GWAMA: A Web-Based Decision Support Tool for Greenwashing Risk Assessment in Sustainable Food Marketing

Abstract

Greenwashing in food marketing undermines consumer trust and impedes Sustainable Development Goal 12 (SDG 12). While prior research has established linkages between greenwashing perception, green skepticism, and purchase intention, no publicly deployed decision support tool has been developed for practitioner use. This study applies Design Science Research (DSR) methodology to translate validated behavioral models into a deployable decision support system rather than re-testing established relationships. We present the development, deployment, and evaluation of the Greenwashing Advertising Message Assessment (GWAMA), a web-based DSR artifact grounded in a validated Stimulus–Organism–Response (S-O-R) structural equation model. GWAMA integrates factor-loading-weighted composite scoring with SEM-derived parameters to generate real-time greenwashing risk diagnostics for food advertising messages. Usability was evaluated with 150 Thai food industry professionals using a Technology Acceptance Model (TAM) instrument applied to the live system. Results provide indicative evidence of stakeholder acceptance, with high perceived usefulness, ease of use, and intention to use. This study contributes by demonstrating how validated behavioral models can be translated into a publicly deployable decision support artifact, with practical implications for sustainable marketing governance and SDG 12 implementation in emerging economies.

1. Introduction

Environmental marketing claims in the food industry have expanded rapidly, often outpacing mechanisms for verification and regulatory oversight. As companies increasingly use sustainability narratives to differentiate products and services, concerns about misleading, unsubstantiated, or exaggerated claims—commonly termed greenwashing—have emerged as a significant governance challenge [1]. Greenwashing erodes consumer trust, distorts market incentives for genuine sustainability investment, and undermines Sustainable Development Goal 12 (SDG 12: Responsible Consumption and Production) [2]. This challenge is particularly acute in the food sector, where sustainability claims directly influence daily consumption decisions and regulatory scrutiny is increasing.

Prior research demonstrates that perceived greenwashing negatively influences consumer responses, including reduced trust, unfavorable brand attitudes, and lower purchase intention [3,4,5]. However, these findings are typically operationalized at the aggregate research level and are not readily translatable into actionable tools for real-time decision-making. This body of knowledge remains largely confined to academic contexts, offering limited practical guidance for marketing teams, sustainability managers, and regulators seeking to evaluate sustainability claims prior to campaign deployment.

To address this implementation gap, this study translates validated behavioral models into a deployable decision support system. Rather than re-testing established relationships, the study demonstrates how empirically grounded models can be operationalized into a system that enables real-time assessment of greenwashing risk. The novelty of this study lies in demonstrating how validated behavioral models can be translated into a publicly deployable decision support system with real-time diagnostic capability.

The result is the Greenwashing Advertising Message Assessment (GWAMA), a web-based decision support system that embeds structural equation model (SEM) parameters into a user interface that generates real-time greenwashing risk diagnostics for food advertising messages. GWAMA is the third study in a sequential research program: Study 1 (quasi-experimental, N = 400) established context-specific message framing effects on greenwashing perception; Study 2 (CB-SEM, N = 400) validated the S-O-R structural relationships—PGC → GST → PI—that serve as the behavioral foundation of the system; and Study 3, reported here, translates those validated parameters into a deployed decision support system using the Design Science Research methodology. The development, deployment, and initial usability evaluation of GWAMA are reported in full below.

Therefore, the research questions addressed in this study are as follows:

RQ1:

How can validated behavioral SEM findings be operationalized into a deployable decision support system while preserving psychometric integrity?

RQ2:

What system architecture, scoring algorithm, and user workflow best translate theoretical S-O-R constructs into an accessible, real-time greenwashing diagnostic tool?

RQ3:

To what extent does the deployed system demonstrate stakeholder acceptance across perceived usefulness, ease of use, and intention to use among food industry professionals?

2. Literature Review

2.1. Greenwashing and the Governance Gap

Table 1. Overview of published greenwashing assessment studies and positioning of GWAMA.

Author and Year of Study	Method/Tool	Deployed Web System	SEM-Based Modeling	Purchase Intention Prediction	Product/Service Context	Regulatory Alerts	Food Industry Context
Chen & Chang (2013) [4]	Survey scale	✗	✓	✗	✗	✗	✗
Nyilasy et al. (2014) [5]	Experiment	✗	✓	✗	✗	✗	✗
Leonidou & Skarmeas (2017) [7]	Survey scale	✗	✓	✗	✗	✗	✗
Schmuck et al. (2018) [8]	Experiment	✗	✗	✗	✗	✗	✗
de Freitas Netto et al. (2020) [3]	Conceptual Taxonomy	✗	✗	✗	✗	✗	✗
Szabo & Webster (2021) [9]	Interview + Experiment	✗	✗	✗	✗	✗	✗
Azazz et al. (2024) [6]	Survey scale	✗	✓	✗	✗	✗	✓
GWAMA (This study, 2025)	Web-based Decision support system	✓	✓	✓	✓	✓	✓

Note: ✓ = present; ✗ = absent.

synthesizes prior greenwashing assessment research and highlights the absence of operationalized, practitioner-oriented system implementations. Prior studies have established validated behavioral frameworks, experimental evidence on greenwashing perception [6,7,8], and mixed-method evidence on perceived greenwashing in digital environments [9], and developed taxonomies of greenwashing typologies [3]. Research in the food and restaurant context further demonstrates that perceived greenwashing reduces consumer trust [4] and has been shown to negatively influence purchase intention [5].

Despite this theoretical and empirical progress, these frameworks are primarily applied at the aggregate research level and are not readily translated into tools for real-time decision-making. As a result, the application of greenwashing research remains largely confined to academic contexts, with limited accessibility for practitioners such as marketing teams, sustainability managers, and regulators. Existing research has not translated greenwashing measurement models into integrated, practitioner-oriented systems capable of real-time application. This gap is particularly pronounced in emerging market contexts such as Thailand, where environmental symbols and eco-label cues have been shown to produce complex and sometimes counterintuitive consumer responses [10]. To address this limitation, this study develops the Greenwashing Advertising Message Assessment (GWAMA), a web-based decision support system that translates validated behavioral models into an accessible tool for evaluating greenwashing risk in food marketing communication.

Therefore, Table 1 reveals two compounding gaps in existing literature. First, academic studies have established recent empirical applications examining greenwashing in the food context [6], a validated behavioral framework [7], and experimental evidence on greenwashing perception [8], but these remain confined to survey and experimental contexts. Recent advances in artificial intelligence and machine learning have begun to extend greenwashing detection into computational domains, as evidenced by systematic analyses of AI applications [11] and emerging empirical detection models [12], yet practitioner-oriented, deployable assessment tools grounded in behavioral models remain absent from the literature. Second, limited evidence of published work, both from academics and practitioners, has operationalized these frameworks into a publicly deployed assessment system. GWAMA addresses both gaps simultaneously.

To further contextualize this gap,

Table 2. Comparative Overview of Open Access Practitioner Tools for Assessing Greenwashing Advertising Messages.

