding storyboards (Section 2.2.3). Immediately afterwards, farmers individually rated the desirability of each scenario on a 5-point Likert scale (1 = very undesirable, 5 = very desirable). These intuitive ratings were elicited without group discussion to capture unfiltered individual responses. Subgroups then deliberated on their views and produced a single group rating and rank order of the four scenarios. In parallel, farmers assessed the 15 sustainability criteria. Using individual worksheets, they rated the urgency of change for each criterion on a five-point scale. Short definitions of each criterion, based on farmers’ own wording from the Photovoice phase, were provided on the same sheet to ensure shared understanding. These urgency ratings provided a baseline picture of which aspects of sustainability farmers considered most in need of transformation before collective deliberation.
The final activity on Day 1 introduced an adapted swing-weighting procedure to elicit relative importance weights for the evaluation criteria. Swing weighting is a widely used technique in multicriteria decision analysis for structuring preferences over improvements from worst to best states [51]. Building on Haag et al.’s [52] cards procedure and Aubert et al. [53] criteria descriptions, farmers received printed cards representing the 15 criteria (six social, five economic, four environmental), each including concise definitions and descriptions of “worst” and “best” states by 2035. In subgroups, farmers first ranked the criteria within each dimension using the prompt: “If you could move one of these criteria from the worst to the best state, which would bring the greatest improvement for your community's future?” After agreeing on a ranking, groups distributed cocoa beans across the ranked criteria to indicate importance: the top-ranked criterion could receive up to ten beans, and lower criteria had to receive strictly fewer beans than the one above, with ties allowed only among lower-ranked items. Once weighting within each dimension was completed, groups compared the top criterion from each dimension against the others and repeated the swing-weighting exercise to establish importance trade-offs across dimensions. Standard swing weighting can pose a high cognitive load on participants [24]. Using cocoa beans helps anchor the prioritisation process in the farmers’ everyday agricultural reality. By using a familiar medium central to their livelihood, this approach helps bridge the gap between formal scenario modelling and the farmers’ situated knowledge. Images of the process can be found in S2 Fig.
At the end of Day 1, all evaluation sheets and weighting results were digitised and sent to one of the co-authors in Switzerland (MC), who calculated preliminary multicriteria decision analysis (MCDA) results based on the swing weights and predefined scenario performance scores. Weights were calculated according to von Winterfeldt and Edwards [51], and all weights added up to 1. The weights and performance scores of the criteria across scenarios were linearly transformed to a 0–1 scale and then used as input into the online MCDA software ValueDecisions [54]. The default parameters used a geometric weighted mean, but we also checked the results with the arithmetic mean to verify robustness. The different (4) groups in each region were reflected as separate stakeholders with their respective weighting, to assess the heterogeneity of preferences within each region, given the relatively high diversity in terms of farming systems (i.e., different intensities, gender balance, farm size, etc.; see S1 Table for anonymised demographic and other descriptive data on the participating farmers). The input and output files for ValueDecisions are available via Zenodo [50]. The results of the MCDA were both continuous value scores (from 0/worst to 1/best) and scenario rankings (1/best to 4/worst), which were then used in the workshops on day 2.
On the morning of Day 2, farmers reconvened. To address the potential for cognitive dissonance between holistic values and structured preferences, we utilised a “cake and ingredients” metaphor as a central pedagogical tool. In this metaphorical framework, a farmer's intuitive preference for a sustainability scenario is likened to the “whole cake,” an integrated, experiential judgment of the scenario as a whole. Conversely, the 15 sustainability criteria and the weights assigned via the cocoa-bean exercise represent the “ingredients,” the individual analytical components that logically constitute the final result. By introducing this metaphor, we clarified that while a specific “recipe” (the criteria weights) might lead to a certain outcome, a person's overall impression of the “cake” might differ, and both the intuitive and analytical perspectives are valid expressions of their lifeworlds.
Farmers received visual summaries that juxtaposed (i) their group's intuitive rankings (the “whole cake”) and (ii) the MCDA-based scenario rankings derived from their weighted criteria (the “ingredients”). This visual comparison served as a springboard for facilitated discussions focused on interpreting the divergences and convergences between the two evaluation methods. Farmers were encouraged to reflect deeply on why certain scenarios performed differently when viewed through an analytical lens versus an intuitive one. These reflections explored how complex trade-offs were represented, such as the tension between economic prosperity and environmental protection, and whether the DMCE results felt convincing and authentic to their personal experiences. By creating this “productive distance” between first instincts and considered judgments, the process helped participants either reconcile their holistic feelings with reasoned analysis or identify the specific value-conflicts that made such reconciliation difficult. After this plenary deliberation, subgroups repeated all evaluation activities from Day 1: individual and group intuitive ratings and rankings, urgency ratings for criteria, and swing-weighting of criteria with cards. This pre–post design enabled us to estimate the effects of deliberation on both scenario preferences and underlying priority structures.
2.4 DMCE workshop output processing
2.4.1 Alignment of MCDA and intuitive scenario ranking.
To examine how deliberation affected the consistency between intuitive and criteria-based scenario preferences, we compared, at subgroup level, the rank orders of scenarios obtained from the intuitive exercise with those from the DMCE. For each subgroup and each day, we constructed paired vectors of scenario ranks: (i) intuitive_rank_day1 and intuitive_rank_day2 from the group-level consensus ranking of scenarios A–D, and (ii) DMCE_rank_day1 and DMCE_rank_day2 from the corresponding MCDA rankings based on the group's swing weights and scenario performance scores.
Using these rank vectors, we calculated Kendall's τ and Spearman's ρ as non-parametric measures of rank alignment between intuitive ranking and MCDA for each subgroup on Day 1 (pre-deliberation) and Day 2 (post-deliberation). We then computed, for each subgroup, the change in correlation between days (Δτ = τ_day2 − τ_day1; Δρ = ρ_day2 − ρ_day1). Positive deltas indicate increased convergence between narrative-based intuitive preferences and criteria-based MCDA outcomes after deliberation; negative deltas indicate divergence.
All calculations were carried out in R [55] using the tidyverse suite [56].
2.4.2 Deliberation indices and intersubjective consistency.
We quantified how deliberation affected (i) individual alignment between criteria-based considerations and preferences and (ii) intersubjective consistency of importance ratings, swing weights and scenario preferences. The analyses followed the pre/post structure of the deliberation: farmers first completed an individual intuitive ranking of scenarios and importance ratings of criteria (pre), then deliberated about the results, and finally repeated the elicitation (post). Analyses were stratified by region (Amazon vs. Coast), group and participating farmer, allowing linkage of pre- and post-observations.
