Advancing occupational and environmental safety in greenhouse agriculture: an extended elaboration likelihood model of safety technology adoption Authors: Amir Alambeigi, Marzieh Keshavarz, Mahla Farzipour, Romina Amiri Published: 2026 ABSTRACT Cleaner greenhouse production depends not only on advanced technologies but also on producers’ adoption of safety innovations that protect workers, consumers, and environmental quality. Despite their importance, safety technologies remain under-adopted compared with productivity-oriented innovations, particularly in developing-country agricultural systems. Existing adoption models provide limited explanation of this paradox because they focus primarily on behavioral outcomes rather than the cognitive mechanisms underlying information evaluation under conditions of occupational, regulatory, and market risk. To address this gap, this study develops and empirically tests an extended Elaboration Likelihood Model (ELM) incorporating perceived risk as a key antecedent of motivation, cognitive ability, and attitude formation. Primary data were collected through a structured questionnaire survey administered to greenhouse owners and managers operating automated or semi-automated production systems in Iran. Using a stratified random sampling approach, 119 valid responses were obtained from major greenhouse production regions. The sample represents technology-exposed producers engaged in commercial greenhouse cultivation. Results from Partial Least Squares Structural Equation Modeling indicate that attitude is the strongest predictor of intention to adopt safety technologies. Perceived risk does not directly influence intention but operates through indirect cognitive pathways by increasing motivation and perceived ability to process information, which activate central and peripheral persuasion routes. While central-route processing leads to systematic evaluation of safety technologies, peripheral cues remain influential under conditions of limited cognitive engagement. Increased motivation reduces reliance on heuristic cues, reinforcing the importance of informed decision-making. These findings highlight the importance of integrated interventions combining risk communication, capacity building, credible endorsements, and supportive policy instruments to promote safer and more sustainable greenhouse production systems. Introduction Greenhouse agriculture has become an increasingly important component of global food production systems due to its ability to enhance yield stability, extend growing seasons, and improve resource-use efficiency. The intensification of protected cultivation systems has enabled farmers to meet rising food demand while maintaining higher levels of control over environmental conditions and production processes1–3. However, this intensification has also led to greater reliance on synthetic chemical inputs such as pesticides and fertilizers, which introduce significant occupational, environmental, and food safety risks4,5. In greenhouse production environments, chemical application practices are often characterized by frequent exposure of workers to hazardous substances through dermal contact, inhalation, and accidental ingestion6–8. Limited adherence to protective protocols, inadequate ventilation systems, and insufficient safety training further exacerbate these risks9–11. In addition to occupational hazards, improper chemical management contributes to environmental contamination and increases the likelihood of pesticide residues entering the food chain12–15, thereby raising concerns for consumer health and regulatory compliance. These challenges highlight the need for more sustainable production approaches that integrate both productivity and safety considerations. In response to these concerns, a range of safety-oriented and environmentally protective technologies has been developed. These include automated and remote-controlled application systems, precision spraying technologies, sensor-based monitoring systems, artificial intelligence (AI)-supported decision tools, and digital traceability platforms16,17. Despite the availability and proven effectiveness of these technologies, their adoption remains limited in many developing-country contexts18,19. This is particularly evident in small and medium-scale greenhouse operations, where financial constraints20–23, limited technical capacity22,24–28, and institutional barriers29,30 restrict investment in safety-related innovations. Interestingly, this low adoption contrasts with the relatively rapid diffusion of productivity-enhancing digital technologies19,31, suggesting that adoption decisions are not driven solely by technical feasibility or economic rationality, but are also influenced by behavioral and cognitive factors. This divergence raises an important theoretical and practical question: why do greenhouse producers adopt technologies that enhance productivity more readily than those designed specifically to reduce occupational and environmental risks? Addressing this question is critical for improving the effectiveness of environmental governance, occupational safety interventions, and sustainable agricultural development strategies. To address this issue, this study investigates the determinants of safety technology adoption intentions among greenhouse producers in Iran. Specifically, it adopts the Elaboration Likelihood Model (ELM) as a theoretical lens to explain how producers process information related to occupational and environmental risks when forming attitudes toward safety technologies. The ELM is particularly suitable for this context because it explains how individuals are persuaded through central-route processing, based on argument quality, and peripheral-route processing, based on heuristic cues. Building on this framework, the study introduces perceived risk as a key antecedent shaping individuals’ motivation and cognitive engagement with safety-related information. In the context of chemical-intensive greenhouse production, perceived risk reflects producers’ awareness of potential occupational hazards, environmental damage, and food safety consequences associated with current practices. Higher perceived risk is expected to increase the relevance of safety-related information and thereby influence attitude formation and behavioral intention regarding safety technology adoption. Despite extensive research on technology adoption in agriculture, existing studies have primarily focused on structural determinants such as cost, performance expectancy, and facilitating conditions, while paying limited attention to the cognitive and informational processes through which risk perceptions influence decision-making. Moreover, empirical evidence on safety technology adoption in developing-country greenhouse systems remains limited, despite the heightened relevance of occupational and environmental risks in these contexts. Accordingly, this study makes three key contributions. First, it extends the application of the Elaboration Likelihood Model to safety technology adoption in greenhouse agriculture by incorporating perceived risk as a central explanatory construct. Second, it provides empirical evidence from a developing-country context characterized by resource constraints and elevated environmental risks. Third, it offers practical insights for policymakers, extension services, and technology providers on how risk communication strategies can enhance the adoption of safety-oriented innovations. The remainder of the paper is structured as follows. The next section presents the literature review and theoretical framework, followed by hypothesis development. The methodology, results, and discussion sections are then presented, concluding with implications and future research directions. Literature review and hypotheses development Safety technologies, cleaner production, and agricultural risk management Safety technologies in agriculture refer to a set of technological innovations designed to reduce occupational exposure, minimize environmental contamination, and improve food safety outcomes in production systems. In greenhouse agriculture, such technologies include precision spraying systems, automated pesticide application devices, sensor-based monitoring platforms, artificial intelligence (AI)-supported decision tools, and digital traceability systems32–36. These technologies collectively aim to reduce dependence on manual chemical handling and enhance control over pesticide use intensity and timing. Within the broader sustainability discourse, these innovations are closely associated with the concept of cleaner production, which emphasizes preventive environmental strategies aimed at minimizing risks to human health and the environment while improving resource efficiency and production sustainability. In greenhouse systems, cleaner production is operationalized through reduced pesticide application, improved targeting of chemical inputs, enhanced monitoring of environmental conditions, and increased transparency in production processes. Despite their technical effectiveness, adoption of such technologies remains uneven across developing-country agricultural systems, limiting their contribution to environmental and occupational risk management. Technology adoption theories in agricultural innovation research Technology adoption in agriculture has been extensively examined through behavioral and innovation diffusion perspectives. Among the most widely applied frameworks are the Perceived Attributes Model (PAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT), both of which provide valuable but partial explanations of adoption behavior. Perceived attributes model (PAM) The Perceived Attributes