Assessment Tool	Function	Methodology	Primary Target User	Claim Components Assessed	Output Format
TerraChoice—Seven Sins of Greenwashing	Identify and classify greenwashing claims to help consumers evaluate environmental sustainability claims	Conceptual taxonomy derived from large-scale marketplace audits of environmental claims	Consumer	Seven claim typologies: hidden trade-off, no proof, vagueness, worshipping false labels, irrelevance, lesser of two evils, fibbing	Checklist document for classifying misleading environmental claims
Office of the Consumer Protection Board (OCPB), Thailand	Protect consumers from misleading or deceptive advertising, including unsubstantiated environmental claims	Legal and regulatory enforcement under Consumer Protection Act B.E. 2522 (Section 22); complaint-based investigation and administrative enforcement	Regulatory authorities, businesses operating in Thailand	False or exaggerated statements (Section 22(1)); statements causing material misunderstanding about goods or services (Section 22(2))	Legal enforcement mechanisms, administrative orders, and regulatory decisions
GWAMA (This study)	Assess greenwashing risk and predict its implications for purchase intention	Survey-based diagnostic tool validated through structural equation modelling (SEM); factor-loading-weighted composite scoring	B2B—food industry practitioners, sustainability managers, marketing agencies, business operating in Thailand	6 items of Perceived Greenwashing Communication (PGC); 4 items of Green Skepticism (GST); predicted Purchase Intention (PI) impact via SEM-derived regression	Interactive risk dashboard with item-level diagnostics, OCPB compliance alerts, and context-specific recommendations on a publicly deployed web platform

provides a comparative overview of existing practitioner-oriented tools and frameworks for assessing greenwashing in advertising messages. The comparison highlights three dominant approaches currently available open access. First, conceptual classification tools such as the TerraChoice “Seven Sins of Greenwashing” [13] offer structured taxonomies for identifying misleading environmental claims; however, these tools primarily function as diagnostic checklists and lack behavioral validation or predictive capability. Second, regulatory frameworks, exemplified by the Office of the Consumer Protection Board (OCPB) in Thailand [14], provide legal criteria for identifying deceptive advertising practices, focusing on enforcement mechanisms rather than proactive assessment or campaign design support. While both approaches play important roles in identifying and regulating greenwashing, they function mainly at the conceptual or regulatory level and do not provide integrated, user-friendly systems for real-time evaluation.

In contrast, GWAMA extends beyond these existing approaches by operationalizing validated behavioral constructs into a deployable decision support system. As shown in Table 2, GWAMA uniquely integrates (1) empirically grounded measurement of perceived greenwashing communication (PGC) and green skepticism (GST), (2) structural equation model-based prediction of purchase intention impact, and (3) an interactive output format that combines diagnostic insights, regulatory alerts, and context-specific recommendations. This comparison reinforces the identified governance gap by demonstrating that prior tools either lack behavioral grounding, real-time application capability, or integration across diagnostic, predictive, and prescriptive functions. GWAMA addresses these limitations by bridging conceptual frameworks, regulatory standards, and behavioral modeling within a single practitioner-oriented system.

2.2. The Behavioral Model Embedded in GWAMA

GWAMA operationalizes a Stimulus–Organism–Response (S-O-R) framework [15]—a theoretical model in which environmental stimuli trigger internal psychological states that in turn drive behavioral responses—in which Perceived Greenwashing Communication (PGC) functions as the stimulus; Green Skepticism (GST) as the organism, defined as consumers’ tendency to question the credibility of environmental claims [7]; and Purchase Intention (PI) as the behavioral response.

The structural relationships embedded in GWAMA are derived from a validated structural equation model estimated from a survey of 400 Thai food consumers. The model identified the following relationships: PGC → GST (β = 0.528, p < 0.01), GST → PI (β = −0.164, p < 0.05), and PGC → PI (β = −0.453, p < 0.01), with good model fit (CFI = 0.975, RMSEA = 0.059). These empirically estimated parameters are embedded within GWAMA’s scoring algorithm to generate predictive assessments of greenwashing impact. All factor loadings, composite reliability, AVE values, and structural coefficients necessary to independently replicate the GWAMA scoring algorithm are reported in full in Appendix A of this paper, ensuring that the system can be evaluated on its own merits without requiring access to companion studies.

Complementary experimental findings further indicate that greenwashing perception is context sensitive. Attribute-based message framing reduces perceived greenwashing for food products, whereas benefit-based framing is more effective for food services (F(1, 396) = 307.888, p < 0.001, η²p = 0.437)—a large effect indicating that message frame and offering type interact substantially in shaping greenwashing perception. This distinction is theoretically grounded in schema congruity theory, which posits that the degree of alignment between message attributes and existing cognitive schemas influences information processing and evaluative responses [16] which is further supported by empirical evidence in sustainability communication research [17] demonstrating that the effectiveness of message framing varies across communication conditions. Furthermore, prior research shows that the impact of greenwashing on consumer responses is contingent on mediating factors such as trust and brand familiarity, which shape how environmental claims are interpreted and subsequently influence purchase intention [18], thereby reinforcing the need for context-specific communication design.

The purchase intention prediction is based on a unified structural model estimated from the full sample (N = 400), reflecting a design choice that prioritizes model stability and generalizability over subgroup-specific parameterization.

2.3. Design Science and TAM as Evaluation Framework

Decision support systems have long been recognized as tools to support managerial decision-making processes [19]. Design Science Research (DSR) focuses on the creation and evaluation of artifacts that address real-world problems [20]. Within this paradigm, artifact development is recognized as a core scholarly contribution, distinct from behavioral theory development [20,21]. March and Smith [21] classify artifacts into four types: constructs, models, methods, and instantiations.

GWAMA represents an instantiation artifact which is a fully implemented system that operationalizes validated S-O-R structural relationships into executable decision logic. Instantiation artifacts provide empirical validation in real-world contexts. Their effectiveness in translating theoretically grounded models into practical applications is demonstrated through deployment and evaluation. Accordingly, the contribution of this study lies in translating and deploying validated behavioral models into a usable governance tool, rather than re-testing the underlying theory.

DSR is particularly appropriate when a real-world problem lacks an adequate solution and relevant knowledge exists but has not yet been operationalized [20]. Both conditions are satisfied in this study. Section 2.1 identifies a governance gap between greenwashing research and practitioner tools, while a validated S-O-R structural equation model estimated from 400 Thai food consumers (CFI = 0.975, RMSEA = 0.059) provides a robust empirical foundation for system design, with all parameters reported in Appendix A. Following the knowledge contribution framework of Gregor and Hevner [22], which extends the artifact taxonomy of March and Smith [21], this study represents an improvement contribution, applying established knowledge to address a persistent problem in a more effective way.

GWAMA was developed following the five-phase Design Science Research (DSR) process proposed by Peffers et al. [23], consisting of problem identification, objective definition, design and development, demonstration and deployment, and evaluation. These phases guide the systematic development and implementation of the artifact.

In addition, the study adheres to the DSR evaluation principles outlined by Hevner et al. [20], which emphasize rigor, relevance, design, and evaluation. These criteria are reflected in the use of validated behavioral models (rigor), the identification of a governance gap in greenwashing assessment (relevance), the development of a functional system artifact (design), and empirical usability testing (evaluation).

System usability is evaluated using the Technology Acceptance Model (TAM) [24], which assesses perceived usefulness, perceived ease of use, and intention to use as key determinants of technology adoption. TAM is appropriate in this context as GWAMA is intended for practitioner deployment. Consistent with the FEDS framework [25], which distinguishes multiple evaluation strategies, this study adopts an ex post, naturalistic evaluation approach involving real users interacting with the deployed system. Evaluation results are reported in Section 4.

3. Methodology and System Development

This study adopts a Design Science Research (DSR) approach to develop and evaluate GWAMA as a practical solution to the identified governance gap between greenwashing research and practitioner application. DSR is appropriate in this context as it enables the translation of validated behavioral knowledge into a functional artifact designed for real-world use.

GWAMA was developed following the five-phase DSR process proposed by Peffers et al. [23]. First, problem identification established the lack of practitioner-oriented tools for assessing greenwashing in food marketing communication. Second, objective definition specified system requirements, including construct fidelity, usability, accessibility, and contextual sensitivity. Third, design and development involved translating validated S-O-R constructs and SEM parameters into a functional web-based system capable of generating real-time greenwashing diagnostics. Fourth, demonstration and deployment involved making the system accessible for real-world use, enabling evaluation under naturalistic conditions. Finally, evaluation was conducted using a Technology Acceptance Model (TAM) instrument with 150 food industry professionals, assessing perceived usefulness, perceived ease of use, and intention to use as indicators of system acceptance—the three core determinants of technology adoption most relevant to the practitioner deployment context [24]. The overall DSR methodology and knowledge flows are summarized in Figure 1.