For each group and day, we constructed two person × item matrices: one for criteria importance (urgency ratings) and one for intuitive scenario ratings. Each cell contained the numeric response of farmer i to criterion or scenario j. To obtain an intersubjective “consensus” index for a given group, we calculated all pairwise Spearman correlations among farmers across items, separately for the importance of criteria and for intuitive ratings of the scenario. Correlations were transformed into a disagreement metric in [0,1] via (1 − ρ)/2 and averaged across all dyads (the unique pairs of farmers within a group). The consensus index was defined as 1 − the mean disagreement, with 0 indicating maximal disagreement and 1 indicating perfect agreement. This yielded one importance-consensus and one scenario preference-consensus value per group and day.
To assess alignment between criteria-based considerations and scenario preferences, we developed a deliberative reason index (DRI) inspired by deliberative quality metrics proposed in recent work on deliberative reasoning [34]. For each group and day, we again computed dyadic disagreements in criteria importance and in scenario preferences. Each pair of farmers can then be represented as a point in a two-dimensional space with axes “disagreement in criteria ratings” and “disagreement in scenario preferences.” The orthogonal distance of each point to the 45° line y = x (where disagreement in criteria ratings equals disagreement in scenario preferences) captures the extent to which disagreement in scenario preferences can be explained by disagreement in underlying considerations. For each farmer, we averaged this distance over all dyads they belonged to and rescaled it to obtain an individual DRI (DRIInd) in [−1, 1], where higher values indicate stronger alignment between considerations and preferences. Group-level DRI values were computed as the mean of DRI per group and day.
Group swing-weight consensus index was computed from the swing-weighting exercises. For each group and day, we normalised swing weights assigned to non-top criteria so that they summed to 1, calculated the standard deviation of these normalised weights, and transformed it into a consensus index as 1 − SD/SD_max, where SD_max is the maximum SD observed across all groups. Higher values indicate more even weight distributions and thus greater agreement on relative criterion importance.
To evaluate pre- and post-deliberation changes, we treated the group (n = 8) as the paired unit for all consensus indices and as the clustering factor for individual indices. For each region, we calculated pre- and post- means, mean change (post − pre), and paired t-tests and Wilcoxon signed-rank tests for (i) importance consensus, (ii) preference consensus, (iii) swing-weight consensus and (iv) DRIInd (restricted to farmers with both pre and post values). For the consensus indices, we additionally obtained non-parametric 95% confidence intervals for the mean change via bootstrap resampling at the group level (2,000 resamples). As a robustness check for DRIInd, we estimated a linear mixed-effects model with fixed effects for time (post vs. pre) and region, their interaction, and random intercepts for farmers nested within groups, as shown in S6 Table. Finally, to visualise intersubjective patterns, we plotted all dyads in the disagreement space for each region and day, with ellipses summarising the joint distribution before and after deliberation.
2.5 Ethics statement
This study was conducted in accordance with the ethical guidelines of the Research Institute of Organic Agriculture (FiBL), Switzerland. The study obtained institutional ethics approval from the ethics committee for social research of the Department of Food System Sciences at FiBL Switzerland, prior to the start of research activities in Ecuador, in alignment with the guidelines for Research Ethics in Ethnography and Anthropology. All participating farmers provided written informed consent prior to the workshops, which included permission to audio-record group discussions and to use anonymised data for scientific publication. Participants were informed of their right to withdraw from the study at any time without consequences. Photographs used in the Photovoice process and in the scenario storyboards were taken by participants themselves; each farmer reviewed the complete image archive and gave explicit written consent for images to be shared in publications. Where photographs depicted identifiable third parties, separate consent was obtained from those individuals. All data were stored securely, and personal identifiers were removed prior to analysis. Participating farmers received financial compensation for each workshop day to reduce extractive dynamics and recognise the value of their time and knowledge. A blank copy of the participant informed-consent form is provided as S1 File.
3 Results
3.1 DMCE scores
Fig 2 shows the MCDA-derived scenario rankings for each farmer subgroup before and after deliberation. In the Amazon region, Day-1 rankings revealed a common preference for Scenario D in first place, with substantial divergence between groups regarding second- and third-place preferences. For example, TE_G1 ranked Scenario B (economically driven with environmental degradation) as the second most desirable, whereas other groups preferred Scenario C (social with economic degradation). This spread reflects initial differences in priorities, likely shaped by local livelihood contexts and individual predispositions. After structured deliberation, all Amazonian groups converged on the same ranking: D > C > B > A, favouring environmentally and socially sustainable futures over economically dominant or status quo scenarios. In contrast, the coastal groups displayed greater consistency from the outset. On both Day 1 and Day 2, all four groups ranked Scenario D highest and the status quo Scenario A lowest. The slight divergence in NA_G2's ranking of Scenarios B and D on Day 1 persisted, and all groups retained identical ranking orders after deliberation. The absolute MCDA scenario values by group, pre- and post-deliberation are shown in S3 Table.
Each line represents the ranking of the four scenarios (A: status quo; B: economically focused with environmental trade-offs; C: socially focused with economic trade-offs; D: environmentally focused with social trade-offs) by a farmer group in the Coast (NA) and Amazon (TE). Day 1 reflects pre-deliberation rankings; Day 2 reflects post-deliberation rankings derived from weighted performance scores across criteria. Lower ranks (1) indicate higher preference. Names assigned to each scenario are used for illustrative purposes relating to scenario narratives only and do not refer to any actual participants.
Fig 3 illustrates the distributions of criteria weights across the 15 sustainability criteria, disaggregated by group, region and day. In the Amazon, several groups markedly reweighted their priorities after deliberation. TE_G1 and TE_G4, for instance, increased the weight assigned to forest protection (+0.02 for TE_G1 and +0.04 for TE_G4) while down-weighting access to technology (−0.04 for TE_G1) and crop diversity (−0.08 for TE_G4). These shifts align with their move towards Scenario C in the final MCDA rankings for TE_G1 and stronger support for environmentally and socially balanced futures more generally. For Coastal groups, weights also evolved, but changes were more moderate and did not alter the final ranking structure, suggesting a robust underlying consensus. The full dataset, including weights and performance for all criteria per group and day, can be accessed via Zenodo [50].
Bar plots show the relative importance (weights) assigned to each decision criterion by farmer groups in the Coast and Amazon regions. Criteria are colour-coded by dimension: environmental (green), social (yellow), and economic (red). Left panels show initial weights from Day 1; right panels show revised weights from Day 2 following the deliberation workshops.