Model, derived from innovation diffusion theory, posits that adoption decisions are influenced by perceived characteristics of innovations, including relative advantage, compatibility, complexity, trialability, and observability. In agricultural contexts, PAM has been widely used to explain farmers’ adoption of new technologies based on how they evaluate the innovation’s functional and economic benefits37,38. However, PAM primarily focuses on structural attributes of innovations and assumes that individuals make relatively rational evaluations of technology characteristics. It provides limited explanation of the cognitive and psychological processes through which individuals interpret risk-related information or form attitudes under uncertainty39,40. Consequently, PAM is less effective in explaining adoption behavior in contexts where risk perception plays a central role, such as occupational and environmental safety in greenhouse production. Unified theory of acceptance and use of technology (UTAUT) The Unified Theory of Acceptance and Use of Technology (UTAUT) integrates multiple behavioral models, including the Technology Acceptance Model (TAM), Theory of Reasoned Action (TRA), and Theory of Planned Behavior (TPB). UTAUT explains behavioral intention through four core constructs: performance expectancy, effort expectancy, social influence, and facilitating conditions, moderated by demographic and contextual variables40,41. In agricultural innovation research, UTAUT has been widely applied to understand farmers’ adoption of digital tools, smart farming technologies, and information systems. The model is particularly effective in capturing perceptions of usefulness, ease of use, and social pressures. Nevertheless, UTAUT remains primarily outcome-oriented and does not explicitly model how individuals process information or how risk-related cognition influences decision formation. In particular, it offers limited insight into how occupational hazards, environmental risks, or food safety concerns shape motivation and cognitive engagement in adoption decisions. This limitation is particularly relevant for safety technologies, where adoption is driven not only by performance considerations but also by risk avoidance motivations. Elaboration likelihood model (ELM) and information processing in adoption decisions The Elaboration Likelihood Model (ELM) provides a process-oriented explanation of attitude formation and change through two distinct cognitive routes: the central route and the peripheral route. Under the central route, individuals engage in careful and systematic evaluation of message content, arguments, and evidence. Under the peripheral route, attitudes are formed based on heuristic cues such as source credibility, social endorsement, or simplicity of messages when motivation or cognitive ability is limited42,43. Unlike PAM and UTAUT, which emphasize determinants of behavioral intention, ELM explains how individuals process information before forming attitudes, making it particularly suitable for contexts where decisions depend on interpretation of complex and uncertain information44,45. In agricultural safety contexts, ELM is highly relevant because producers are often exposed to mixed information regarding risks, benefits, and uncertainties associated with both chemical use and alternative technologies. Despite its usefulness, the original formulation of ELM does not explicitly define the antecedents of motivation and ability to process information. This limitation is critical in safety-related decision contexts, where external risk conditions strongly influence cognitive engagement. Perceived risk as a key antecedent in extended ELM framework Perceived risk refers to individuals’ subjective assessment of potential losses or negative consequences associated with a specific behavior or decision. In agricultural contexts, perceived risk encompasses occupational health risks, environmental risks, economic uncertainty, and food safety concerns arising from production practices. In the context of greenhouse agriculture, perceived risk associated with chemical use plays a central role in shaping producers’ cognitive engagement with safety-related information. When producers perceive higher levels of occupational or environmental risk, they are more likely to consider safety-related messages as personally relevant46,47, thereby increasing their motivation to engage in central-route processing as described in ELM. In addition, perceived risk can enhance perceived cognitive involvement by increasing attention to risk mitigation strategies and improving willingness to seek information about protective technologies. This suggests that perceived risk not only influences attitude formation indirectly but also shapes the depth and direction of information processing. However, the relationship between perceived risk and behavioral intention is not necessarily direct or linear29,48. Prior research indicates that even when risk perception is high, adoption may remain limited due to financial constraints, lack of trust, or perceived inefficacy of available technologies29,49,50. This indicates that perceived risk must be integrated into a broader cognitive framework that explains how risk information is processed and translated into attitudes and intentions. Accordingly, this study extends the ELM framework by positioning perceived risk as a central antecedent that influences both motivation and cognitive engagement in processing safety-related information. This extension provides a more comprehensive explanation of how risk awareness translates into behavioral intention through structured cognitive pathways. Hypotheses development Perceived risk and information processing in safety contexts Within the Elaboration Likelihood Model (ELM), attitude formation is determined by the degree to which individuals engage in effortful cognitive processing of persuasive information. This level of elaboration is primarily driven by motivation and ability. In the context of greenhouse agriculture, perceived risk associated with chemical use constitutes a critical motivational factor that determines whether producers engage deeply with safety-related information or rely on heuristic cues. When greenhouse producers perceive higher levels of occupational exposure, environmental contamination, or food safety risks, the relevance of safety technologies becomes more personally salient. This heightened salience increases cognitive involvement and encourages systematic evaluation of risk mitigation alternatives. Accordingly, perceived risk is expected to enhance motivation to process information. H1 Perceived risk positively influences motivation to process information in the context of safety technology adoption. In addition to motivation, perceived risk may also affect perceived cognitive engagement ability by prompting producers to seek additional information, consult external sources, and enhance their understanding of safety technologies. As individuals become more concerned about potential losses, they are more likely to invest cognitive effort in understanding technological solutions. H2 Perceived risk positively influences perceived ability to process safety-related information. Central and peripheral route processing and attitude formation According to ELM, attitude formation is primarily determined by the route through which individuals process information. When individuals engage in central-route processing, they carefully evaluate the quality, relevance, and strength of arguments related to safety technologies. In the context of greenhouse agriculture, producers who systematically assess evidence regarding the effectiveness of safety technologies are more likely to develop stable and favorable attitudes toward adoption. Central-route processing leads to stronger, more persistent, and behaviorally predictive attitudes because it is based on analytical evaluation rather than superficial cues. H3 Central-route information processing positively influences attitude toward safety technology adoption. In contrast, peripheral-route processing is based on heuristic cues such as perceived credibility of information sources, recommendations from peers, and institutional endorsement. In agricultural decision-making environments characterized by uncertainty and limited technical capacity, producers may rely on such cues when cognitive resources or motivation for deep processing are limited. Peripheral cues can still shape favorable attitudes toward safety technologies, although these attitudes are typically less stable than those formed through central processing. H4 Peripheral-route cues positively influence attitude toward safety technology adoption. Perceived risk and attitude toward safety technologies Beyond its role in shaping ability to process information, perceived risk is also expected to indirectly influence attitude formation. Producers who perceive higher occupational and environmental risks associated with chemical use are more likely to evaluate safety technologies positively, as these technologies are viewed as mechanisms for risk mitigation and harm reduction. However, this relationship is contingent upon the perceived effectiveness of safety technologies and contextual constraints such as cost and accessibility. In general, higher perceived risk increases the perceived necessity and value of safety innovations, thereby strengthening favorable attitudes. H5 Perceived risk positively influences attitude toward safety technology adoption. Attitude and behavioral intention Attitude toward a technology represents an individual’s overall evaluative judgment regarding its adoption, encompassing perceived benefits, risks, and compatibility with