3.1. System Overview

GWAMA functions as a web-based decision support system designed to support decision-making under conditions of uncertainty and incomplete information, consistent with established definitions of decision support systems in the information systems literature [19]. The system enables users to input advertising content, assess it across validated dimensions of perceived greenwashing, and receive an integrated risk assessment together with context-specific recommendations through a structured four-step evaluation process. To ensure input validity, the system incorporates an automated pre-screening step that verifies whether submitted text constitutes a food-related advertisement containing environmental or sustainability claims. A set of example advertisements is also provided to guide first-time users and support consistent interpretation of the evaluation criteria. The system operates bilingually in Thai and English via an interface language toggle, enhancing accessibility across user groups. It is publicly accessible online, enabling real-world application and evaluation.

Regarding the model integration strategy, experimental evidence (N = 400, η²p = 0.437) indicates that attribute-based message framing reduces perceived greenwashing for food products, whereas benefit-based framing is more effective for food services. These context-specific findings are incorporated in GWAMA as a qualitative recommendation layer rather than as differentiated predictive algorithms, reflecting practical implementation considerations and avoiding overfitting. The purchase intention prediction component is based on a single structural equation model estimated from the full sample (N = 400; CFI = 0.975, RMSEA = 0.059), operating as a unified equation without subgroup-specific parameterization. All model parameters are reported in Appendix A. This design prioritizes stability and generalizability in translating validated behavioral relationships into a deployable decision support system.

3.2. System Architecture: Four Functional Clusters

GWAMA is structured around four functional clusters, each addressing a specific role within the decision support process. Figure 2 illustrates the system architecture and the interaction between these clusters.

Risk Mitigation: This cluster provides pre-campaign diagnostic assessment by identifying which dimensions of perceived greenwashing contribute to elevated risk. Based on these diagnostics, the system generates context-specific recommendations. For food product advertisements, guidance emphasizes verifiable attributes. For food service advertisements, recommendations focus on operational practices and experiential sustainability disclosures. Results are presented using role-specific interpretation formats to enhance usability across different user groups.

Governance Strengthening: This cluster supports regulatory awareness by generating item-level compliance alerts when an individual PGC item score exceeds a predefined threshold. Specifically, a compliance alert is triggered when any single PGC item receives a mean rating above 3.50 on the five-point scale—a value chosen because it exceeds the PGC population mean (M = 3.05) by approximately half a standard deviation (SD = 0.83), representing a meaningful and statistically grounded deviation from the average consumer baseline rather than an arbitrary cut-off. This threshold identifies individual claim dimensions (e.g., vagueness, unprovability, or exaggeration) that are statistically elevated relative to the population and therefore warrant regulatory scrutiny. Alerts indicate potential alignment with relevant provisions of Thailand’s Consumer Protection Act (B.E. 2522), including Section 22 on misleading or exaggerated claims [14]. Each alert is linked to the specific evaluation item that triggered the flag and is presented bilingually to support accessibility.

Behavioral Impact Prediction: This cluster translates diagnostic results into estimated behavioral outcomes by applying the structural equation model to predict the potential impact on purchase intention. The prediction is based on a unified population-level model (N = 400) and is applied consistently across both product and service contexts.

Capability Development: This cluster supports continuous improvement by enabling repeated assessments and providing interpretive guidance. It facilitates the development of organizational capabilities in sustainability communication through structured feedback and learning mechanisms.

3.3. GWAMA User Workflow

The GWAMA interface is structured as a sequential workflow designed to operationalize the validated constructs of the S-O-R framework within a user-friendly decision support process. While the workflow is not a direct theoretical representation, it is aligned with the underlying behavioral model by guiding users from stimulus evaluation to behavioral outcome interpretation.

In the pre-assessment stage, users optionally configure their role (e.g., Food Entrepreneur, Marketing/Advertising Agency, Marketing/Advertising Manager, or Other), which enables role-specific interpretation of results without altering the underlying scoring algorithm. This configuration ensures that identical diagnostic outputs are translated into stakeholder-relevant managerial insights. Users are additionally presented with an informed consent statement prior to system access, confirming voluntary participation, data anonymity, and the academic research purpose of the platform.

The assessment stages involve entering the advertisement text and selecting the food offering type (product or service). This selection activates a context-specific recommendation pathway, whereby attribute-based guidance is emphasized for food products and benefit-based guidance for food services. Importantly, this contextual adaptation affects only the recommendation logic, while the scoring algorithm and predicted Purchase Intention (PI) remain invariant across conditions. Subsequently, these steps correspond to the measurement of Perceived Greenwashing Communication (PGC) and Green Skepticism (GST), respectively. Users evaluate the advertisement using validated multi-item Likert scales (see Appendix A), with the system preserving item-level diagnostics for subsequent analysis. The assessment workflow is illustrated in Figure 3.

The final stage presents the integrated results dashboard, which combines composite PGC and GST scores with SEM-based PI (Purchase Intention) prediction. An example output is shown in Figure 4. The results dashboard is structured across four integrated output clusters. The first cluster presents an overall risk classification using a color-coded five-level scale. This is accompanied by legal compliance alerts aligned with OCPB Section 22, along with item-specific revision guidance. The second cluster presents a Population Comparison Analysis via a three-axis radar chart plotting the respondent’s PGC, GST, and PI scores against population mean benchmarks derived from the validation study, with z-scores and directional status labels (Above Average, Near Average, Below Average) enabling practitioners to contextualize their advertisement’s performance relative to the Thai food consumer baseline. The third cluster delivers a behavioral summary with plain-language interpretation of the predicted PI impact, and a role-specific managerial interpretation tailored to the user’s declared role, translating the statistical output into a stakeholder-relevant action directive. The fourth cluster provides item-level recommendations for each PGC dimension, differentiated by offering type (food product or food service), alongside general PI-level recommendations such as strengthening claim credibility and obtaining third-party endorsements. A summary scorecard and social sharing functionality complete the report. This illustrates how GWAMA translates psychometric inputs into interpretable, multi-layered managerial insights without requiring statistical expertise from the user.

3.4. Scoring Algorithm

The GWAMA scoring engine operationalizes user responses through a three-stage computational procedure that translates observed ratings into interpretable risk and behavioral outcome metrics.

First, construct-level composite scores are computed using factor-loading-weighted aggregation:

$$\mathrm{Composite} = {\displaystyle \frac{\sum ({\lambda }_i\times r_i)}{\sum {\lambda }_i}}$$

where ${\lambda }_i$ irepresents standardized factor loadings derived from the validated measurement model, and $r_i$ denotes user ratings for item $i$. This approach preserves the relative contribution of each indicator in accordance with the reflective measurement structure [26,27]—that is, each item is treated as an effect of the underlying construct, so indicators with stronger factor loadings contribute more to the composite score, ensuring that the weighting reflects the empirically estimated importance of each item.

Second, composite scores are standardized using population parameters obtained from the validation sample:

PGC: M = 3.05, SD = 0.83

GST: M = 4.65, SD = 0.48

Third, predicted Purchase Intention (PI) is calculated using the standardized regression equation derived from the structural model:

PI_z = −0.453 (PGC_z) − 0.164 (GST_z)

The standardized PI score is then transformed back to the original five-point Likert scale:

PI = (PI_z × 1.02) + 3.33

The resulting PI value (M = 3.33, SD = 1.02) is used to enhance interpretability for end users. This transformation assumes linear relationships and approximate normality consistent with SEM estimation procedures. To support managerial decision-making, the back-transformed PI score is presented as a complementary behavioral indicator alongside the PGC-based risk diagnostics. It is critical to note that greenwashing risk in GWAMA is determined by PGC composite scores, which directly reflect the degree to which advertisement claims are perceived as greenwashing. The PI prediction serves a distinct function: it quantifies the behavioral consequence of that greenwashing perception on consumer purchase intention. Because the structural model embeds negative path coefficients (PGC → PI: β = −0.453; GST → PI: β = −0.164), greenwashing perception suppresses purchase intention. A higher predicted PI therefore indicates that this suppression is limited—the advertisement is performing well behaviorally. A lower predicted PI signals that greenwashing perception is substantially suppressing purchase intention, warranting intervention. The PI classification reflects the degree to which greenwashing perception suppresses consumer purchase intention. While lower PI values indicate stronger negative behavioral impact, greenwashing risk itself is determined by PGC-based diagnostics, with PI serving as a complementary behavioral indicator. The five PI bands are interpreted as follows: Very High Behavioral Suppression (PI = 1.00–1.80) indicates the most severe negative impact on purchase intention; High Behavioral Suppression (PI = 1.81–2.60) indicates substantial impact; Moderate Behavioral Impact (PI = 2.61–3.60) indicates near-average impact, anchored by the population mean (M = 3.33, SD = 1.02); Limited Behavioral Suppression (PI = 3.61–4.20) indicates minimal negative impact on purchase intention; and Negligible Behavioral Suppression (PI = 4.21–5.00) indicates that the advertisement is unlikely to negatively influence consumer behavior. The lower the PI, the greater the priority for revising claim content. These classifications are accompanied by color-coded indicators and context-specific recommendations.