3.2 Intuitive evaluation scores and rankings
During the intuitive ranking exercise, farmers first evaluated the desirability of the four scenarios through the visual-narrative storyboards. Group members started with individual assessments and then reached consensus on both a 1–5 desirability rating and a rank order of the scenarios. Fig 4 presents the consensus-based desirability ratings for all eight groups before and after deliberation. On the Coast, ratings shifted towards convergence around Scenario D as the most preferred scenario after deliberation. NA_G2 and NA_G4, for example, increased their ratings for Scenario D, while several groups downgraded Scenario B, which had initially been attractive due to its economic promises. This pattern suggests a growing appreciation for scenarios that balance environmental and economic dimensions following group dialogue. In the Amazon, group ratings were more stable. TE_G2, TE_G3, and TE_G4 already favoured Scenario C or D before deliberation and maintained this pattern. In both regions, Scenario A (the status quo) consistently received low scores for most groups, underlining a shared desire for systemic change.
Bar plots show consensus ratings (1 = very undesirable, 5 = very desirable) for each scenario (A–D) before and after deliberation, disaggregated by group. The top panels represent the Amazon region (TE); the bottom panels, the coastal region (NA).
3.3 Alignment between MCDA and intuitive ratings
We compared farmers’ intuitive scenario rankings with their deliberative evaluations generated through MCDA. Several groups show substantial positive shifts in rank alignment after deliberation (i.e., NA_G2, TE_G1, and TE_G2), indicating a closer alignment between intuitive preferences and structured, criteria-informed evaluations (Fig 5, see S4 Table for full alignment metrics). In these cases, deliberation appears to have enhanced the internal consistency of preferences through reflection and collective judgement.
Bar plots show the changes in Kendall's τ (orange) and Spearman's ρ (blue) for each subgroup (n = 8), comparing consensus-intuitive rankings of scenarios with MCDA-based rankings. Positive values indicate improved alignment after deliberation. Coastal farmers are represented with NA groups and Amazonian with TE groups.
In contrast, groups such as NA_G3, TE_G3, and TE_G4 exhibit smaller negative deltas, suggesting a divergence between intuitive and MCDA-based rankings. This likely reflects the introduction of new framings or trade-offs during deliberation that moved final, criteria-based choices away from initial affective inclinations. For instance, TE_G4 intuitively ranked Scenario C highest after deliberation, but their criteria-weighted MCDA placed Scenario D first, reflecting a tension between their experiential preference for social cohesion and the analytical weight they assigned to environmental criteria. Rather than indicating failure, such decoupling may signal that farmers reconsidered their first impressions in light of additional information and group discussion. Groups NA_G1 and NA_G4 show no change in alignment because their MCDA ranks remained identical on both days. Here, deliberation affected intuitive perceptions more than structured evaluations, reinforcing the idea that, in some cases, the underlying trade-off structure was already stable. Alignment metrics and delta values are displayed in S4 Table.
3.4 Deliberation indices and intersubjective consistency
Baseline intersubjective consistency was moderate to high between both regions (Fig 6). At the group level, the mean importance-consensus index in Amazonian groups increased from 0.64 before deliberation to 0.69 afterwards (on a 0–1 scale where 1 indicates perfect agreement; mean change Δ = 0.04; n = 4 groups; paired t-test p = 0.032; Wilcoxon p = 0.1; bootstrap 95% CI: 0.02, 0.07). On the Coast, importance consensus remained essentially unchanged (pre: 0.72; post: 0.72; Δ ≈ 0.00; n = 4 groups; paired t-test p = 0.95; Wilcoxon p = 1.0; bootstrap 95% CI: −0.08, 0.06) (Fig 6b). Thus, deliberation slightly tightened agreement on which criteria matter in Amazonian groups but not in coastal groups.
Panel (a) shows the distribution of the individual deliberative reason index (DRIInd) by region and time (pre/post). Panel (b) shows the group-level consensus index for importance ratings; panel (c) the corresponding consensus index for intuitive scenario ratings; and panel (d) the swing-weight consensus index derived from dispersion of group swing weights. Boxplots summarise the distribution of values, and annotated Δ and p(t) values report the mean pre- and post-difference and the corresponding paired t-test p-value for each region.
Scenario-preference consensus showed larger but statistically less certain improvements (Fig 6c). In the Amazon, the preference-consensus index rose from 0.63 to 0.75 (Δ = 0.12; paired t-test p = 0.35; Wilcoxon p = 0.58; bootstrap 95% CI: −0.09, 0.29), while on the Coast it increased from 0.69 to 0.79 (Δ = 0.10; paired t-test p = 0.19; Wilcoxon p = 0.2; bootstrap 95% CI: −0.02, 0.18). The direction of change is consistently positive, but confidence intervals broadly overlap zero, indicating small average effects relative to between-group variability.
Swing-weight consensus indices revealed regionally differentiated patterns (Fig 6d). Amazonian groups showed a modest, non-significant increase in the evenness of swing weights (pre: 0.11; post: 0.19; Δ = 0.08; paired t-test p = 0.30; Wilcoxon p = 0.58; bootstrap 95% CI: −0.03, 0.19). On the Coast, where baseline swing consensus was higher, the mean index increased from 0.33 to 0.54 (Δ = 0.21; paired t-test p = 0.11; Wilcoxon p = 0.2), and the bootstrap 95% CI for the mean change (0.05, 0.35) lay entirely above zero, suggesting a genuine though small strengthening of agreement on relative criterion weights.
At the individual level, the farmers’ deliberative reason index (DRIInd) also tended to increase, particularly in the Amazon (Fig 6a). In Amazonian groups, the mean DRIInd rose from 0.44 to 0.54 (Δ = 0.10; n = 20 paired participating farmers; paired t-test p = 0.096; Wilcoxon p = 0.086). On the Coast, the mean DRIInd changed only from 0.67 to 0.69 (Δ = 0.02; n = 15; paired t-test p = 0.74; Wilcoxon p = 0.8). Thus, Amazonian farmers showed a trend towards stronger alignment between disagreement in scenario rankings and disagreement in underlying criteria ratings. In contrast, coastal farmers were already at a high level, with little room for improvement. The mixed-effects model, which accounts for repeated measures within groups, yielded similar estimates: a small, positive but not statistically significant fixed effect for “post” and no strong time × region interaction. In sum, Amazonian farmers’ post-deliberation decisions were more closely linked to their stated criteria, while coastal farmers maintained a high level of consistency. All presented values and the results of the mixed-effects model are shown in S5 and S6 Tables.
Dyadic disagreement plots (Fig 7) complement these summaries. Before deliberation, most dyads in both regions were distributed around or slightly above the diagonal of the disagreement space, with moderate disagreement over criteria ratings but greater disagreement over scenario preferences. The post ellipses show that after deliberation, dyads move closer to the origin (less disagreement on both axes) and slightly closer to the diagonal, particularly on the Coast, where the post-ellipse shifts downwards and leftwards and becomes more compact. This pattern suggests that farmers’ preferences and importance judgements become somewhat more mutually consistent across individuals, even if effect sizes remain modest.