existing practices. Within behavioral theory, attitude is one of the most consistent predictors of behavioral intention. In the context of safety technology adoption, a favorable attitude reflects a belief that such technologies are beneficial for reducing occupational hazards, improving environmental outcomes, and ensuring compliance with safety standards. Consequently, individuals with more positive attitudes are more likely to intend to adopt such technologies. H6 Attitude toward safety technology adoption positively influences behavioral intention to adopt safety technologies. Materials and methods Study area Greenhouse agriculture in Iran has expanded rapidly over the past decade in response to increasing water scarcity, climate variability, and the need to improve land and input productivity while maintaining a stable, year-round food supply. In 2022, the total greenhouse area reached approximately 18,239 hectares, producing over 1.3 million tons of vegetables and horticultural crops, including cucumber, tomato, bell pepper, and leafy greens51. This expansion reflects a strategic shift toward more resource-efficient production systems, however, it also concentrates chemical use, labor exposure, and environmental risks within enclosed production environments. These characteristics make greenhouse systems a critical focal point for managing occupational health, food safety, and cleaner-production outcomes. Greenhouse development in Iran is spatially uneven and concentrated in regions with favorable agro-climatic and infrastructural conditions, particularly in Kerman and Fars provinces (Fig. 1). Mild winter temperatures, high solar radiation, and extended frost-free periods support off-season vegetable production while reducing heating requirements and associated energy consumption. Within these provinces, greenhouse units are clustered in fertile plains, river corridors, and areas with relatively reliable groundwater access, benefiting from adequate soil depth, suitable textures, and manageable water tables. Altitude further shapes spatial distribution, as lower-elevation areas experience frost-free winters that reduce energy demand for climate control. In addition, infrastructure availability and market access strongly influence location choices, with many facilities situated near major transport corridors to facilitate timely delivery of perishable produce to wholesale markets, packing centers, and export terminals. Collectively, this cluster-based and infrastructure-oriented spatial pattern supports economic competitiveness but also concentrates environmental and occupational risks, increasing the importance of effective risk management practices. Fig. 1Provincial suitability for vegetable and horticultural production in Iran. Darker color shades indicate higher suitability levels. The map was generated using ArcGIS Desktop version 10.8 based on spatial data obtained from the Ministry of Jihad Agriculture of Iran. Provincial suitability for vegetable and horticultural production in Iran. Darker color shades indicate higher suitability levels. The map was generated using ArcGIS Desktop version 10.8 based on spatial data obtained from the Ministry of Jihad Agriculture of Iran. From a technological standpoint, greenhouse vegetable systems in the study region are predominantly equipped with basic to semi-advanced environmental control technologies, including thermostats, simple sensors, timers, and control units used to manage temperature, humidity, ventilation, and irrigation. These systems contribute to yield stability, reduced input use, and improved production efficiency52. In recent years, interest in smart greenhouse solutions—such as locally developed automation and decision-support systems—has increased, supported by domestic knowledge-based firms and policy incentives for technological upgrading. Nevertheless, adoption of safety-oriented technologies remains limited. Technologies designed to reduce agrochemical exposure, monitor spraying operations, protect worker health, and prevent chemical contamination are not widely implemented, despite their relevance to occupational health and food-safety management. This gap is particularly consequential in southern Iran, where intensive pesticide use, enclosed production environments, and year-round cultivation amplify risks related to chemical exposure and residue accumulation. As a result, Kerman and Fars provinces represent highly relevant settings for examining how greenhouse managers perceive and respond to environmental and occupational risks, and how these perceptions influence engagement with safety-related information and technologies. Studying these regions therefore provides a suitable empirical basis for analyzing behavioral mechanisms that condition the effectiveness of environmental risk management interventions in intensive greenhouse production systems. Sampling and survey instrument The target population consisted of owners and managers of commercial greenhouse operations in southern Iran that had installed and actively operated automated production systems, such as fertigation and irrigation control technologies. This population was purposively defined to capture producers who had already demonstrated engagement with technological solutions for productivity enhancement and resource-use management. Focusing on this group allowed the study to isolate the cognitive and persuasive factors influencing the uptake of safety-oriented technologies, rather than general barriers to basic mechanization. Greenhouse owners operating purely traditional (non-automated) systems and non-operational facilities were excluded to ensure comparability in technological baseline and decision-making capacity. In addition, in this study, “safety technologies” were conceptualized as a unified construct encompassing any technological system or tool designed to reduce occupational exposure to agrochemicals, minimize environmental contamination, and enhance food safety in greenhouse production. This includes a broad spectrum of solutions ranging from basic automated systems (e.g., timers, thermostats, fertigation controllers) to more advanced technologies (e.g., precision spraying systems, sensor-based monitoring, and digital traceability tools). Respondents were therefore asked to evaluate their perceptions and adoption intentions toward safety technologies as an integrated category rather than as discrete technological subtypes. The sampling frame was constructed using contractor installation records, which provided verified information on greenhouse location, size, and automation status. A stratified random sampling strategy was employed, with strata defined by province (Kerman and Fars) and greenhouse area. Stratification by area was methodologically justified because greenhouse size is closely associated with capital availability, access to finance, labor intensity, and managerial capacity—factors that condition both environmental risk exposure and the ability to evaluate and adopt safety technologies. Within each stratum, respondents were randomly selected to minimize selection bias. Eligibility was confirmed through preliminary phone contact, and appointments were arranged for survey administration. The final analytical sample comprised 119 completed questionnaires. Sample adequacy was evaluated using two complementary approaches commonly applied in structural equation modeling (SEM). First, the “10-times rule”53 was applied, requiring a minimum sample size equal to ten times the maximum number of structural paths directed at any latent construct. As the maximum number of incoming paths in the model was four, the minimum required sample size was 40 observations. Second, a post-hoc statistical power analysis was conducted using G*Power 3.154. Assuming a medium effect size (f2 = 0.15), a significance level of α = 0.05, and the observed number of predictors, the analysis yielded statistical power (1 − β) exceeding 0.80, indicating sufficient sensitivity to detect meaningful relationships. Model fit indices and parameter stability were further examined to confirm robustness. This study employed an anonymous and voluntary questionnaire survey administered to adult participants and did not involve clinical interventions, experimental treatments, biological sampling, collection of sensitive personal information, or participation of vulnerable populations. The study protocol was reviewed by the Research Ethics Committee of the College of Agriculture and Natural Resources, University of Tehran, which determined that the research was exempt from formal ethical review and approval because it constituted non-interventional survey research involving competent adults, with no collection of personally identifiable or sensitive information. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and applicable institutional and national regulations governing research involving human participants. Prior to data collection, all participants were informed about the objectives of the study, the voluntary nature of participation, their right to refuse or discontinue participation at any time without consequence, and the confidentiality and anonymity of their responses. Verbal informed consent was obtained from all participants before the interview commenced. Data were collected using a structured questionnaire administered through face-to-face interviews conducted by trained enumerators. Given the technical nature of the research, particularly concepts related to occupational chemical exposure, safety technologies, and information processing, interviewers followed a standardized survey protocol. This protocol included a brief explanation of the study objectives, clarification of key technical concepts, and confirmation that the research focused exclusively on management and decision-making practices rather than regulatory compliance. These procedures were implemented to improve respondent comprehension, enhance response accuracy, and minimize measurement error. To evaluate potential non-response bias, the characteristics of early and late respondents were compared across key observable variables. No statistically significant differences were identified, suggesting that non-response bias was unlikely to affect the validity of the findings. Measurement All latent constructs were measured using five-point Likert scales (1 = strongly disagree to 5 = strongly agree) adapted from previously validated instruments in the technology adoption, risk perception, and persuasion literature. The questionnaire was translated into Persian using a forward–backward translation procedure involving professional translation, independent back-translation, and reconciliation by the research team in consultation with bilingual subject-matter experts. The operationalization aligned with the theoretical components of the extended ELM and complementary adoption frameworks (PAM and UTAUT), with constructs including: (i) Central route cues: Argument quality, perceived relevance of safety technologies; (ii) Peripheral route cues: Source credibility, peer influence, normative cues; (iii) Perceived risk: Susceptibility, severity, and uncertainty of chemical exposure; (iv) Attitude toward safety technologies; and (v) Intention to adopt safety technologies. Variable definitions and measurement items are reported in Table 1. Table 1Study variables.VariablesNumber of itemsExplanationReference(s)/theoryPerceived risk (PR)5The owner’s subjective assessment of negative consequences from not adopting safety technologiesWang and Zhao55, Hu et al.56Motivation for processing (MPI)4The owner’s intrinsic drive to engage in thoughtful evaluation of safety information (personal relevance, need for cognition). Exemplar: “I actively seek detailed information before deciding about new safety technologies.”Susanto et al.57, Shin58Ability to process information (API)4Cognitive resources and situational conditions that enable elaboration (technical knowledge, time availability). Exemplar: “I have enough technical knowledge to evaluate whether a safety technology will work on my farmKim and Park59Central route processing (CRP)4Systematic, effortful evaluation of diagnostic information about safety technologies. Items: a) I carefully evaluate the quality and effectiveness of safety technologies before considering adoption, b) I rely on scientific evidence and research findings when assessing safety technologies, c) Logical and well-supported arguments about risk reduction strongly influence my judgment, and d) My evaluation of safety technologies is based primarily on their actual effectiveness in reducing occupational hazardsZhang and Benbasat60Peripheral route processing (PRP)4Heuristic-based evaluation relying on social endorsement, authority signals, and superficial attributes. a) I tend to rely on opinions of trusted farmers or experts when evaluating safety technologies, b) If a reputable organization recommends a safety technology, I am more likely to adopt it, c) The visual appearance or modern design of a safety technology influences my perception of it, and d) If a safety technology appears easy to use, I am more likely to prefer itShen and Bigman61Attitude toward safety technologies (AST)4An individual’s positive or negative mindset, beliefs, and feelings about using technological tools designed to prevent harmTRA-TPB constructsIntention to adopt safety technologies (IAST)4Self-reported likelihood and readiness to implement safety technologies within a defined future timeframeUTAUT/TAM intention measures Table 1Study variables.VariablesNumber of itemsExplanationReference(s)/theoryPerceived risk (PR)5The owner’s subjective assessment of negative consequences from not adopting safety technologiesWang and Zhao55, Hu et al.56Motivation for processing (MPI)4The owner’s intrinsic drive to engage in thoughtful evaluation of safety information (personal relevance, need for cognition). Exemplar: “I actively seek detailed information before deciding about new safety technologies.”Susanto et al.57, Shin58Ability to process information (API)4Cognitive resources and situational conditions that enable elaboration (technical knowledge, time availability). Exemplar: “I have enough technical knowledge to evaluate whether a safety technology will work on my farmKim and Park59Central route processing (CRP)4Systematic, effortful evaluation of diagnostic information about safety technologies. Items: a) I carefully evaluate the quality and effectiveness of safety technologies before considering adoption, b) I rely on scientific evidence and research findings when assessing safety technologies, c) Logical and well-supported arguments about risk reduction strongly influence my judgment, and d) My evaluation of safety technologies is based primarily on their actual effectiveness in reducing occupational hazardsZhang and Benbasat60Peripheral route processing (PRP)4Heuristic-based evaluation relying on social endorsement, authority signals, and superficial attributes. a) I tend to rely on opinions of trusted farmers or experts when evaluating safety technologies, b) If a reputable organization recommends a safety technology, I am more likely to adopt it, c) The visual appearance or modern design of a safety technology influences my perception of it, and d) If a safety technology appears easy to use, I am more likely to prefer itShen and Bigman61Attitude toward safety technologies (AST)4An individual’s positive or negative mindset, beliefs, and feelings about using technological tools designed to prevent harmTRA-TPB constructsIntention to adopt safety technologies (IAST)4Self-reported likelihood and readiness to implement safety technologies within a defined future timeframeUTAUT/TAM intention measures Study variables. Content validity and contextual relevance were evaluated by a panel of five experts, including agricultural extension specialists, occupational health researchers, and greenhouse technology practitioners. A pilot survey (n = 30), conducted outside the main study sample, was used to test item clarity, response distributions, and survey logistics. Based on pilot feedback, ambiguous or redundant items were revised or removed. Also, indicator reliability, internal consistency (Cronbach’s α, CR), convergent validity (AVE), and discriminant validity (Fornell-Larcker, HTMT) were confirmed in the “Results” section. Data analysis Data analysis followed a structured PLS-SEM approach, designed to evaluate: (i) Measurement model: reliability, convergent validity, discriminant validity, (ii) Structural model: paths implied by the extended ELM, hypothesis testing, and (iii) Model robustness and predictive relevance. Prior to estimation, the dataset was examined for completeness, plausibility, and response distributions. Missing values (< 5%) were imputed using item-level means, and sensitivity analyses confirmed that substantive conclusions were robust to the imputation approach. Univariate and multivariate outliers were identified and analyzed both with and without extreme cases to assess the stability of parameter estimates, ensuring that inferences regarding risk perception, attitude formation, and adoption intention were not unduly influenced by anomalous observations. The measurement model was evaluated to confirm indicator reliability, internal consistency, convergent validity, and discriminant validity. Indicator reliability was assessed via standardized loadings, with items < 0.40 considered for removal and items between 0.40 and 0.70 evaluated based on theoretical justification and contribution to composite reliability. Internal consistency was assessed using Cronbach’s alpha and composite reliability (CR ≥ 0.70). Convergent validity was confirmed using Average Variance Extracted (AVE > 0.50), and discriminant validity was evaluated using the Fornell–Larcker criterion and cross-loadings. Full measurement results are reported in the “Results” section. Given the self-reported, cross-sectional nature of the data, potential common method variance (CMV) was addressed through both procedural and statistical remedies. Procedurally, questionnaires ensured respondent anonymity, conceptually separated distinct constructs, and counterbalanced item order to minimize bias. Statistically, Harman’s single-factor test, a marker-variable approach, and the inclusion of a latent common method factor in the PLS model were employed to detect CMV. All substantive conclusions were verified to be robust to these adjustments, ensuring that observed relationships reflect genuine behavioral and risk-management processes rather than methodological artifacts. Finally, the structural model was evaluated following standard PLS-SEM procedures. First, multicollinearity among predictors was assessed using variance inflation factors (VIF < 3.3). Second, path coefficients were estimated, and statistical significance was tested via bootstrapping with 5000 resamples. Third, key model-fit and explanatory statistics were reported to evaluate the adequacy of the model in representing the decision-making processes underlying safety-technology adoption. Results Sample characteristics The sample consists of 119 greenhouse owners and managers with a mean age of 43.3 years and a standard deviation of 11.2 years. Participants reported an average of 6.4 years of professional experience in greenhouse production with a standard deviation of 5.5 years. Educational attainment was relatively low to moderate, as 75.6 percent of