Recommendation pathways are adapted based on offering type. Food product assessments emphasize attribute-based guidance, whereas food service assessments emphasize benefit-based guidance. This differentiation is grounded in experimental evidence demonstrating a significant crossover interaction between message framing and offering type (F(1, 396) = 307.888, p < 0.001, η²p = 0.437), as demonstrated in Study 1 of the research program. The GWAMA scoring mechanism integrates insights from both empirical studies. The recommendation rules are derived from the experimental findings of Study 1, while the structural parameters, including factor loadings, path coefficients, and population statistics, are estimated from the SEM analysis conducted in Study 2. These SEM-derived parameters are applied as fixed constants within the scoring engine to ensure consistency between the validated behavioral model and the system implementation. Full parameter specifications are provided in Appendix A. These findings further highlight that sustainable decision-making is inherently complex and influenced by multiple psychological and contextual factors [28], reinforcing the importance of structured, model-based assessment approaches in interpreting greenwashing risk.

It is important to note that the PI prediction model is applied uniformly across offering types. The regression equation is estimated from the full sample (N = 400) without stratification; therefore, predicted values represent generalized population-level expectations rather than context-specific forecasts. Accordingly, PI outputs are presented to users as indicative estimates rather than precise predictions. This design reflects a deliberate methodological choice: employing a unified population-level model prioritizes parameter stability, replicability, and deployment feasibility over subgroup-specific precision. Although contextual differences between food products and food services were identified in the experimental phase of the research program, developing separate predictive models would require sufficiently large subgroup samples to support stable multi-group SEM estimation. In addition, formal measurement invariance testing would be a necessary prerequisite before comparing or implementing context-specific structural coefficients across offering types. Both requirements exceed the scope of this initial deployment study. Consequently, the contextual sensitivity of the GWAMA system is implemented primarily at the recommendation layer rather than the predictive layer. Future research should extend the system through multi-group SEM and measurement invariance testing to determine whether context-specific recalibration of the predictive engine is empirically justified.

For illustration, advertisements containing vague or unverifiable environmental claims tend to produce elevated PGC composite scores—indicating higher greenwashing risk—and correspondingly lower predicted PI values, signaling greater suppression of consumer purchase intention. Both outputs trigger recommendations to incorporate specific, evidence-based claims prior to campaign deployment. Full measurement items, factor loadings, and confirmatory factor analysis results are reported in Appendix A. The complete system architecture and data flow are illustrated in Figure 5, detailing how user inputs are transformed through the scoring pipeline into the four functional outputs presented in the GWAMA dashboard.

3.4.1. Prediction Uncertainty and Interpretation

The PI prediction generated by GWAMA represents a population-level expected value derived from the structural equation model. While this point estimate provides a useful indicator of likely consumer response, individual purchase intention may vary due to unobserved factors not captured by the model.

The explanatory power of the PI equation is reflected in an R² value of 0.311, indicating that approximately 31.1% of the variance in purchase intention is explained by perceived greenwashing communication (PGC) and green skepticism (GST). The remaining unexplained variance reflects heterogeneity in consumer responses, contextual influences, and measurement noise. Accordingly, PI outputs should be interpreted as indicative rather than deterministic. It is important that practitioners do not overinterpret the precision of predicted PI values. The dashboard presents a population-level point estimate rather than a deterministic forecast for any individual advertisement or consumer response. The structural model explains approximately one-third of the variance in purchase intention; the remaining unexplained variance reflects factors beyond the scope of this system, including brand equity, pricing, competitive context, and consumer-specific characteristics. Accordingly, the PI output should be interpreted as a directional indicator of behavioral risk rather than as an exact behavioral prediction. This interpretation is consistent with the intended role of GWAMA as a decision support tool designed to facilitate structured managerial reflection and pre-campaign diagnostic assessment, rather than as a predictive forecasting engine or behavioral oracle. For example, an advertisement yielding a predicted PI above the population mean (M = 3.33) suggests that greenwashing perception has a relatively limited negative impact on purchase intention, indicating a lower risk profile on average. Conversely, a predicted PI below the population mean signals that greenwashing perception is suppressing purchase intention below the baseline, warranting closer review of claim content.

To maintain usability for non-technical users, GWAMA presents the point estimate as the primary diagnostic output. However, the system design explicitly acknowledges prediction uncertainty by framing results as expected values and by providing item-level diagnostics and qualitative recommendations that support informed managerial interpretation. Future versions of the system will incorporate confidence intervals or uncertainty ranges alongside the point estimate to further support transparent and calibrated managerial interpretation.

3.4.2. Effect Size of Embedded Regression Coefficients

The regression coefficients embedded in GWAMA’s scoring engine are derived from a previously validated structural equation model and are implemented as fixed parameters within the system. The direct effect of perceived greenwashing communication (PGC) on purchase intention (PI) is represented by a standardized coefficient of β = −0.453, while the effect of green skepticism (GST) on PI is represented by β = −0.164.

Within the system, these coefficients are used to translate user-generated PGC and GST scores into an estimated purchase intention outcome. The relative magnitude of the coefficients results in PGC exerting a stronger suppressive influence on predicted PI compared to GST. Because both coefficients are negative, higher PGC scores produce lower predicted PI values, signaling greater suppression of purchase intention. This weighting is reflected in the prioritization of PGC diagnostics within the Risk Mitigation cluster of the system outputs.

In addition to the direct effect, the model structure incorporates an indirect pathway from PGC to PI via GST, which is operationalized within the system through the sequential computation of PGC and GST composites prior to PI estimation. The full model explains approximately 31.1% of the variance in purchase intention, and this level of explanatory power defines the predictive scope of the system. These parameters are estimated using a Thai food consumer sample (N = 400) and are applied in GWAMA as population-level coefficients. Accordingly, the system outputs should be interpreted as expected values at the population level rather than deterministic predictions for individual consumers. Application in alternative contexts may require recalibration of the underlying model parameters.

3.4.3. Sensitivity Analysis: Weighting Scheme Robustness

To examine the robustness of the composite scoring procedure, a sensitivity analysis was conducted comparing three alternative weighting schemes: (1) factor-loading-weighted scoring (GWAMA’s default approach), (2) equal-weight simple mean, and (3) reliability-weighted scoring using squared loadings (λ²). The analysis was performed using a single advertisement (Chinese Kale food product context). The same set of user ratings for all PGC and GST items was applied across the three weighting schemes, ensuring that any variation in composite scores and predicted outcomes was attributable solely to differences in weighting methodology rather than input variation. The results are reported in

Table 3. Sensitivity Analysis.

Weighting Method	PGC Composite	GST Composite	Predicted PI	PI Behavioral Classification
Factor-loading weighted (Σλᵢrᵢ/Σλᵢ)	2.805	3.753	3.779	Limited Suppression
Equal weights (simple mean)	2.833	3.750	3.764	Limited Suppression
Reliability-weighted (Σλᵢ2rᵢ/Σλᵢ2)	2.777	3.756	3.793	Limited Suppression
Maximum divergence	0.056	0.006	0.029	—

.