Scatterplots of pairwise disagreement in considerations (x-axis; importance ratings) and pairwise disagreement in preferences (y-axis; intuitive scenario ratings) for all participating farmer pairs in the Amazon and Coast regions, before (grey) and after (orange) the deliberative MCDA workshops. Each point represents one pair of farmers; the dashed diagonal indicates equal disagreement in considerations and preferences, and ellipses summarise the main concentration of pairwise distances for each time point.
4 Discussion
4.1 DMCE scores
This study set out to explore whether a structured DMCE process could foster a coherent sustainability vision among cocoa farmers and how it would affect their priorities. The results show that the effect of deliberation on scenario rankings was strongly context-dependent. In the Amazonian region, where initial scenario preferences diverged, the DMCE workshops produced clear convergence: groups moved from heterogeneous rankings to a shared preference for the same sustainability scenarios. This supports the idea that structured dialogue can help participants work through trade-offs and align around a collective vision when starting from disagreement. By contrast, in the coastal region, initial rankings were already homogeneous: nearly all groups ranked the environmentally focused scenario as most desirable and the status-quo scenario as least desirable before deliberation. Here, deliberation primarily reinforced and articulated an existing consensus; groups retained the same ranking order after deliberation, even where underlying reasoning shifted.
Deliberation also affected the criteria that farmers considered most important. In the Amazon, several groups substantially increased the importance they attached to environmental and social criteria and reduced the weight they gave to other more economic aspects, such as access to technology or short-term income. These shifts were consistent with their move toward scenarios emphasising ecological integrity and social cohesion. This result is in line with findings from a deliberative multicriteria study of ecosystem values [24], which observed that while instrumental or economic concerns were frequently discussed in a DMCE setting, the final group consensus shifted towards scenarios focused on environmental and social values. In the Coastal region, changes in weights were more modest and did not overturn the already strong preference for a transformative sustainability scenario. This observation indicates that deliberation isn't just a tool for resolving conflicts between opposing groups, but also a powerful way to increase the precision of consensus within groups that already agree. Taken together, these patterns suggest that DMCE can play different roles depending on the starting point: in more heterogeneous settings, it can reshape both priorities and outcomes, while in more homogeneous settings, it primarily consolidates and refines an already shared direction of preferences.
4.2 Alignment between DMCE and intuitive rankings
Comparing intuitive scenario rankings with DMCE results provides further insight into how the process worked. For some groups, rank correlations between intuitive and criteria-weighted evaluations increased after deliberation, indicating that participating farmers’ feelings became more consistent with what their collective, reasoned analysis suggested was right. In these cases, discussion appears to have helped farmers integrate experiential impressions with explicit arguments about long-term consequences, producing a more coherent preference structure. This is in line with deliberative theories that emphasise the role of argumentation in making preferences more reflective and internally consistent [25,34].
Three out of eight groups reduced the alignment between intuitive and DMCE rankings after deliberation, though with a much smaller effect size than those that increased alignment. We interpret this not as a failure but as evidence that deliberation can also introduce productive distance between “first instincts” and considered judgments. Exposure to alternative framings and new information may have led some farmers to endorse a scenario they did not initially favour, while still feeling ambivalent about it [57,58]. Such divergence suggests that deliberation can increase the salience of latent shared values and prompt critical reflection on how these frames are applied to specific issues, facilitating an adjustment of views rather than a simple confirmation of prior inclinations [59].
4.3 Deliberation indices
The quantitative deliberation indices help to summarise these dynamics. In the Amazon, intersubjective agreement on the importance of criteria increased significantly, from a moderately high baseline to a higher level. Coastal groups started from a higher baseline and showed virtually no change, consistent with an already shared set of priorities. Consensus on scenario preferences also increased in both regions, but the changes were small and not statistically robust, suggesting only a subtle shift toward more shared visions. The swing-weight consensus index, designed to capture agreement in how farmers distributed weights across criteria, showed the strongest increase on the Coast, with bootstrap confidence intervals suggesting a real improvement. Amazonian groups showed a smaller, non-significant rise. Thus, in the Coast, deliberation did not change which scenario was preferred, but helped farmers converge on why it was preferred by aligning the fine-grained pattern of trade-offs and resolving internal cognitive dissonance between values and preference choices.
At the individual level, the deliberative reason index (DRI) increased in the Amazon, while remaining high and almost unchanged on the Coast. Amazonian farmers thus left the workshops with decisions that were more coherent with their own stated criteria, whereas coastal farmers were already highly consistent at the outset. Across indices, effect sizes were modest and conventional significance thresholds were crossed only for Amazonian importance consensus. Nevertheless, the consistent direction of change toward higher alignment and agreement suggests that the DMCE process modestly tightened the logic and mutual intelligibility of farmers’ judgments, in line with previous work on deliberative reasoning indices in citizen forums [34].
4.4 Relation to other studies
Our working hypothesis, grounded in deliberative and participatory decision-making theory, was that deliberation would increase consensus and yield more informed, collectively justified preferences. The Amazonian case clearly supports this expectation: initially divided groups converged on a shared scenario and developed more aligned criteria priorities, illustrating deliberation's capacity to build consensus around sustainability goals. Similar patterns have been reported in participatory planning studies where local stakeholders reshaped adaptation or land-use priorities once their perspectives were explicitly incorporated into multicriteria or scenario exercises [23,32,60].
Our findings also align with the broader literature on participatory DMCE and social multicriteria evaluation, which highlights that these tools can structure complex, value-laden problems while making trade-offs transparent [22,23,28]. We extend this line of work by explicitly measuring how deliberation shifts both the convergence of preferences and the internal coherence between farmers’ criteria and their scenario choices, using indices adapted from the deliberative democracy literature [34]. Furthermore, working with imagery our participants produced themselves, representing their own work, families, and natural environments, led farmers to feel ownership of the futures they were evaluating.
Our approach also speaks to a broader methodological debate about how sustainability is assessed in agriculture. Standardised indicator frameworks, including SAFA, SMART, and the Tool for Agroecology Performance Evaluation, provide comparability across sites but rely on externally fixed indicator sets and expert scoring, which can understate locally salient social and relational concerns [12,26]. Indicator-based assessments of Ecuadorian cocoa farms display the same tendency [36]. By eliciting criteria, weights, and scenario judgements from farmers themselves, the present design retains comparability through its multicriteria structure while keeping the content grounded in local meaning, and it renders the trade-offs between the two visible rather than concealing them within an aggregate score [6]. A parallel can be drawn with research on climate change and health, where quantitative exposure metrics are increasingly complemented by accounts of lived experience in order to capture dimensions that aggregate indicators miss [61]. Deliberative multicriteria evaluation offers one structured way to hold both registers together in the assessment of agricultural sustainability.