respondents held a diploma or lower qualification, while 16.8 percent held a bachelor’s degree and 7.6 percent held postgraduate degrees. Crop production systems were diverse, including cucumber (32.8%), strawberry (21.8%), bell pepper (18.5%), leafy vegetables (8.4%), eggplant (0.8%), and mixed cultivation systems (17.6%), indicating heterogeneity in production orientation. Operational scale varied substantially across respondents, with a strongly skewed distribution characterized by a mean greenhouse area of 22,255 square meters and a median of 5000 square meters. A small proportion of respondents operated extremely large facilities exceeding 100,000 square meters. The sample is further characterized by a relatively high level of technological exposure, as respondents predominantly operate within greenhouse systems that include some degree of automation or mechanization. This indicates that the sample represents a technologically experienced subgroup within the broader greenhouse sector. Perceived risk, attitudes, and intention to adopt safety technologies Table 2 presents respondents’ evaluations of agrochemical risks, attitudes toward safety technologies, and intentions to adopt such technologies. Overall, greenhouse owners reported moderate levels of perceived risk associated with agrochemical use. Among the perceived-risk indicators, concerns related to acute poisoning consequences received relatively higher ratings, whereas perceptions regarding the long-term health effects of pesticide exposure and the unsafety of current spraying practices were comparatively lower. In several risk-related items, a substantial proportion of respondents selected neutral response categories, indicating considerable variation in risk perceptions across the sample. Table 2Descriptive analysis of perceived risk, attitudes, and intention to adopt safety technologies.ItemsStrongly disagreeDisagreeNeutralAgreeStrongly agreePerceived risk I believe that pesticide use in greenhouses poses a serious health risk to workers36 (30.3%)9 (7.6)25 (21.0)34 (28.6)15 (12.6) I am concerned about the long-term consequences of exposure to pesticides25 (21)26 (21.8)42 (35.3)18 (15.1)8 (6.7) I think the likelihood of pesticide poisoning in our greenhouse is high26 (21.8)20 (16.8)19 (16.0)40 (33.6)14 (11.8) I think that the current spraying methods in our greenhouse are not sufficiently safe25 (21)26 (21.8)47 (39.5)19 (16.0)2 (1.7) The probability that I or my colleagues will experience health problems from pesticide use is high37 (31.1)2 (1.7)40 (33.6)40 (33.6)0 (0.0)Attitude toward safety technologies Overall, I hold a favorable view toward the adoption of safety technologies in greenhouse production1 (0.8)9 (7.6)33 (27.7)51 (42.9)25 (21.0) Investing in safety technologies represents a valuable and worthwhile investment for greenhouse operations0 (0.0)1 (0.8)44 (37.0)35 (29.4)39 (32.8) The adoption of safety technologies reflects a strong commitment to protecting workers’ health and safety0 (0.0)9 (7.6)27 (22.7)53 (44.5)30 (25.2) I have a positive emotional response toward using safety technologies in my greenhouse0 (0.0)12 (10.1)19 (16.0)28 (23.5)60 (50.4)Intention to adopt safety technologies I intend to use safety technologies in the near future0 (0.0)1 (0.8)19 (16)46 (38.7)53 (44.5) I actively seek information about how to procure and use safety technologies1 (0.8)14 (11.8)14 (11.8)35 (29.4)55 (46.2) I am willing to pay to use safety technologies2 (1.7)1 (0.8)26 (21.8)42 (35.3)48 (40.3) I am likely to recommend safety technologies to colleagues1 (0.8)15 (12.6)11 (9.2)34 (28.6)58 (48.7) Table 2Descriptive analysis of perceived risk, attitudes, and intention to adopt safety technologies.ItemsStrongly disagreeDisagreeNeutralAgreeStrongly agreePerceived risk I believe that pesticide use in greenhouses poses a serious health risk to workers36 (30.3%)9 (7.6)25 (21.0)34 (28.6)15 (12.6) I am concerned about the long-term consequences of exposure to pesticides25 (21)26 (21.8)42 (35.3)18 (15.1)8 (6.7) I think the likelihood of pesticide poisoning in our greenhouse is high26 (21.8)20 (16.8)19 (16.0)40 (33.6)14 (11.8) I think that the current spraying methods in our greenhouse are not sufficiently safe25 (21)26 (21.8)47 (39.5)19 (16.0)2 (1.7) The probability that I or my colleagues will experience health problems from pesticide use is high37 (31.1)2 (1.7)40 (33.6)40 (33.6)0 (0.0)Attitude toward safety technologies Overall, I hold a favorable view toward the adoption of safety technologies in greenhouse production1 (0.8)9 (7.6)33 (27.7)51 (42.9)25 (21.0) Investing in safety technologies represents a valuable and worthwhile investment for greenhouse operations0 (0.0)1 (0.8)44 (37.0)35 (29.4)39 (32.8) The adoption of safety technologies reflects a strong commitment to protecting workers’ health and safety0 (0.0)9 (7.6)27 (22.7)53 (44.5)30 (25.2) I have a positive emotional response toward using safety technologies in my greenhouse0 (0.0)12 (10.1)19 (16.0)28 (23.5)60 (50.4)Intention to adopt safety technologies I intend to use safety technologies in the near future0 (0.0)1 (0.8)19 (16)46 (38.7)53 (44.5) I actively seek information about how to procure and use safety technologies1 (0.8)14 (11.8)14 (11.8)35 (29.4)55 (46.2) I am willing to pay to use safety technologies2 (1.7)1 (0.8)26 (21.8)42 (35.3)48 (40.3) I am likely to recommend safety technologies to colleagues1 (0.8)15 (12.6)11 (9.2)34 (28.6)58 (48.7) Descriptive analysis of perceived risk, attitudes, and intention to adopt safety technologies. In contrast, respondents generally expressed favorable attitudes toward safety technologies. Most attitude indicators exhibited positive evaluations, suggesting that greenhouse owners largely viewed safety technologies as beneficial and worthwhile investments. Similarly, intention-to-adopt indicators demonstrated a positive orientation toward future adoption. Respondents reported a willingness to seek information about available technologies and expressed readiness to consider implementing safety solutions within their greenhouse operations. Drivers of intention to adopt safety technologies Measurement model assessment The measurement model was evaluated using Partial Least Squares Structural Equation Modeling. Indicator reliability was assessed through outer loadings, most of which exceeded the recommended threshold of 0.70 (Table 3). Two indicators (MPI4 and PRP4) exhibited outer loadings below the recommended threshold of 0.50 and were therefore removed prior to estimating the final measurement model. The remaining indicators demonstrated satisfactory reliability and validity. Also, internal consistency reliability was confirmed as Cronbach’s alpha, composite reliability, and Dijkstra–Henseler’s rho values exceeded the minimum threshold of 0.70 for all constructs. Table 3Measures of reliability and validity.Variables and itemsFactor loadingCronbach’s alphaComposite reliability (rho_a)Composite reliability (rho_c)Average variance extracted (AVE)Motivation to process information (MPI) MPI10.9480.7980.8330.8770.652 MPI20.870 MPI30.850 MPI40.481Ability to process information (API) API10.7570.8590.8590.9050.704 API20.860 API30.884 API40.849Central route processing (CRP) CRP10.8440.7890.8080.8640.615 CRP20.800 CRP30.658 CRP40.822Peripheral route processing (PRP) PRP10.9150.7920.8160.8750.700 PRP20.791 PRP30.798 PRP40.381Attitude toward safety technologies (AST) AST10.9010.8820.8860.9200.744 AST20.729 AST30.933 AST40.874Intention to adopt safety technologies (IAST) IAST10.9310.9240.9440.9470.817 IAST20.948 IAST30.768 IAST40.956Perceived risk (PR) PR10.9680.9590.9670.9690.861 PR20.893 PR30.946 PR40.919 PR50.911 Table 3Measures of reliability and validity.Variables and itemsFactor loadingCronbach’s alphaComposite reliability (rho_a)Composite reliability (rho_c)Average variance extracted (AVE)Motivation to process information (MPI) MPI10.9480.7980.8330.8770.652 MPI20.870 MPI30.850 MPI40.481Ability to process information (API) API10.7570.8590.8590.9050.704 API20.860 API30.884 API40.849Central route processing (CRP) CRP10.8440.7890.8080.8640.615 CRP20.800 CRP30.658 CRP40.822Peripheral route processing (PRP) PRP10.9150.7920.8160.8750.700 PRP20.791 PRP30.798 PRP40.381Attitude toward safety technologies (AST) AST10.9010.8820.8860.9200.744 AST20.729 AST30.933 AST40.874Intention to adopt safety technologies (IAST) IAST10.9310.9240.9440.9470.817 IAST20.948 IAST30.768 IAST40.956Perceived risk (PR) PR10.9680.9590.9670.9690.861 PR20.893 PR30.946 PR40.919 PR50.911 Measures of reliability and validity. Convergent validity was established as all Average Variance Extracted values were above 0.50 (Table 3). Discriminant validity was confirmed using the Fornell–Larcker criterion (Table 4), as each construct demonstrated greater shared variance with its own indicators than with other constructs. Table 4Discriminant validity of variables (Fornell-Larcker criterion).ConstructsAPIASTCRPIASTMPIPRPRPAbility to process information (API) 0.839 Attitude toward safety technologies (AST)0.649 0.863 Central route processing (CRP)0.7910.741 0.784 Intention to adopt safety technologies (IAST)0.7260.8080.750 0.904 Motivation to process information (MPI)0.5940.3000.2960.374 0.808 Perceived risk (PR)0.5010.2970.2240.3440.795 0.928 Peripheral route processing (PRP)0.4950.7500.7010.6570.0610.156 0.837 Significant values are in bold. Table 4Discriminant validity of variables (Fornell-Larcker criterion).ConstructsAPIASTCRPIASTMPIPRPRPAbility to process information (API) 0.839 Attitude toward safety technologies (AST)0.649 0.863 Central route processing (CRP)0.7910.741 0.784 Intention to adopt safety technologies (IAST)0.7260.8080.750 0.904 Motivation to process information (MPI)0.5940.3000.2960.374 0.808 Perceived