Across all three methods, composite scores and predicted outcomes exhibited minimal variation. PGC scores ranged from 2.777 to 2.833 (maximum difference: 0.056 Likert units), GST scores from 3.750 to 3.756 (maximum difference: 0.006), and predicted purchase intention from 3.764 to 3.793 (maximum difference: 0.029 Likert units, equivalent to less than 1% of the five-point scale range). Importantly, all three methods yielded the same PI behavioral outcome classification (Limited Behavioral Suppression: PI = 3.61–4.20), indicating that the system’s governance outputs are stable across alternative weighting specifications. A PI in this range indicates that the advertisement’s predicted purchase intention exceeds the population mean (M = 3.33), reflecting limited suppression of consumer purchase intention by greenwashing perception. Note that this PI classification describes a behavioral consequence, not the greenwashing risk level itself, which is determined separately by the PGC composite score.

These findings suggest that while different weighting schemes produce numerically similar results in this case, the factor-loading-weighted approach remains theoretically appropriate, as it preserves the relative contribution of each indicator as estimated in the measurement model. In contrast, equal weighting assumes uniform indicator importance, which may not align with the underlying construct structure. Therefore, the selection of factor-loading-weighted scoring in GWAMA is primarily theory-driven rather than outcome-sensitive. The robustness observed in this analysis indicates that the system’s diagnostic outputs are not dependent on a specific weighting configuration, enhancing confidence in its practical applicability.

All three weighting methods yield consistent PI behavioral outcome classification (Limited Behavioral Suppression: PI = 3.61–4.20), confirming that a predicted PI in this range reflects limited suppression of purchase intention relative to the population baseline. Maximum variation across methods is minimal, with predicted purchase intention differing by less than 0.03 Likert units. Factor-loading weights for PGC range from λ = 0.720–0.841 and for GST λ = 0.768–0.793. Results are based on the Chinese Kale advertisement example (food product context). Critically, all three methods produced the same PI behavioral outcome classification (Limited Behavioral Suppression: PI = 3.61–4.20), confirming that GWAMA’s governance outputs are robust to weighting scheme selection. The factor-loading-weighted method is retained as the primary scoring approach because it preserves the relative importance of each indicator as estimated in the measurement model while producing substantively indistinguishable composite scores in practice. However, this sensitivity analysis is based on a single illustrative case and is intended as a robustness check rather than a formal statistical comparison; future research may extend this validation across a broader set of advertising stimuli.

Therefore, Figure 5 illustrates the four-stage data flow from user inputs (advertisement text, offering type, PGC and GST ratings) through factor-loading-weighted composite scoring, population-based standardization, SEM-derived purchase intention prediction, and five-band risk classification into the four governance-oriented output clusters presented in the GWAMA dashboard.

4. Initial System Evaluation Results

4.1. Evaluation Approach and Sample

The initial evaluation of GWAMA was conducted using the Technology Acceptance Model (TAM) to assess user perceptions of system usability and behavioral intention to adopt the tool in practice. TAM was selected due to its established validity in evaluating user acceptance of information systems [24], particularly in decision support contexts which perceived usefulness and ease of use influence adoption behavior [19].

A total of 150 respondents participated in the evaluation under live system conditions, interacting directly with the publicly deployed GWAMA platform. Participants were Thai food industry practitioners recruited through purposive snowball sampling, initiated via the researchers’ personal and professional networks. The inclusion criterion required that participants were actively engaged in, or responsible for, developing and reviewing food advertising prior to campaign deployment—that is, practitioners with direct occupational relevance to pre-campaign greenwashing risk screening. Eligible roles included food entrepreneurs, marketing managers, and marketing agency professionals operating within the Thai food sector. This criterion-based approach ensured that respondents possessed the contextual expertise necessary to meaningfully evaluate the system’s practical utility, and that the sample reflected diverse decision-making perspectives across food business ownership, in-house marketing management, and agency practice.

In addition to the three core TAM constructs, supplementary system-level usability indicators—user interface design, application performance, and data security—were included to provide a broader assessment of system readiness for real-world deployment. This multi-dimensional evaluation approach is consistent with design science research (DSR), which emphasizes both user acceptance and the practical utility of the artifact [25]. It is important to note that the present evaluation assesses usability acceptance—specifically practitioners’ perceptions of the system’s usefulness, ease of use, and intention to adopt it—rather than the actual effectiveness of the system in reducing greenwashing practices. The evaluation is based on users declared perceptions following interaction with the live system, not on longitudinal organizational outcomes such as measurable improvements in advertising claim quality or verified reductions in greenwashing behavior. This distinction is consistent with early-stage Design Science Research (DSR) evaluation logic [25], which prioritizes feasibility, usability, and practitioner acceptance before long-term effectiveness can be rigorously assessed. Accordingly, the current findings should be interpreted as evidence of initial deployment acceptance rather than as proof of real-world behavioral or organizational impact. Larger-scale, longitudinal, and outcome-based evaluations remain important directions for future research.

4.2. TAM Usability Results

The TAM evaluation results across the three core dimensions, perceived usefulness, perceived ease of use, and intention to use, are presented in

Table 4. TAM Usability Evaluation Results—GWAMA Live System (N = 150).

TAM Dimension	Items	M	SD	Interpretation	Highest-Scoring Item
Perceived Usefulness	7	4.18	0.77	Very High	PU5: Efficiency (M = 4.63)
Perceived Ease of Use	6	4.03	0.74	High	PEOU4: Convenience (M = 4.48)
Intention to Use	5	4.25	0.66	Very High	IU5: Intended Use (M = 4.66)

.

Table 4 presents the results of the Technology Acceptance Model (TAM) evaluation across three core dimensions: perceived usefulness, perceived ease of use, and intention to use.

Perceived Usefulness was rated highest overall (M = 4.18, SD = 0.77), indicating that respondents considered the system effective for assessing greenwashing advertising messages. The highest-scoring item was “It helps organizations conduct greenwashing assessments more efficiently” (M = 4.63, SD = 0.76), followed by “It is beneficial for assessing greenwashing advertising messages” (M = 4.55, SD = 0.75), both at the very high level. These results suggest that users recognize the system’s practical value in supporting sustainability communication evaluation.

Perceived Ease of Use was also rated highly (M = 4.03, SD = 0.74), indicating that the system interface is clear and accessible. The highest-rated item, “It makes work easier, more convenient, and faster” (M = 4.48, SD = 0.66), reflects the system’s ability to streamline the evaluation process. Other items related to usability and clarity were consistently rated at the high level, supporting the system’s usability for non-technical users.

Intention to Use was similarly high (M = 4.25, SD = 0.66), indicating strong behavioral intention to adopt the system. The highest-scoring item was “I intend to use this system” (M = 4.66, SD = 0.64), followed by willingness to recommend and future usage intention. These findings suggest positive adoption potential among food industry practitioners.

The internal consistency of the TAM instrument was confirmed, with Cronbach’s α = 0.811, exceeding the recommended threshold of 0.70 [26], indicating good reliability and supporting construct-level interpretation [27]. This finding is particularly important, as behavioral intention is widely recognized as a strong predictor of actual system adoption [24], suggesting that GWAMA demonstrates substantial potential for real-world implementation.

Beyond the TAM constructs, additional system-level indicators were assessed to evaluate the operational readiness of the deployed system. Among these indicators, user interface design included the highest-rated individual item, with “The evaluation format is appropriate and correct” achieving the highest score across all measured items (M = 4.67, SD = 0.69). This suggests that users perceive the system interface as logically structured and methodologically appropriate. Application performance was generally rated at a high level, with scores ranging from 3.77 to 4.38. The highest-rated item, “The application has potential to further develop for commercialization” (M = 4.38, SD = 0.74), indicates strong perceived scalability, although other operational aspects were rated slightly lower. Data security perceptions were consistently rated at a high level (M range: 3.80–4.00), indicating acceptable but comparatively lower confidence relative to other system dimensions. This suggests that while basic security expectations are met, further enhancement may strengthen user trust in future system iterations. Adoption intention findings were consistent with the high intention-to-use scores observed in the TAM evaluation, suggesting that willingness to adopt the system is driven primarily by perceived usefulness and usability rather than system constraints. Where hesitation was observed, it was associated with contextual factors such as organizational size or limited engagement with sustainability communication rather than deficiencies in system functionality.