Nevertheless, the results of our coastal region nuance the working hypothesis. When a community enters deliberation with a strong, explicit consensus, large changes should not be expected. All groups in this region independently identified similar preferred and rejected scenarios before discussion, likely reflecting a shared experience of chronic insecurity, lack of state support, and environmental degradation that makes transformative scenarios attractive and the status quo untenable. Comparable observations have been made in studies where communities facing similar structural pressures use deliberative forums more to articulate and legitimise existing preferences than to renegotiate them [62]. In such settings, deliberation plays a confirmatory and justificatory role, consolidating agreement and clarifying reasons rather than altering outcomes.
Our case also differs from many one-off participatory experiments in that the groups were not “deliberatively naïve.” Three years prior to the DMCE, all of our participating farmers had engaged in an intensive Photovoice process that involved collectively defining sustainability criteria and publicly debating them. That earlier intervention already fostered social learning, built trust, and generated a locally grounded conceptualisation of sustainability. From a longitudinal perspective, the DMCE should thus be seen as a second step in an ongoing deliberative trajectory rather than a standalone event. The relatively high baseline consensus indices and the strong initial alignment of coastal rankings reflect this history. Participatory research in agriculture has emphasised that social learning is a key benefit of farmer discussions about sustainability [63], which can create “prepared ground” for subsequent deliberation, leading to ceiling effects in measurable change [64].
Interpreted in this light, the small effect sizes in our study are less a sign of limited impact and more an indication that much of the heavy deliberative lifting had already occurred during the Photovoice project and through ongoing community interaction. The DMCE mainly refined and documented a discourse that was already evolving in everyday life. This reading is reinforced by the absence of signs of groupthink or coercive conformity: no group abandoned deeply held views simply to “fit in,” and some individuals maintained intuitive preferences that diverged from MCDA rankings. This pattern aligns with deliberative ideals that stress mutual understanding and reason-giving without erasing pluralism or dissent [65]. It suggests a form of “consensus without uniformity” in which farmers agree on core criteria and broad directions, while retaining nuance and difference in their judgments.
4.5 Limitations and generalizability
Several limitations qualify our findings. First, the study involved 36 farmers from two specific cocoa-growing regions in Ecuador and used purposive recruitment of volunteers who were already willing to participate. These communities have particular agro-ecological conditions and social histories. While we tried to engage farmers who otherwise can not afford to participate in time-consuming workshops through monetary compensation, our sample likely over-represents farmers who are comparatively engaged in community affairs. The results, therefore, cannot be assumed to represent all cocoa farmers, even within the same regions. Comparative work across other cocoa systems and cultural settings would be needed to assess the robustness of these patterns.
Second, the DMCE intervention was layered on top of substantial prior participatory work. The earlier Photovoice process and subsequent community interactions probably increased baseline consensus and familiarity with deliberative practices. While this history is analytically interesting, it makes it difficult to isolate the effect of the DMCE alone. In communities without such a background, deliberation might produce larger shifts or, conversely, encounter more conflict and resistance. Comparative designs that include communities with and without prior engagement would help clarify how starting conditions shape deliberative outcomes.
Third, the DMCE had a 2-day time frame per site. Some attitude and relationship changes may require longer or repeated engagements to mature. We measured only immediate pre- and post-workshop changes; we do not know whether the modest improvements in consensus and reasoning persisted, intensified, or faded over time. Longitudinal follow-up would be needed to examine the durability of deliberative effects.
Finally, as in all quantitative evaluations of social processes, our indices are partial proxies. They capture convergence, weight alignment, and internal consistency, but not all dimensions of deliberation quality, such as emotional resonance, narrative shifts, or subtle power dynamics. More ethnographic or discourse-analytic work would be needed to unpack how gender, age, or status shaped who spoke, who listened, and how compromises were reached.
4.6 Contributions and implications
Despite these limitations, the study makes several contributions to sustainability science and practice. Methodologically, it offers a concrete way to integrate narrative- and image-based participatory methods with formal multicriteria evaluation. By combining Photovoice-derived scenario narratives with both intuitive and deliberative MCDA rankings, we bridged experiential and analytical modes of reasoning. This procedure responds to calls in the MCDA literature for more transparent, reproducible, and stakeholder-centred evaluation frameworks [22,60,66]. Furthermore, adapting the swing-weighting procedure to physical cocoa beans and printed criterion cards rather than digital interfaces demonstrates how formal MCDA tools can be made accessible to smallholder farmers. Finally, the deliberation indices we adapted provide a transferable toolkit for assessing whether participatory processes are merely symbolic or actually changing how people organise their judgments.
Theoretically, the work contributes to understandings of sustainability as a co-constructed, dialogic process rather than a purely expert-driven assessment [6,20,67]. It shows that smallholder farmers can engage in sophisticated trade-off analysis, explicitly link their preferences to criteria, and collectively articulate long-term, future-oriented pathways when given appropriate formats and time. This contribution aligns with transdisciplinary and co-production perspectives that foreground extended peer communities in defining sustainability problems and solutions [21,68]. At the same time, our findings connect with political-ecology insights that local visions of sustainability often challenge dominant productivist or purely economic narratives [69–71]. In our case, farmers prioritised ecological integrity, civil security, and social cohesion over short-term profit maximisation.
In practice, the study offers lessons for policymakers, development practitioners, and community leaders. First, it indicates that well-designed, context-specific deliberative processes can help uncover and consolidate latent community preferences for transformative sustainability pathways. In both regions, farmers clearly rejected the status quo and incremental change, favouring scenarios that entailed substantial environmental restoration and strengthened social infrastructure. Such bottom-up visions can inform public investment and policy design, helping to align interventions with what communities themselves regard as desirable and feasible [72,73]. Second, by grounding planning in locally defined criteria and creating space for reasoned disagreement, participatory DMCE can enhance the perceived legitimacy and salience of resulting decisions and encourage participation in implementation, consistent with evidence from other co-produced environmental initiatives [64,74].
Finally, the approach we developed is scalable and adaptable. It could be used to structure dialogues not only among farmers, but also between farmers, policymakers, scientists and supply-chain actors, thereby opening up decision spaces rather than closing them down [28,75,76]. The methodology and the farmer-defined criteria it yields can serve as a baseline for designing future development projects and orienting public investment toward priorities identified by farming communities themselves. When embedded in longer-term engagement, such processes can contribute to building the social capital, shared language and mutual accountability needed for just and durable sustainability transitions.
5 Conclusions
This study has shown how deliberative, participatory methods can help bridge globally framed sustainability agendas and the lived realities of smallholder cocoa farmers. By enabling farmers in two distinct Ecuadorian regions to define their own sustainability criteria and evaluate future scenarios grounded in their own narratives and images, we documented how they collectively envision desirable futures and negotiate trade-offs. The DMCE workshops revealed that deliberation can transform fragmented views into a shared vision, as in the Amazon case, or consolidate and articulate an existing consensus, as on the Coast. In both contexts, farmers converged on pathways that depart from narrow growth-oriented trajectories and instead prioritise ecological integrity, community cohesion, and civil security.