risk (PR)0.5010.2970.2240.3440.795 0.928 Peripheral route processing (PRP)0.4950.7500.7010.6570.0610.156 0.837 Significant values are in bold. Discriminant validity of variables (Fornell-Larcker criterion). Significant values are in bold. Structural model results The structural model (Fig. 2; Tables 5, 6 and 7) was estimated using Partial Least Squares Structural Equation Modeling with 5,000 bootstrap resamples and bias-corrected confidence intervals. All significant path coefficients exhibited confidence intervals that did not include zero (Tables 5 and 6). Also, variance inflation factor (VIF) diagnostics indicated no problematic multicollinearity, with most predictors ≈ 1.0 and predictors of attitude (AST) ranging 1.97–1.98, confirming stable estimates and reliable parameterization. The model explained 65% of the variance in attitude toward safety technologies and 78% of the variance in intention to adopt safety technologies (Table 5), indicating substantial explanatory power. Fig. 2Drivers of intention to adopt safety technologies. Table 5Direct path estimates and predictive power of the structural model.PathPath coefficient95% CIt-valueP valueVIFR2R2AdjPR → API0.500[0.389, 0.600]9.4700.0011.000.250.25PR → MPI0.847[0.782, 0.905]27.1750.0011.000.720.72API → CRP0.788[0.716, 0.859]21.3800.0011.000.620.62MPI → PRP− 0.361[− 0.603, − 0.122]2.8600.0011.000.080.08API → PRP0.692[0.404, 0.974]4.6370.0011.00CRP → AST0.433[0.229, 0.616]4.4120.0011.970.650.65PRP → AST0.444[0.263, 0.629]4.6610.0011.98AST → IAST0.884[0.858, 0.914]59.8640.0011.000.780.78 Drivers of intention to adopt safety technologies. Table 5Direct path estimates and predictive power of the structural model.PathPath coefficient95% CIt-valueP valueVIFR2R2AdjPR → API0.500[0.389, 0.600]9.4700.0011.000.250.25PR → MPI0.847[0.782, 0.905]27.1750.0011.000.720.72API → CRP0.788[0.716, 0.859]21.3800.0011.000.620.62MPI → PRP− 0.361[− 0.603, − 0.122]2.8600.0011.000.080.08API → PRP0.692[0.404, 0.974]4.6370.0011.00CRP → AST0.433[0.229, 0.616]4.4120.0011.970.650.65PRP → AST0.444[0.263, 0.629]4.6610.0011.98AST → IAST0.884[0.858, 0.914]59.8640.0011.000.780.78 Direct path estimates and predictive power of the structural model. Table 6Indirect effects of perceived risk on intention to adopt safety technologies.Indirect pathwayStructural chainIndirect effect (β)95% CIt-valueP valueCentral-route mediationPR → API → CRP → AST → IAST0.151[0.074, 0.236]2.090.04Peripheral-route (ability-based) mediationPR → API → PRP → AST → IAST0.136[0.056, 0.258]3.050.000Peripheral-route suppression (motivation-based)PR → MPI → PRP → AST → IAST− 0.120[− 0.250, − 0.032]2.080.04Total indirect effectPR → (API, MPI, CRP, PRP, AST) → IAST0.167[0.053, 0.279]2.990.000 Table 6Indirect effects of perceived risk on intention to adopt safety technologies.Indirect pathwayStructural chainIndirect effect (β)95% CIt-valueP valueCentral-route mediationPR → API → CRP → AST → IAST0.151[0.074, 0.236]2.090.04Peripheral-route (ability-based) mediationPR → API → PRP → AST → IAST0.136[0.056, 0.258]3.050.000Peripheral-route suppression (motivation-based)PR → MPI → PRP → AST → IAST− 0.120[− 0.250, − 0.032]2.080.04Total indirect effectPR → (API, MPI, CRP, PRP, AST) → IAST0.167[0.053, 0.279]2.990.000 Indirect effects of perceived risk on intention to adopt safety technologies. Table 7Structural model’s predictive power.ConstructsQ2 predictRMSEMAECramér-von Mises testPLS lossIA lossAverage loss differencet valuep valueStatisticp valueAPI0.2430.8860.7203.2600.0010.8070.965− 0.1593.5990.001CRP0.6060.6370.4981.6370.0010.5981.030− 0.4336.4950.001MPI0.7120.5450.3964.1710.0010.6981.493− 0.7956.8390.001PRP0.2590.8740.6732.2610.0010.5720.697− 0.1251.5560.122AST0.6380.6110.4358.2520.0010.4590.851− 0.3926.0130.001IAST0.7720.4880.3714.8650.0010.3280.922− 0.5947.1210.001Overall0.5771.006− 0.42910.9920.001 Table 7Structural model’s predictive power.ConstructsQ2 predictRMSEMAECramér-von Mises testPLS lossIA lossAverage loss differencet valuep valueStatisticp valueAPI0.2430.8860.7203.2600.0010.8070.965− 0.1593.5990.001CRP0.6060.6370.4981.6370.0010.5981.030− 0.4336.4950.001MPI0.7120.5450.3964.1710.0010.6981.493− 0.7956.8390.001PRP0.2590.8740.6732.2610.0010.5720.697− 0.1251.5560.122AST0.6380.6110.4358.2520.0010.4590.851− 0.3926.0130.001IAST0.7720.4880.3714.8650.0010.3280.922− 0.5947.1210.001Overall0.5771.006− 0.42910.9920.001 Structural model’s predictive power. The results indicate that attitude toward safety technologies shows the strongest standardized path coefficient to intention to adopt safety technologies (β = 0.884, p < 0.001). Perceived risk is significantly associated with motivation to process information (β = 0.847, p < 0.001) and perceived ability to process information (β = 0.500, p < 0.001). Perceived ability exerts significant effects on both central route processing and peripheral route processing. Central route processing has a significant positive effect on attitude toward safety technologies, and peripheral route processing similarly shows a significant positive effect on attitude. Motivation to process information shows a significant negative effect on peripheral route processing with a coefficient of minus 0.361 at the 0.01 significance level. The results confirm that perceived risk (PR) influences intention to adopt safety technologies exclusively through indirect pathways (Fig. 2 and Table 6). Three mediation chains are notable: First, central-route pathway: PR enhances perceived ability (API), which promotes central-route processing (CRP) and generates favorable attitudes (β = 0.151, p < 0.05); Second, peripheral-route pathway via ability: PR increases API, which fosters reliance on peripheral cues (PRP), contributing positively to attitudes and adoption intention (β = 0.136, p < 0.001); and third, motivation-driven suppression pathway: PR increases motivation (MPI), which reduces reliance on peripheral cues, producing a negative indirect effect (β = − 0.120) by redirecting cognitive effort toward systematic evaluation. The total indirect effect of perceived risk on intention is positive (β ≈ 0.167) and statistically significant, indicating that facilitative pathways outweigh inhibitory mechanisms. PLS predict and loss-comparison analyses (Table 7) show that the PLS-SEM model outperforms a linear benchmark for most endogenous constructs (API, CRP, MPI, AST, and IAST), with lower RMSE/MAE and statistically significant loss differences (p < 0.001). Positive Q2predict values for API (Q2 predict ≈ 0.243; p = 0.001), MPI (Q2 predict ≈ 0.712; p = 0.001), CRP (Q2 predict ≈ 0.606; p = 0.001), AST (Q2 predict ≈ 0. 638; p = 0.001), and IAST (Q2 predict ≈ 0. 772; p = 0.001) indicate predictive relevance of the model for these constructs. In contrast, PRP exhibited comparatively weaker predictive performance (Q2 predict ≈ 0.259; p = 0.122), suggesting that additional antecedent variables may be necessary to improve prediction of peripheral information processing. Discussion Sample characteristics and implications for safety technology adoption The sample characteristics provide important contextual insights for interpreting safety technology adoption within greenhouse production systems. Respondents were predominantly middle-aged greenhouse owners and managers with several years of operational experience, suggesting familiarity with routine production practices, pesticide application, labor management, and day-to-day occupational safety challenges. Such experience may contribute to the development of practical knowledge regarding production risks while simultaneously shaping perceptions of the costs, benefits, and feasibility of adopting safety-related innovations. Consequently, technology adoption decisions in this context are likely to reflect not only perceptions of risk but also accumulated operational experience and prior exposure to technological change. The educational profile of the sample reveals substantial heterogeneity in formal educational attainment. From the perspective of the Elaboration Likelihood Model (ELM), differences in educational background may influence how individuals engage with safety-related information. Although education was not directly modeled in the present study, previous persuasion and technology-adoption research suggests that individuals with greater educational attainment often possess stronger capacities for evaluating technical information, whereas less formally educated audiences may rely more heavily on simplified communication, trusted information sources, and practical demonstrations. This observation highlights the importance of designing extension and risk communication programs that accommodate multiple forms of information processing rather than assuming a uniform decision-making style among greenhouse operators. The diversity of crop production systems represented in the sample further underscores the heterogeneous nature of occupational and environmental risk exposure within greenhouse agriculture. Different crops involve distinct pest-management requirements, agrochemical application frequencies, labor demands, and market-quality standards. For example, producers cultivating high-value crops may face stronger economic incentives to adopt technologies that improve residue management, product quality assurance, and regulatory compliance. Consequently, safety technology adoption should not be viewed as a uniform process across greenhouse systems but rather as a context-dependent decision shaped by both production characteristics and market conditions. Substantial variation in greenhouse size similarly suggests differences in resource availability and organizational capacity. Larger operations may possess greater financial resources, managerial specialization, and access to technical