4.3. Construct Reliability of the Scoring Engine

The fidelity of the GWAMA scoring engine was assessed by verifying the accurate implementation of the structural equation modeling (SEM) parameters within the system algorithm. This evaluation is conceptually distinct from the TAM-based usability assessment, as it focuses on the internal consistency between the computational logic of the system and the validated measurement and structural models. Specifically, the scoring engine incorporates the factor loadings associated with the Perceived Greenwashing Communication (PGC) and Green Skepticism (GST) constructs, with loading ranges of λ = 0.720–0.841 (PGC) and λ = 0.768–0.793 (GST), as established in the underlying SEM analysis. Composite scores generated by the system were cross-checked against manually computed SEM-based values, confirming that the implementation produces numerically consistent results within acceptable rounding tolerance.

Importantly, the GWAMA system does not re-estimate measurement parameters; rather, it operates previously validated constructs by embedding their factor loadings directly into the scoring algorithm. The reliability and validity of these constructs—evidenced by composite reliability where CR = 0.906 (PGC) and 0.859 (GST) and average variance extracted (AVE = 0.605–0.616), with discriminant validity confirmed using the Fornell–Larcker criterion [26]—are established in the foundational model (see Appendix A). Accordingly, the contribution of this section lies in demonstrating implementation fidelity, ensuring that the system faithfully translates a validated theoretical model into a functioning decision support tool.

4.4. Structural Path Stability Analysis

To further examine the robustness of the system under real-world usage conditions, an exploratory structural path analysis was conducted using the evaluation dataset (N = 150). This analysis was not intended to revalidate the theoretical model, but rather to assess the stability of key relationships when applied in a smaller, practitioner-oriented sample.

The analysis indicated that the path from perceived usefulness (PU) to intention to use (IU) was statistically significant (β = 0.310, p = 0.029), consistent with the positive direction predicted by established TAM theory. Given that TAM consistently predicts a positive relationship between perceived usefulness and behavioral intention [24], this result provides exploratory support for the structural coherence of the system evaluation, while acknowledging that sample size constraints limit the strength of any structural inference.

Accordingly, these findings are interpreted as exploratory and provide initial structural support for the TAM framework as applied in this practitioner-oriented context. They reinforce the decision to treat TAM results in Section 4.2 as descriptive indicators of user acceptance, while suggesting that the directional relationships are broadly consistent with established theory. Overall, the stability analysis supports the robustness of the GWAMA system by demonstrating that, within this small-sample structural estimation, the system’s core functionality, based on previously validated SEM parameters, remains theoretically grounded and operationally reliable.

This finding is consistent with the functional validity of the system, as TAM is applied descriptively to assess user acceptance rather than to test structural relationships.

4.5. Illustrative Numerical Example

To demonstrate the operational logic of the GWAMA scoring engine,

Table 5. Illustrative Scoring Example: Food Product Advertisement.

Item	Measurement Statement (Abbreviated)	Rating (r)	Loading (λ)	λ × r
Panel A: Perceived Greenwashing Communication (PGC)
PGC1	Claim is difficult to verify	3	0.765	2.295
PGC2	Claim is vague or unprovable	4	0.731	2.924
PGC3	Overstates environmental performance	2	0.840	1.680
PGC4	Omits important information	3	0.841	2.523
PGC5	Uses information that appears false	2	0.808	1.616
PGC6	Does not accurately represent impact	3	0.720	2.160
PGC Composite	Σ(λ × r) = 13.198/Σλ = 4.705 → 2.805
Panel B: Green Skepticism (GST)
GST1	Skeptical: Environmentally friendly?	4	0.777	3.108
GST2	Skeptical: less damaging?	4	0.793	3.172
GST3	Skeptical: meets high standards?	3	0.768	2.304
GST4	Skeptical: better for environment?	4	0.771	3.084
GST Composite	Σ(λ × r) = 11.668/Σλ = 3.109 → 3.753

presents a worked example using a hypothetical food product advertisement in the Thai context. The advertisement message “Our vegetables are grown using environmentally friendly agricultural practices that protect the soil and reduce chemical use” was entered into the system under the food product pathway and evaluated for illustrative purposes using a single set of ratings.

Table 5 reports on the item-level ratings, SEM-derived factor loadings, and weighted contributions used to compute the composite scores for Perceived Greenwashing Communication (PGC) and Green Skepticism (GST).

From the table, the scoring process can be explained in the following stages:

In brief, the four-stage process proceeds as follows: (1) factor-loading-weighted composite scores are computed for PGC and GST; (2) scores are standardized using population parameters from the validation sample; (3) predicted Purchase Intention (PI) is calculated via the SEM-derived regression equation; and (4) outputs are classified into risk and behavioral impact bands and presented on the results dashboard. A fully worked numerical example, including step-by-step calculations for a Chinese Kale food product advertisement, is provided in Appendix B.

5. Discussion

The development of GWAMA contributes to an emerging body of scholarship at the intersection of greenwashing governance and practitioner-oriented decision support. Existing reviews of the greenwashing literature have documented the proliferation of misleading environmental claims across consumer, investor, and institutional contexts, and have identified the failure to bridge the translation gap to practitioners as a key shortcoming—observing that the field has yet to offer comprehensible and practical solutions to well-meaning executives, activists, and policymakers who seek to address greenwashing in practice [29]. Efforts to develop actionable assessment tools have been proposed, with one interdisciplinary framework offering a first attempt at analyzing the quality and truthfulness of green claims across diverse organizational actors [30]. However, such frameworks remain conceptual checklists that are not grounded in validated consumer behavioral models and have not been deployed as real-time decision support systems. GWAMA addresses this gap by demonstrating that psychometrically validated structural equation models can serve not only as explanatory instruments but as the computational foundation for a practitioner-facing diagnostic system—translating validated behavioral constructs into operational governance infrastructure accessible to non-specialist users. The present findings suggest that the principal barrier to such translation is not theoretical but architectural: the challenge lies in preserving psychometric integrity while rendering model outputs interpretable without requiring statistical expertise from the user. Future decision support systems in adjacent domains—including sustainability reporting, environmental labeling, and CSR communication—may draw on the scoring architecture demonstrated here as a replicable methodological template.

5.1. Key Findings and Interpretation

The findings of this study demonstrate the feasibility of translating behavioral models of greenwashing perception into a functioning decision support system. By integrating SEM-derived relationships with a structured scoring architecture, GWAMA operationalizes constructs such as perceived greenwashing communication (PGC) and green skepticism (GST) into real-time diagnostic outputs that can be interpreted without statistical expertise. A key insight from the system design is the balance between predictive consistency and interpretive flexibility. While purchase intention is predicted using a unified population-level SEM, the recommendation layer adapts to contextual differences between food products and services. This reflects a deliberate design trade-off between generalizability and practical relevance, enabling the system to provide context-sensitive guidance while maintaining theoretical coherence.

The results also highlight the central role of claim clarity and verifiability in shaping greenwashing perceptions. Both the scoring outputs and the underlying model indicate that vague and unsubstantiated claims are primary drivers of perceived greenwashing. This reinforces the importance of transparent communication practices in sustainability marketing and suggests that improving claim specificity may be a critical lever for reducing consumer skepticism.

An important conceptual distinction should be made explicit here. GWAMA measures perceived greenwashing communication risk—that is, the extent to which a food advertisement’s sustainability claims are likely to be interpreted by consumers as misleading, vague, exaggerated, or insufficiently substantiated—rather than the objective level of greenwashing as determined through independent verification of a firm’s actual environmental practices or sustainability performance. These two constructs are related but conceptually distinct: a firm may engage in legitimate sustainability practices yet communicate them in ways that trigger elevated consumer skepticism, while conversely, highly polished sustainability communication may obscure limited underlying environmental performance. Accordingly, GWAMA is specifically designed to diagnose communication-level perceptual risk grounded in consumer interpretation and psychological response. Practitioners should therefore use GWAMA diagnostics as a starting point for improving communication quality, not as a substitute for substantive environmental performance assessment.