The analysis of deliberation indices complemented these substantive findings. While effect sizes were modest (explained by farmers’ extensive prior deliberation engagement), the general pattern was toward greater agreement on what matters and clearer internal alignment between criteria and choices. For Amazonian farmers, the process improved the coherence of individual reasoning and increased consensus on importance weights; for coastal farmers, it strengthened an already high level of agreement about trade-offs without erasing diversity in intuitive judgments. Qualitative observations from the workshops, such as greater confidence in expressing views, recognition of others’ perspectives, and emerging leadership roles since the initial photovoice process, underscore that the deliberative process itself constituted an important outcome, building social learning and readiness for collective action.
Looking ahead, our findings point to several avenues for further research and practice. Comparative studies in communities without prior participatory experience could clarify how much deliberation alone can build consensus and whether preparatory capacity-building is a necessary precursor. Longitudinal work would be needed to examine whether shifts in consensus and reasoning endure, and how they interact with changing market and policy conditions. Extending the approach to multi-stakeholder settings (bringing together farmers, cooperatives, government agencies, and cocoa traders) could help identify alignment and friction between levels of governance and support more coherent sustainability strategies. Finally, combining quantitative indices like those used here with ethnographic or discourse-analytic methods would deepen understanding of the micro-dynamics of deliberation, including power relations and the role of affect and narrative.
In sum, the study provides evidence that participatory deliberation is both feasible and fruitful in smallholder farming contexts. It yields contextually grounded, democratically owned sustainability pathways and strengthens the social fabric needed to pursue them. These are precisely the kinds of outcomes that global sustainability frameworks aspire to but often struggle to achieve through top-down implementation. By creating spaces where local and scientific knowledges interact on more equal terms, and by carefully documenting both the processes and their outcomes, deliberative approaches such as the one presented here can contribute to more just, resilient, and legitimate sustainability transitions.
Supporting information
S1 Fig. Scenario storyboard posters displayed during the DMCE workshops (pre-annotation).
https://doi.org/10.1371/journal.pstr.0000279.s001
(TIF)
S2 Fig. Adapted swing-weighting procedure: step-by-step photo documentation.
https://doi.org/10.1371/journal.pstr.0000279.s002
(TIF)
S1 Table. Participant demographic and descriptive data.
https://doi.org/10.1371/journal.pstr.0000279.s003
(DOCX)
S2 Table. Full scenario performance scores by criterion and region.
https://doi.org/10.1371/journal.pstr.0000279.s004
(DOCX)
S3 Table. Absolute MCDA scenario values by group, pre- and post-deliberation.
https://doi.org/10.1371/journal.pstr.0000279.s005
(DOCX)
S4 Table. Alignment metrics between intuitive scenario rankings and MCDA scores before and after deliberation across eight farmer subgroups.
https://doi.org/10.1371/journal.pstr.0000279.s006
(DOCX)
S5 Table. Effect sizes and bootstrap confidence intervals for all deliberation indices.
https://doi.org/10.1371/journal.pstr.0000279.s007
(DOCX)
S6 Table. Mixed-effects model coefficients for the individual deliberative reason index (DRIInd).
https://doi.org/10.1371/journal.pstr.0000279.s008
(DOCX)
S1 Text. Extended cross-impact balance (CIB) methodology and scenario derivation.
https://doi.org/10.1371/journal.pstr.0000279.s009
(DOCX)
S1 File. Participant informed-consent form (blank template).
https://doi.org/10.1371/journal.pstr.0000279.s010
(DOCX)
Acknowledgments
We extend our sincere gratitude to the entire project team of the Deliberative Diets project (www.deliberative-diets.net) for their continuous support and constructive feedback. Special thanks to the entire team at ESPOL University in Guayaquil for their generous hospitality, support, and facilitation of the research activities. We are grateful to the farmers’ associations Maquita and Costa Azul in the Coastal region, and to the cooperative Kallari in the Amazonian region, for their committed participation in this investigation. Above all, our most profound appreciation goes to the local co-investigators, the farmers themselves, for sharing their profound knowledge, insights, and personal experiences regarding the sustainability of cocoa farming in Ecuador. Thus, we gratefully acknowledge the significant contributions from the following co-investigators:
Farmers from the Coast:
Luis Arturo Malavé Constante, Guillermo Llagua Chinachi, Moisés Vega Proaño, Ortencia Alexandra Chinachi Guato, Verónica Ruth Llagua Ortega, Miguel Ángel Fajardo Basurto, Miguel Ramón Murillo Figueroa, Antonio Orlando Sudario Ribera, Henry Joffre Rosado Bravo, Amanda Jéssica Cepeda Andrade, Lidia Bravo Chica, Lorgia Basurto, Ángel Remache, Ruperto Gonzalo Calle Barba, Ángel Sumba Lucero, Vicente Antonio Morán Bueno, Lucía, Jeimmy ASSU, MIBL, Ana Luisa Alarcón Tapia, Naranjaliño.
Farmers from the Amazon:
Tapuy Tapuy Liveya Lizbeth, Rocío, Evelyn, Katherin Andrea, Poveda Tapuy Reny Marcelo, Vargas Tapuy Katia Yadira, Grefa Mamallacta Darío Humberto, Shiguango López Rosa Marlene, Mary Mamallacta, Mamallacta Grefa Carmela Alicia, Mamallacta Grefa Hugo Vicente, Benito Nantipa, Vargas Tapuy Alexis Jhon, Grefa Calapucha Jessica Yadira, Grefa Karla, Grefa Chimbo José Luis, Yumbo Pizango Francisco Raúl, Cirilo, Alex PG, Solano Lloclla Juan Calixto, Palermo, Villamarin Jipa Delfina Mercedes.
References
- 1. Purvis B, Mao Y, Robinson D. Three pillars of sustainability: in search of conceptual origins. Sustain Sci. 2019;14:681–95.
- 2. Rist S, Bottazzi P, Jacobi J. Critical sustainability sciences: intercultural and emancipatory perspectives. 1st ed. London: Routledge. 2023.
- 3. Walsh ML, Thorbjørnsen TL, Jones RC. Framework assessment of sustainability: Methodology for evaluating the conformance of fishery and aquaculture management systems to FAO guidelines. PLOS Sustain Transform. 2024;3(9):e0000124.
- 4. Altieri MA. Agroecology: the science of natural resource management for poor farmers in marginal environments. Agriculture, Ecosystems & Environment. 2002;93(1–3):1–24.