expertise, potentially facilitating the evaluation and implementation of new technologies. In contrast, smaller operations may face tighter budget constraints and greater sensitivity to investment costs, increasing the importance of affordability and perceived economic returns. These structural differences imply that a single policy approach is unlikely to be equally effective across all greenhouse enterprises. Instead, adoption-support programs may benefit from tailoring financial incentives, technical assistance, and training activities to the specific circumstances of different producer groups. Perceived risk, attitudes, and intention to adopt safety technologies The descriptive findings reveal an important and theoretically meaningful pattern: greenhouse owners reported only moderate levels of perceived agrochemical risk while simultaneously expressing favorable attitudes toward safety technologies and relatively strong intentions to adopt them. This pattern suggests that the relationship between risk perception and safety-related behavioral intention is more complex than a simple linear association. Rather than functioning as a direct determinant of adoption, perceived risk appears to operate through intervening cognitive and attitudinal processes. This observation provides an initial indication of the relevance of the extended Elaboration Likelihood Model (ELM) framework employed in this study and foreshadows the mediation mechanisms subsequently confirmed by the structural model. A first explanation concerns the heterogeneous nature of risk perception among greenhouse owners. Aggregate means may obscure substantial variation in how producers perceive occupational and environmental hazards associated with agrochemical use. Individuals who perceive higher levels of risk are likely to experience greater personal relevance and involvement, two factors that ELM identifies as important antecedents of elaborative information processing. Under such conditions, risk perception may stimulate motivation to seek information, evaluate technical evidence, and carefully assess the effectiveness of available safety technologies. Consequently, higher-risk individuals are more likely to form attitudes through central-route processing, generating stronger and more persistent evaluations of technological solutions. Conversely, producers who perceive lower levels of risk may arrive at favorable attitudes through alternative mechanisms that are less dependent on extensive cognitive elaboration. A second explanation relates to the coexistence of multiple persuasion pathways in agricultural decision-making. The positive evaluations of safety technologies observed in this study indicate that favorable attitudes can emerge through both central and peripheral routes of information processing. Some greenhouse owners may evaluate safety technologies on the basis of objective evidence regarding their effectiveness in reducing pesticide exposure, improving environmental protection, or enhancing compliance with safety regulations. Others may rely more heavily on heuristic cues, including recommendations from extension personnel, experiences of neighboring producers, supplier credibility, or broader perceptions that technological modernization represents a desirable production strategy. This finding is particularly important because it highlights that occupational and environmental safety decisions are not exclusively driven by technical risk assessments. Social influence, trust, and institutional credibility may also play substantial roles in shaping technology evaluations. Consequently, extension and risk communication programs should combine evidence-based information with trusted social and institutional channels to maximize their persuasive effectiveness. A third explanation concerns contextual factors that may attenuate perceived risk while simultaneously supporting positive evaluations of safety technologies. Many greenhouse owners increasingly operate in technologically intensive production environments where automation and precision-management systems are becoming more common. Within such contexts, safety technologies may be perceived not only as protective mechanisms but also as instruments for improving operational efficiency, product quality, regulatory compliance, and market competitiveness. As a result, adoption intentions may be motivated by a broader set of expected benefits than risk reduction alone. Furthermore, psychological mechanisms such as optimism bias, habituation to routine agrochemical exposure, and limited awareness of chronic health consequences may contribute to relatively modest reported risk perceptions despite objectively significant hazards. Similar patterns have been observed in occupational health research, where prolonged exposure to familiar hazards often reduces perceived vulnerability without necessarily reducing support for preventive interventions. Drivers of intention to adopt safety technologies Attitude as the proximal determinant of safety technology adoption The present study identifies attitude toward safety technologies as the strongest predictor of adoption intention among greenhouse managers. This finding aligns with major behavioral theories, including the Theory of Planned Behavior (TPB), the Technology Acceptance Model (TAM), and the Unified Theory of Acceptance and Use of Technology (UTAUT), all of which recognize attitudinal evaluations as central determinants of behavioral intention. However, the current findings extend these perspectives by demonstrating how attitudes emerge through multiple cognitive pathways rather than functioning as an isolated psychological construct. In greenhouse production systems, adoption decisions often involve substantial financial investments, operational adjustments, and uncertainty regarding technological performance. Consequently, managers are unlikely to adopt safety technologies solely because they recognize occupational hazards. Instead, they first develop evaluative judgments regarding the usefulness, effectiveness, feasibility, and overall desirability of the technology. The exceptionally strong relationship between attitude and adoption intention observed in this study suggests that greenhouse managers’ willingness to invest in safety innovations is fundamentally shaped by how positively they evaluate the technology itself. Perceived risk as a catalyst for cognitive engagement A key contribution of this study is the identification of perceived risk as an upstream driver of information processing rather than a direct determinant of adoption intention. Perceived risk significantly increased both motivation to process information and perceived ability to evaluate information, while exerting no direct effect on adoption intention. This pattern suggests that risk perception operates primarily by activating cognitive engagement mechanisms. This finding is consistent with risk perception theory and protection-oriented decision-making frameworks, which argue that awareness of potential hazards stimulates information seeking and deliberative evaluation processes. Within greenhouse agriculture, occupational hazards represent salient threats that may motivate managers to seek information regarding protective technologies. However, recognizing a risk does not automatically translate into behavioral intention. Instead, risk perception appears to trigger a cognitive process through which individuals actively evaluate available solutions before forming adoption intentions. The indirect nature of this relationship provides important theoretical insight. Previous agricultural technology adoption studies frequently conceptualize risk as a direct predictor of behavior. The current findings suggest that this assumption may oversimplify the decision-making process for safety-oriented innovations. In the context of occupational and environmental protection, risk awareness appears to function primarily as a motivational mechanism that initiates information processing and attitude formation. Consequently, risk communication campaigns may be most effective when they not only increase awareness of hazards but also provide sufficient information to facilitate informed evaluation of potential solutions. The coexistence of central and peripheral processing routes The findings provide strong support for the central proposition of the Elaboration Likelihood Model, namely that attitudes can be formed through both systematic and heuristic information processing pathways. Both central-route processing and peripheral-route processing exerted significant positive effects on attitudes toward safety technologies, indicating that greenhouse managers rely on multiple forms of cognitive evaluation when considering adoption decisions. The positive influence of central-route processing is theoretically expected. Safety technologies often involve technical specifications, operational requirements, and long-term consequences that encourage careful assessment of objective information. Managers who engage in systematic evaluation are more likely to examine evidence regarding effectiveness, reliability, compatibility with existing production systems, and expected safety outcomes. Because attitudes formed through central processing are generally more stable and resistant to change, this pathway may contribute to sustained adoption behaviors over time. Equally important, however, is the significant contribution of peripheral-route processing. Even in a context involving occupational health and environmental protection, managers appear to rely on heuristic cues such as recommendations from trusted peers, endorsements from extension services, supplier reputation, and perceived ease of implementation. This