5.2. Contributions

This study makes three primary contributions to the literature on sustainability communication, behavioral modeling, and applied system design.

Theory-to-Practice Contribution: First, this study demonstrates how a validated behavioral framework can be systematically operationalized into a functioning decision support system. Grounded in the Stimulus–Organism–Response (S–O–R) framework, GWAMA translates the relationships between PGC, GST, and purchase intention into a structured, user-facing diagnostic tool. A key design feature is the separation of recommendation pathways by offering type. Food product contexts receive attribute-focused guidance (e.g., verifiable claims and certifications), whereas food service contexts receive benefit-focused guidance (e.g., operational practices and experiential sustainability cues). This enables context-sensitive interpretation while maintaining a unified predictive model.

To clarify the hierarchical structure of this study’s contributions: the primary contribution is the DSR instantiation artifact—GWAMA as a publicly deployed decision support system that demonstrates the feasibility of translating validated behavioral models into operational governance tools. This artifact-level contribution is what distinguishes the present study from conventional SEM-based research; it provides empirical proof-of-concept that theoretical constructs can be operationalized into a real-world system without loss of psychometric integrity. The secondary contribution is methodological: the factor-loading-weighted scoring architecture with SEM-derived regression parameters represents a replicable template that other researchers and system developers can apply to analogous behavioral translation problems across sustainability communication domains. These two contribution levels are complementary but distinct—the artifact demonstrates that translation is possible in practice, while the methodology provides the blueprint for how such translation can be achieved with theoretical fidelity.

Methodological Contribution: Second, this study introduces a scoring architecture that integrates factor-loading-weighted composite measures with SEM-derived regression parameters. This approach preserves the relative contribution of each indicator as established in the validated measurement model, in contrast to conventional equal-weighted scoring approaches. Sensitivity analysis indicates that this method is robust, with minimal variation across alternative weighting schemes and stable classification outcomes. The use of SEM-derived coefficients further enables behaviorally interpretable predictions while maintaining transparency in how outputs are generated. By incorporating population-based standardization, the system also provides a consistent reference framework for interpreting results across different inputs.

Governance Contribution: Third, this study contributes to sustainability governance by providing an applied tool that supports responsible communication practices aligned with Sustainable Development Goal 12 (Responsible Consumption and Production). By translating abstract constructs into actionable diagnostics, GWAMA enables firms to evaluate environmental marketing claims prior to market release. The system also provides a structured mechanism for identifying high-risk claim characteristics which can inform both managerial decision-making and the development of clearer regulatory standards. This is particularly relevant in emerging market contexts, where sustainability communication frameworks are still evolving.

5.3. Positioning Within Design Science Research

The present study follows the design science research (DSR) paradigm, which conceptualizes research as a problem-solving approach that generates knowledge through the development and evaluation of artifacts [20]. Within this paradigm, GWAMA represents an instantiation that operationalizes validated behavioral constructs into a functioning system for real-world application.

Recent research suggests that artificial intelligence can reduce greenwashing behavior and improve transparency in environmental disclosures [31]. However, addressing such complex sustainability challenges requires structured problem-solving approaches such as Design Science Research, which enables the development of artifact-based solutions [32]. Prior studies have demonstrated the effectiveness of DSR in designing system-based tools to support sustainability-oriented decision-making [33], highlighting the relevance of this approach for developing practitioner-oriented solutions such as GWAMA.

The broader family of sustainability-oriented decision support systems—spanning carbon footprint tools and supply chain transparency platforms [33]—similarly demonstrates how validated frameworks can be operationalized into practitioner-facing artifacts. Prior DSR studies in sustainability contexts have demonstrated the importance of aligning artifact design with underlying theoretical constructs to ensure both methodological rigor and practical relevance [30,31]. In this study, this alignment is achieved through the direct integration of SEM-derived parameters into the scoring engine, enabling transparent traceability between the underlying behavioral model and system outputs.

GWAMA’s contribution can be situated within the design science research (DSR) knowledge contribution framework proposed by Gregor and Hevner [22] building on the artifact taxonomy established by March and Smith [21] which classifies research according to the maturity of both the problem domain and the solution space. In this framework, GWAMA can be interpreted as an improvement-type artifact, where the problem of greenwashing in sustainability communication is well-established, but practical, deployable solutions remain limited. By developing and implementing a functioning assessment system grounded in behavioral theory, this study contributes a novel artifact addressing a recognized gap between theoretical understanding and practical application.

Within the DSR artifact taxonomy [21], GWAMA represents an instantiation, that is, a fully operational system rather than a conceptual model or methodological proposal. This perspective is consistent with prior DSR emphasizing alignment between conceptual models and system implementation [34]. While prior research has established and validated constructs such as perceived greenwashing communication and green skepticism, their application has largely remained within survey or experimental settings. The present study extends this work by embedding these constructs into a real-time scoring engine, thereby demonstrating their operational feasibility in a practitioner-facing context.

The design of GWAMA reflects the dual-cycle logic of DSR [20]. The relevance cycle is addressed through the identification of a practical governance gap in sustainability communication, particularly in the food sector, where ambiguous environmental claims are prevalent. The rigor cycle is maintained by grounding the system architecture in validated SEM parameters, ensuring that the scoring outputs remain theoretically consistent with the underlying behavioral model. This integration enhances traceability between theory and artifact, a key requirement in DSR-based system development.

In addition to its methodological contribution, GWAMA provides a domain-specific contribution to sustainability communication and greenwashing research. By identifying vagueness and lack of verifiability as primary drivers of perceived greenwashing, the system operationalizes abstract theoretical constructs into actionable diagnostic outputs. This supports practitioners in improving claim transparency and provides policymakers with behavioral evidence relevant to the design of environmental claim standards, particularly in emerging market contexts.

Finally, the live deployment of GWAMA enables initial demonstration and evaluation of the artifact in a real-world setting. Following the design science research methodology [23], the system is made accessible for direct use, allowing for feasibility assessment beyond controlled experimental conditions. The Technology Acceptance Model (TAM) evaluation (N = 150) provides preliminary evidence of usability and adoption potential, consistent with early-stage evaluation frameworks for DSR artifacts [25]. While further large-scale validation is required, these results suggest that the system is both functionally viable and practically relevant.

5.4. Implications

The findings and system design of GWAMA generate implications for three primary audiences: food industry practitioners, regulatory and policy bodies, and the academic research community.

Implications for Food Industry Practitioners: For food manufacturers, food service operators, and marketing agencies, GWAMA addresses a practical gap in sustainability communication: the lack of a structured, pre-campaign diagnostic tool grounded in consumer behavioral evidence. Current practices typically rely on legal review or internal judgment, which may not fully capture how consumers interpret environmental claims. GWAMA enables practitioners to assess greenwashing risk prior to campaign release using validated psychometric constructs. A key insight is that vagueness and lack of verifiability—captured by PGC1 (difficulty of verification) and PGC2 (vagueness or lack of substantiation)—emerge as primary drivers of perceived greenwashing. This suggests that sustainability claims may be strengthened by anchoring them to verifiable evidence, such as third-party certifications or quantified environmental metrics, rather than relying on general descriptors such as “eco-friendly.” The system may be integrated into pre-campaign evaluation workflows, particularly for sustainability claims involving certifications, sourcing practices, or environmental impact reduction. In addition, its context-sensitive recommendation structure—attribute-focused guidance for food products and benefit-focused guidance for food services—provides differentiated communication strategies aligned with the nature of the offering, enhancing the practical usability of the system.