- 5. Brunori G, Galli F, Barjolle D, Van Broekhuizen R, Colombo L, Giampietro M, et al. Are Local Food Chains More Sustainable than Global Food Chains? Considerations for Assessment. Sustainability. 2016;8(5):449.
- 6. Sala S, Ciuffo B, Nijkamp P. A systemic framework for sustainability assessment. Ecological Economics. 2015;119:314–25.
- 7. Whyte P, Lamberton G. Conceptualising Sustainability Using a Cognitive Mapping Method. Sustainability. 2020;12(5):1977.
- 8. Janker J, Mann S, Rist S. What is Sustainable Agriculture? Critical Analysis of the International Political Discourse. Sustainability. 2018;10(12):4707.
- 9. Krauss JE. Representing environment and development – tracing links between drivers, representations and power dynamics in cocoa sustainability and beyond. Journal of Political Ecology. 2018;25(1).
- 10. Fraser EDG, Dougill AJ, Mabee WE, Reed M, McAlpine P. Bottom up and top down: analysis of participatory processes for sustainability indicator identification as a pathway to community empowerment and sustainable environmental management. J Environ Manage. 2006;78(2):114–27. pmid:16095806
- 11. Scott JC. The moral economy of the peasant: rebellion and subsistence in Southeast Asia. New Haven, CT: Yale University Press. 1977.
- 12. Schader C, Grenz J, Meier MS, Stolze M. Scope and precision of sustainability assessment approaches to food systems. E&S. 2014;19(3).
- 13. Sterling EJ, Filardi C, Toomey A, Sigouin A, Betley E, Gazit N, et al. Biocultural approaches to well-being and sustainability indicators across scales. Nat Ecol Evol. 2017;1(12):1798–806. pmid:29062123
- 14. Eizenberg E, Jabareen Y. Social sustainability: A new conceptual framework. Sustainability. 2017;9:68.
- 15. Binder CR, Feola G, Steinberger JK. Considering the normative, systemic and procedural dimensions in indicator-based sustainability assessments in agriculture. Environmental Impact Assessment Review. 2010;30(2):71–81.
- 16. Grund J, Singer-Brodowski M, Büssing AG. Emotions and transformative learning for sustainability: a systematic review. Sustain Sci. 2024;19:307–24.
- 17. Gaventa J. Finding the Spaces for Change: A Power Analysis. IDS Bulletin. 2006;37(6):23–33.
- 18. Masterson VA, Mahajan SL, Tengö M. Photovoice for mobilizing insights on human well-being in complex social-ecological systems: case studies from Kenya and South Africa. E&S. 2018;23(3).
- 19. Arnstein SR. A Ladder Of Citizen Participation. Journal of the American Institute of Planners. 1969;35(4):216–24.
- 20. Blackstock KL, Kelly GJ, Horsey BL. Developing and applying a framework to evaluate participatory research for sustainability. Ecological Economics. 2007;60(4):726–42.
- 21. Funtowicz SO, Ravetz JR. Uncertainty, complexity and post-normal science. Environmental Toxicology and Chemistry. 1994;13(12):1881–5.
- 22. Stirling A. Analysis, participation and power: justification and closure in participatory multi-criteria analysis. Land Use Policy. 2006;23(1):95–107.
- 23. Garmendia E, Gamboa G. Weighting social preferences in participatory multi-criteria evaluations: A case study on sustainable natural resource management. Ecological Economics. 2012;84:110–20.
- 24. Mavrommati G, Borsuk ME, Kreiley AI, Larosee C, Rogers S, Burford K, et al. A methodological framework for understanding shared social values in deliberative valuation. Ecological Economics. 2021;190:107185.
- 25. Niemeyer S, Veri F, Dryzek JS, Bächtiger A. How Deliberation Happens: Enabling Deliberative Reason. Am Polit Sci Rev. 2024;118:345–62.
- 26. De Olde EM, Moller H, Marchand F, McDowell RW, MacLeod CJ, Sautier M, et al. When experts disagree: the need to rethink indicator selection for assessing sustainability of agriculture. Environ Dev Sustain. 2017;19:1327–42.
- 27. Turnpenny J, Jones M, Lorenzoni I. Where Now for Post-Normal Science?: A Critical Review of its Development, Definitions, and Uses. Science, Technology, & Human Values. 2010;36(3):287–306.
- 28. Burgess J, Stirling A, Clark J, Davies G, Eames M, Staley K, et al. Deliberative mapping: a novel analytic-deliberative methodology to support contested science-policy decisions. Public Underst Sci. 2007;16(3):299–322.
- 29. Proctor W, Drechsler M. Deliberative Multicriteria Evaluation. Environ Plann C Gov Policy. 2006;24:169–90.
- 30. Dean M. Including multiple perspectives in participatory multi-criteria analysis: A framework for investigation. Evaluation. 2022;28(4):505–39.
- 31. Aubert AH, Lienert J, von Helversen B. Gamified environmental multi‐criteria decision analysis: information on objectives and range insensitivity bias. Int Trans Operational Res. 2022;30(6):3738–70.
- 32. Garmendia E, Gamboa G, Franco J, Garmendia JM, Liria P, Olazabal M. Social multi-criteria evaluation as a decision support tool for integrated coastal zone management. Ocean & Coastal Management. 2010;53(7):385–403.
- 33. Mavrommati G, Borsuk ME, Howarth RB. A novel deliberative multicriteria evaluation approach to ecosystem service valuation. E&S. 2017;22(2).
- 34. Niemeyer S, Veri F. Deliberative Reason Index. In: Ercan SA, Asenbaum H, Curato N, Mendonça RF. Research Methods in Deliberative Democracy. 1st ed. Oxford University Press. 2022. 99–114.
- 35. Charry A, Perea C, Ramírez K, Zambrano G, Yovera F, Santos A, et al. The bittersweet economics of different cacao production systems in Colombia, Ecuador and Peru. Agricultural Systems. 2025;224:104235.
- 36. Heredia-R M, Blanco-Gutiérrez I, Esteve P, Puhl L, Morales-Opazo C. Assessment of sustainability in cocoa farms in Ecuador: application of a multidimensional indicator-based framework. International Journal of Agricultural Sustainability. 2024;22(1).
- 37.
Ramírez K, Zambrano G, Santos A, Charry A, Chávez E. Análisis de la cadena de valor de cacao del Ecuador y percepciones sobre la regulación europea de cadmio en chocolates y otros derivados del cacao y el cambio climático. Publicación CIAT. 2022. https://cgspace.cgiar.org/bitstream/10568/122788/3/report.pdf
- 38. Freeman W. Can Ecuador avoid becoming a narco-state?. Current History. 2024;123:56–62.
- 39. Rivera-Rhon RA, Bravo-Grijalva CE. Gobernanzas criminales y enclaves productivos de la minería ilegal en Ecuador. RLCT. 2023;15:49–69.