finding challenges the implicit assumption that high-stakes safety decisions are driven exclusively by rational and evidence-based evaluations. Instead, it suggests that social and contextual cues remain influential components of agricultural decision making. The comparable influence of both routes on attitude formation indicates that persuasion strategies targeting greenhouse managers should integrate both informational and social dimensions. Technical evidence alone may be insufficient to maximize adoption. Communication strategies that combine scientific information with credible endorsements, peer demonstrations, and social proof may be more effective in promoting favorable attitudes toward safety technologies. Reconsidering the role of perceived ability in technology adoption One of the most theoretically interesting findings concerns the dual influence of perceived ability on both central and peripheral processing routes. Consistent with traditional ELM assumptions, managers who perceived themselves as more capable of understanding and evaluating information were more likely to engage in systematic information processing. Greater confidence in one’s ability to assess technological information naturally facilitates deeper cognitive elaboration and more comprehensive evaluation of available evidence. More unexpectedly, perceived ability also positively influenced peripheral-route processing. This finding suggests that cognitively capable individuals do not necessarily abandon heuristic decision-making strategies. Rather, they may selectively use heuristic cues as efficient mechanisms for reducing information-processing demands in complex decision environments. Greenhouse managers frequently operate under significant constraints related to time, labor availability, production schedules, and market uncertainty. Under such circumstances, even highly capable decision makers may rely on trusted information sources and credible recommendations to simplify complex evaluations. Therefore, heuristic processing should not necessarily be interpreted as a sign of cognitive limitation. Instead, it may represent an adaptive strategy that enables efficient decision making under practical constraints. This finding contributes to ongoing discussions regarding the flexibility of cognitive processing mechanisms within ELM. Rather than functioning as mutually exclusive alternatives, central and peripheral processing may coexist and complement one another in real-world agricultural decision-making contexts. Future ELM applications in agriculture may benefit from conceptualizing cognitive processing as a continuum of complementary strategies rather than as strictly separate pathways. Conclusion and policy implications This study applied an extended Elaboration Likelihood Model (ELM) to examine how greenhouse owners form intentions to adopt occupational safety technologies aimed at reducing agrochemical exposure and improving environmental and worker safety. The findings demonstrate that adoption intention is primarily determined by attitude, which functions as the central mediating mechanism linking cognitive processing routes and behavioral intention. This result reinforces established technology acceptance theories while extending them by explicating the cognitive mechanisms underlying attitude formation in safety–critical agricultural contexts. The empirical evidence shows that perceived risk does not directly influence intention to adopt safety technologies. Instead, it operates indirectly by increasing motivation to process information and perceived ability to evaluate technical content. These cognitive states subsequently activate both central and peripheral processing routes, which jointly shape attitude formation. In this sense, perceived risk functions as a cognitive activation mechanism rather than a behavioral determinant, highlighting its indirect but structurally fundamental role in safety technology adoption. Both central and peripheral processing routes significantly influence attitudes toward safety technologies. Central route processing reflects systematic evaluation of scientific evidence and functional effectiveness, while peripheral route processing reflects heuristic reliance on social influence, institutional endorsement, and perceived usability. The coexistence of both pathways indicates that greenhouse owners do not rely exclusively on analytical reasoning even in high-risk occupational environments. Instead, cognitive processing is hybrid in nature, combining systematic and heuristic evaluations depending on informational and contextual constraints. Theoretical contributions This study makes three primary theoretical contributions to the literature on technology adoption and persuasion. First, the integration of the Extended Elaboration Likelihood Model with dominant adoption frameworks such as the Theory of Planned Behavior, Technology Acceptance Model, Unified Theory of Acceptance and Use of Technology, and Protection Motivation Theory advances understanding from static explanatory models toward a process-based account of how adoption intentions are cognitively constructed. The findings demonstrate that adoption behavior emerges through sequential cognitive mechanisms rather than direct causal relationships between exogenous variables and intention. Second, the study refines the role of perceived risk in ELM-based models of occupational safety behavior. Rather than functioning as a direct predictor of behavioral intention, perceived risk operates upstream by enhancing both motivation and perceived ability, thereby shaping the extent and direction of cognitive elaboration. This positions risk perception as a structural trigger of information processing, particularly relevant in domains characterized by health hazards and environmental exposure risks. Third, the findings challenge the classical assumption that cognitive ability primarily reduces reliance on peripheral cues. The positive relationship between perceived ability and peripheral processing suggests that cognitively capable individuals may strategically employ heuristic cues as efficient decision-making tools. This extends ELM theory by introducing a more context-sensitive interpretation of heuristic use as a complement to, rather than a substitute for, systematic processing. Practical contributions The results offer several implications for policymakers, extension systems, and agricultural technology providers. Effective promotion of safety technologies requires a dual-route intervention strategy. On the one hand, central-route strategies should emphasize scientific evidence, technical validation, and performance-based demonstrations of safety technologies. These approaches are particularly effective for operators with higher motivation and cognitive capacity. On the other hand, peripheral-route strategies should leverage trusted social actors, institutional endorsements, certification signals, and observable demonstrations of usability to influence decision-making among operators who rely more heavily on heuristic cues. Risk communication strategies should avoid purely fear-based messaging and instead focus on transforming risk awareness into informed cognitive engagement. This requires coupling risk information with capacity-building interventions such as training programs, advisory support, and simplified decision tools. Without such integration, risk perception may not translate into effective adoption behavior. Institutional and financial mechanisms also play a critical role in facilitating adoption. Subsidy programs, leasing models, pay-per-use systems, and certification-linked incentives can reduce financial barriers while simultaneously enhancing the perceived credibility of safety technologies. Embedding safety technologies within existing automated greenhouse systems further reduces adoption costs by aligning safety improvements with productivity-oriented investments. Limitations and future research This study has several limitations that should be considered when interpreting the findings. First, the sample is drawn primarily from greenhouse owners operating automated or semi-automated production systems, which implies that the respondents represent relatively technologically advanced and commercially oriented producers. This introduces a potential early-adopter bias and limits the generalizability of the findings to smallholder farmers or traditional greenhouse systems, where access to technology, financial resources, and information may differ substantially. Second, the cross-sectional research design restricts the ability to infer causal relationships or observe temporal changes in cognitive processing and adoption behavior. Third, all constructs were measured using self-reported survey data, which may introduce common method variance and social desirability bias. Finally, although PLS-SEM provides robust estimation for complex mediation models, future studies using longitudinal or experimental designs would be valuable for confirming causal pathways and strengthening external validity. Future research should extend this work by examining actual adoption behavior using longitudinal or panel data rather than behavioral intention alone. Experimental or quasi-experimental designs manipulating message framing, risk salience, or institutional signals would also strengthen causal interpretation. Additionally, future models should incorporate social network structures, peer influence dynamics, and institutional pressures to better explain peripheral processing mechanisms. Comparative studies across different technological maturity levels and regulatory environments would further enhance the generalizability of the extended ELM framework.