Implications for Regulatory and Policy Bodies: For regulatory agencies such as the Office of the Consumer Protection Board (OCPB), GWAMA provides a complementary, behaviorally informed framework for evaluating sustainability claims. By mapping dimensions of perceived greenwashing to provisions of the Consumer Protection Act (B.E. 2522), the system translates legal criteria into observable patterns of consumer interpretation. The item-level compliance alerts—linking exaggerated, misleading, or unverifiable claims to Sections 22(1) and 22(2)—may support more structured screening processes and improve consistency in claim evaluation. While not intended to replace formal regulatory procedures, such tools could assist in early-stage assessment or voluntary self-regulation by firms. At the policy level, the identification of vagueness and unverifiability as key drivers of greenwashing perception provides empirical support for the development of clearer environmental claim standards. Policies encouraging substantiation through certification or measurable indicators may help reduce ambiguity in sustainability communication. These implications are particularly relevant in emerging market contexts, where sustainability marketing is expanding rapidly and regulatory frameworks are still evolving, contributing to broader efforts aligned with Sustainable Development Goal 12 (Responsible Consumption and Production).

Implications for Research: For the research community, this study demonstrates a structured approach to translating validated behavioral models into operational decision support systems. By embedding SEM-derived parameters into a deployable artifact, GWAMA illustrates how psychometrically validated constructs can be extended beyond traditional survey and experimental settings. The proposed scoring architecture—combining factor-loading-weighted composites with regression-based prediction—offers a methodological template for preserving measurement validity while enabling practical application. This approach may be extended to related domains, including sustainability reporting, environmental labeling, and social responsibility communication. The study also highlights the importance of maintaining alignment between theoretical constructs and system implementation. Ensuring traceability between measurement models and artifact design is critical for preserving both rigor and interpretability in applied systems. Future research may extend this work through larger-scale evaluations, cross-cultural validation, and the incorporation of additional communication elements such as visual or symbolic cues.

A particularly important direction for future research concerns the cross-cultural validation and international extension of the GWAMA system. The current behavioral model is grounded in empirical data from Thai food consumers, and while Thailand provides a meaningful initial deployment context as one of Southeast Asia’s most active green marketing research environments [10], the broader transferability of the embedded parameters requires systematic investigation. Cultural values, regulatory environments, consumer awareness levels, and market maturity are known to shape how environmental claims are perceived and interpreted across national contexts. Cross-cultural analyses demonstrate that national culture significantly moderates the relationship between corporate environmental claims and actual environmental behavior at the firm level [35], suggesting that cultural context shapes how sustainability communication is both produced and interpreted across national settings. At the consumer level, research conducted in the United States confirms that green skepticism mediates the relationship between greenwashing claims and consumer behavioral outcomes including brand attitude and purchase intention—a pathway structurally consistent with the S-O-R logic embedded in GWAMA—yet cautions that replication in emerging markets is needed to establish cross-context validity [36]. Future studies should therefore treat the current Thai deployment as an initial naturalistic evaluation episode within a broader research program—consistent with DSR evaluation logic, which requires artifact utility to be demonstrated first in a specific real-world context before generalizability across settings can be systematically assessed [25,37]. Extending the evaluation to diverse national contexts would constitute the next stage of summative validation, enabling assessment of whether the structural path coefficients and population-level standardization parameters remain stable or require context-specific recalibration. Such cross-cultural validation would strengthen the external validity of the GWAMA scoring architecture and contribute to the broader literature on how behavioral models of greenwashing perception generalize—or diverge—across international sustainability communication contexts.

5.5. Limitations and Future Research Directions

While the GWAMA system demonstrates strong theoretical grounding and initial empirical support, several limitations should be acknowledged.

First, the usability evaluation was conducted with a relatively small sample, which limits the generalizability of user acceptance findings. Although appropriate for early-stage design science evaluation, larger-scale studies are needed to confirm these results across broader user populations.

Second, the predictive model is calibrated using a single-country dataset (Thailand), which represents a substantive limitation on the scholarly and empirical generalizability of the system. Cross-cultural differences in consumer behavior, greenwashing perception, and psychological responses to sustainability communication are well-documented in the marketing literature. Perceptions of environmental claims, trust in sustainability messaging, and skepticism toward green advertising are strongly influenced by cultural values, regulatory environments, consumer awareness, and market maturity—factors that vary considerably across national contexts. Consequently, the behavioral assumptions embedded in the GWAMA scoring model—including the S-O-R path coefficients and population-level standardization parameters—may not be fully transferable to other countries or cultural settings without recalibration. For instance, consumers in markets with stronger environmental regulatory frameworks or higher sustainability literacy may exhibit systematically different PGC and GST response patterns compared to the Thai food consumer baseline used in this study. Future research should prioritize cross-cultural validation studies, ideally replicating the SEM estimation across multiple national contexts, to assess the stability and generalizability of the embedded behavioral parameters. Recalibration of model parameters for specific cultural or regional contexts may be necessary before broader international deployment of the system. The behavioral parameters embedded in GWAMA were estimated from Thai food consumers and may reflect culturally specific interpretations of environmental communication, trust formation, and green skepticism; accordingly, caution should be exercised when generalizing the system directly to other cultural or regulatory contexts without additional validation. Consumer interpretation of sustainability claims may also vary according to differences in environmental literacy, market maturity, and exposure to sustainability communication across countries. At its current stage, GWAMA is therefore intended primarily as a context-specific decision support prototype for the Thai food industry rather than a universally calibrated international assessment platform, with cross-national deployment requiring prior validation of the embedded behavioral parameters in the target cultural context.

Third, the system employs a unified predictive model without segmentation by food category or communication channel. While this enhances usability, it may overlook potential heterogeneity in greenwashing perception across contexts.

Fourth, the current implementation focuses on claim-based (textual) assessment and does not incorporate executional greenwashing elements, including implicit visual cues that shape consumer judgment. For example, the use of green packaging has been shown to increase perceived environmental sustainability, even among informed consumers [38]. Beyond perception, such cues can also influence actual consumer behavior, as green-colored packaging has been found to increase product choice and market share in experimental settings [39]. As the field evolves toward AI-assisted multimodal analysis, future iterations of GWAMA should integrate computer vision modules capable of evaluating visual stimuli alongside textual claims, enabling a more comprehensive assessment of greenwashing risk in advertising materials.

Finally, the TAM-based usability evaluation is applied descriptively rather than as a confirmatory structural model. As such, the findings should be interpreted as indicative rather than conclusive, and future research should employ larger samples and longitudinal designs to assess sustained system adoption. Further collaboration with regulatory bodies may also support the integration of such tools into formal governance frameworks aligned with Sustainable Development Goal 12 (Responsible Consumption and Production).

6. Conclusions

This study presents GWAMA, a publicly deployed web-based decision support system designed to operationalize behavioral insights on greenwashing into a practical tool for sustainability communication assessment in the food industry. By integrating a validated Stimulus–Organism–Response (S–O–R) framework into a structured scoring architecture, the system demonstrates how constructs such as perceived greenwashing communication and green skepticism can be translated into real-time, user-accessible diagnostics.

The findings indicate that behavioral models of greenwashing perception can be effectively embedded within a deployable system without compromising their underlying theoretical structure. In particular, the integration of SEM-derived relationships with a factor-loading-weighted scoring approach enables consistent and interpretable prediction of purchase intention, while maintaining robustness across alternative scoring specifications. In addition to predictive functionality, system design highlights the importance of contextual adaptation in sustainability communication. While the predictive model remains unified, the recommendation layer differentiates between food product and food service contexts, enabling more relevant and actionable guidance for practitioners. This reflects a practical balance between model generalizability and application-specific relevance.

Initial usability evaluation with 150 food industry professionals provides indicative evidence of system acceptance, with high levels of perceived usefulness, ease of use, and intention to use. These results suggest that the system is both functionally viable and practically relevant, although further large-scale validation is required to establish generalizability. The live deployment of GWAMA (https://gwamathai.com) enhances the transparency of the research by allowing direct access to the artifact. This supports the design science objective of demonstrating artifact feasibility in real-world contexts and enables external stakeholders to engage with the system beyond controlled experimental settings.

Overall, the contribution of this study lies in demonstrating how validated behavioral models can be translated into an operational governance tool that supports sustainability communication practice. This approach complements existing empirical research by extending its application into decision support environments, particularly in emerging market contexts where sustainability governance mechanisms are still developing.