- 40. Rivera-Rhon RA, Bravo-Grijalva C. Crimen organizado y cadenas de valor: el ascenso estratégico del Ecuador en la economía del narcotráfico. Urvio. 2020;:8–24.
- 41.
Amazon Watch. Gold, Gangs, and Governance: Indigenous Communities in the Grip of Organized Crime. 2024. https://amazonwatch.org/assets/files/2024-09-gold-gangs-and-governance.pdf
- 42. Avadí A. Environmental assessment of the Ecuadorian cocoa value chain with statistics-based LCA. Int J Life Cycle Assess. 2023;28(11):1495–515.
- 43. Guillen M, Calderon J, Espinoza F, Ayol L. Aerial Spraying and Its Impacts on Human Health in Banana-Growing Areas of Ecuador. Healthcare (Basel). 2024;12(20):2052. pmid:39451467
- 44. Zamora-Acosta G, León-Hernández E. La agroindustria del banano en el litoral sur ecuatoriano: una mirada crítica desde la coremática. Revista de Investigación TALENTOS. 2021;8:62–83.
- 45.
FAO. Amazonian Chakra - Globally Important Agricultural Heritage Systems. GIAHS. 2023. https://www.fao.org/giahs/giahs-around-the-world/ecuador-amazonian-chakra/en
- 46. Wang C, Burris MA. Photovoice: concept, methodology, and use for participatory needs assessment. Health Educ Behav. 1997;24:369–87.
- 47. Weimer-Jehle W. Cross-impact balances: A system-theoretical approach to cross-impact analysis. Technological Forecasting and Social Change. 2006;73(4):334–61.
- 48. Weimer-Jehle W. Einführung in die Cross-Impact-Bilanzanalyse (CIB): Wege zur qualitativen System- und Szenarioanalyse. Wiesbaden: Springer Fachmedien Wiesbaden. 2023.
- 49. Kearney NM. Prophesy. Zenodo. 2025.
- 50. Egger M. Data and outputs for: consolidation of shared visions for sustainability scenarios through deliberative multicriteria evaluation in Ecuadorian cocoa farming communities. Zenodo. 2026.
- 51.
von Winterfeldt D, Edwards W. Decision analysis and behavioral research. Cambridge: Cambridge Univ. Press. 1993.
- 52. Haag F, Reichert P, Maurer M, Lienert J. Integrating uncertainty of preferences and predictions in decision models: An application to regional wastewater planning. J Environ Manage. 2019;252:109652. pmid:31629174
- 53. Aubert AH, Schmid S, Beutler P, Lienert J. Innovative online survey about sustainable wastewater management: What young Swiss citizens know and value. Environmental Science & Policy. 2022;137:323–35.
- 54. Haag F, Aubert AH, Lienert J. ValueDecisions, a web app to support decisions with conflicting objectives, multiple stakeholders, and uncertainty. Environmental Modelling & Software. 2022;150:105361.
- 55.
R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. 2025.
- 56. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R, et al. Welcome to the Tidyverse. JOSS. 2019;4:1686.
- 57. Druckman JN, Nelson KR. Framing and deliberation: how citizens’ conversations limit elite influence. American Journal of Political Science. 2003;47:729–45.
- 58. Heong KL, Escalada MM. Perception Change in Rice Pest Management: A Case Study of Farmers’ Evaluation of Conflict Information. Journal of Applied Communications. 1997;81(2).
- 59. Zimmermann A, Albers N, Kenter JO. Deliberating Our Frames: How Members of Multi-Stakeholder Initiatives Use Shared Frames to Tackle Within-Frame Conflicts Over Sustainability Issues. J Bus Ethics. 2021;178(3):757–82.
- 60. Huerta E, Kampichler C, Ochoa-Gaona S, De Jong B, Hernandez-Daumas S, Geissen V. A multi-criteria index for ecological evaluation of tropical agriculture in southeastern Mexico. PLoS One. 2014;9(11):e112493. pmid:25405980
- 61. Desai Z, Zhang Y. Climate change and women’s health: A scoping review. GeoHealth. 2021;5:e2021GH000386.
- 62. Dryzek JS, Braithwaite V. On the Prospects for Democratic Deliberation: Values Analysis Applied to AustralianPolitics. Political Psychology. 2000;21(2):241–66.
- 63. Stringer LC, Dougill AJ, Fraser E, Hubacek K, Prell C, Reed MS. Unpacking “Participation” in the Adaptive Management of Social–ecological Systems: a Critical Review. E&S. 2006;11(2).
- 64. Reed MS, Evely AC, Cundill G, Fazey I, Glass J, Laing A. What is social learning?. Ecol Soc. 2010;15:resp1.
- 65.
Mansbridge J, Bohman J, Chambers S, Christiano T, Fung A, Parkinson J. A systemic approach to deliberative democracy. Cambridge University Press. 2012.
- 66. Macharis C, de Witte A, Ampe J. The multi‐actor, multi‐criteria analysis methodology (MAMCA) for the evaluation of transport projects: Theory and practice. J of Advced Transportation. 2009;43(2):183–202.
- 67. Ely A. Transformative pathways to sustainability: Learning across disciplines, cultures and context. Taylor & Francis. 2022.
- 68. Cockburn J. Knowledge integration in transdisciplinary sustainability science: Tools from applied critical realism. Sustainable Development. 2021;30(2):358–74.
- 69. Altieri MA, Toledo VM. The agroecological revolution in Latin America: rescuing nature, ensuring food sovereignty and empowering peasants. Journal of Peasant Studies. 2011;38(3):587–612.
- 70.
Escobar A. Designs for the pluriverse: Radical interdependence, autonomy, and the making of worlds. Duke University Press. 2018.
- 71. Martinez-Alier J. The environmentalism of the poor. Geoforum. 2014;54:239–41.
- 72. Fourat E, Blanchart E, Cué Rio M, Darias MJ, Diedhiou A, Droy I, et al. Holistic approaches to assess the sustainability of food systems in low- and middle-income countries: A scoping review. PLOS Sustain Transform. 2024;3(7):e0000117.
- 73. Theis S. From local to global: Leveraging localized environmental actions for scalable sustainability effects. PLOS Sustain Transform. 2025;4(7):e0000185.
- 74. Cash D, Clark WC, Alcock F, Dickson NM, Eckley N, Jäger J. Salience, credibility, legitimacy and boundaries: linking research, assessment and decision making. Rochester, NY: Social Science Research Network. 2002.
- 75. Macharis C, Turcksin L, Lebeau K. Multi actor multi criteria analysis (MAMCA) as a tool to support sustainable decisions: State of use. Decision Support Systems. 2012;54(1):610–20.
- 76.
Stirling A. Precaution, foresight and sustainability: Reflection and reflexivity in the governance of science and technology. 2006.