Promoting the Adoption of Agricultural Green Production Technologies for Sustainable Farming: A Multi-Attribute Decision Analysis

Abstract

Stakeholders have become worried about the environmental problems of agricultural production activities. Therefore, there is pressure on smallholder farmers to observe environmental regulations and embed sustainable green technologies in their production. The literature on green production has thoroughly emphasized the critical role of behavioral factors in adopting environmental sustainability practices. We develop a probabilistic linguistic preference selection index method to assess the behavioral constructs that promote the adoption of agricultural green production technologies (AGPTs) among smallholder rice farmers in an emerging economy. The result shows that the five most-important factors promoting the adoption of AGPTs include knowledge (0.828), perceived cost and benefit (0.819), descriptive norm (0.810), moral and environmental concern (0.809), and injunctive norm (0.807). The study findings offer insightful directions for examining rice farmers’ decisions on the adoption of AGPTs. Our findings imply that policymakers should consider multiple behavioral factors when designing policies that promote AGPTs. This study enriches farmers’ adoption decisions by modeling the uncertainties in the decision-making process.

1. Introduction

The increasing population growth among developing countries and the imminent threat of food insecurity have instigated sustained research attention on agricultural production activities [1,2,3,4]. Due to the centrality of the consumption of agricultural products in the livelihoods of many economies, local crop production has seen a significant upsurge in recent decades. Agricultural production activities have become a threat to the environment as they are seen as determinants of climate change. The growing increase in the usage of agricultural inputs such as fertilizer application, diesel-powered heavy-duty implements, and others has potentially adverse impacts on the environment [5,6]. As a result, stakeholders are concerned about environmental issues of agricultural production. To this end, there is pressure on smallholder farmers to adopt agricultural green production technologies (AGPTs) for sustainable production. Environmentally friendly technologies are at the forefront of efforts to create an economically sustainable future in light of the move from linear to circular economies [7]. Agricultural green production helps achieve sustainable production in the following ways: (i) increasing agricultural productivity to support food security and broader development goals, (ii) increasing adaptive capacity and resilience to climate variability, and (iii) decreasing greenhouse gas emissions.

Agricultural green production consists of processes and product innovation via enhancements in production processes utilizing environmentally friendly technologies. AGPTs aid in promoting ecofriendly products and help to protect the environment from harmful emissions [8]. AGPTs increase agricultural productivity without impacting the environment negatively [9]. Hence, it has been accepted as a vital strategic intervention for agricultural sustainability and rural food security. Agricultural green production shifts the attention from low-productive farming methods towards high-productivity sustainable techniques. Notwithstanding their enormous benefits, previous studies have revealed low rates of AGPT adoption in developing countries [10,11], causing soil degradation and nutrient depletion as the main limitations to productivity and sustainability in crop production.

The growing design and implementation of AGPTs across developing countries have aroused research interests over the last decades [12]. Nevertheless, the existing literature is overly skewed to analyze the determining factors of AGPTs by crop growers [13,14,15,16]. Meanwhile, the studies on green innovation adoption have highlighted the critical role of attitudinal factors in environmental conservative practices’ adoption and implementation [17,18,19]. Surprisingly, the role of behavioral factors in AGPT adoption has been marginalized in the current literature. Additionally, the prioritization of the AGPT adoption factors remains unexplored in the extant literature. Research on AGPTs adoption is important to serve as a policy guide locally and as a baseline analysis for a complete grasp of the factors of agricultural-related GHG emissions in developing countries. In light of this, we identify a comprehensive list of behavioral factors and examine their importance in promoting the adoption of AGPTs in rice production.

Rice production expansion has been recognized as a sure policy approach to improving food and nutrition security because the crop is fast becoming a primary staple food around the globe. For instance, rice serves as a direct daily energy and nutrition source in sub-Saharan Africa [20]. Due to rice consumption’s role in the livelihoods of many developing economies, local crop production has seen a significant upsurge in recent decades. Many developing countries like Ghana aim to achieve self-sufficiency through supporting commercial or large-scale rice cultivation. As a staple food, rice is grown in all of Ghana’s agroecological zones. Between 2008 and 2019, paddy rice production increased by around 10% yearly, with 2019 seeing tremendous growth of 25%. [21] There have been several government policy initiatives in recent years to limit rice imports while increasing the ability of local farmers to meet expanding demand. However, rice farming in commercial quantities has some environmental concerns [1,22]. Typically, rice production is associated with some harmful GHG emissions, including carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O) [1,22]. Therefore, there is a need to encourage the adoption of AGPTs (e.g., green fertilizer, soil testing formula for fertilization, biological pesticides, insecticidal lamps, etc.). Through the government of Ghana’s (GOG) policy dubbed “planting for food and jobs”, many young individuals have been trained to adopt green farming practices. To motivate these young farmers, the government provides financial incentives such as loans and insurance policies to reduce the risks of adopting AGPTs. Because behavioral tendencies are important in adopting innovative technologies, the question that comes to mind is, “what behavioral factors influence rice farmers’ adoption decisions on AGPTs considering the uncertainties of the decision-making process?”.

Identifying behavioral factors in promoting AGPT adoption can be regarded as a multi-attribute decision analysis (MADA) problem. MADA is the procedure of assessing various alternatives based on multiple conflicting attributes [23,24,25]. One MADA technique scholars focus on is the preference selection index (PSI) [26]. The PSI method is specifically practical when there is conflict in deciding the relative importance between attributes, which is the advantage of the PSI method. The PSI method’s strength lies in the fact that it does not require the computation of the attributes’ weights before its utilization. Instead, determining the attributes’ importance is part of its integral mechanism [26]. Furthermore, the PSI method offers a simple, logical, and systematic way to unravel complex decision-making issues [27]. The PSI technique has been employed to solve diverse prioritization problems [27,28,29,30,31]. To effectively implement policies that promote AGPT adoption, scientific data on farmers’ adoption determinants and their behavioral aspects of the decision-making process become the crucial take-off point. Decision-makers usually rely on their habits, conjectures, instincts, and other irrational psychological traits, rather than complete rationality [32]. As a result, farmers’ knowledge and thought are mostly limited when it comes to the adoption of AGPTs. In light of this, we employ the concept of a probabilistic linguistic term set (PLTS) [33] to model the farmers’ linguistic evaluations. Due to the merits of the PLTS, it has often been used in various disciplines to solve group decision-making [32,34,35,36,37,38]. Due to the capabilities of the PSI method and the PLTS, we integrate the two techniques and develop a probabilistic linguistic preference selection index (PL-PSI) method to identify the factors encouraging AGPT adoption among rice farmers. This novel method provides a way to model the uncertainty and vagueness in farmer adoption decision-making.

In a nutshell, this paper aims to identify and prioritize the behavioral factors that drive the adoption of AGPTs in rice production using the PL-PSI MADA method. The findings of this study will indicate to policymakers and implementers the attitudinal factors that improve or impede AGPT adoption by farmers. In summary, this research contributes immensely to literature in the following ways:

(1)

This study enriches the literature on farmers’ adoption decisions by analyzing a comprehensive framework of behavioral factors in promoting agricultural green production.

(2)

Considering the uncertainties and ambiguity in the decision-making process of adopting AGPTs, this study employs PLTSs to model farmers’ knowledge and thoughts.

(3)

This study develops a novel MADA approach by extending the PSI method into the PLTSs environment to help policymakers understand farmers’ adoption decisions on AGPTs.

(4)

Based on the proposed PL-PSI method, we analyze the relative importance of the behavioral factors promoting the adoption of AGPTs among rice growers in Ghana.

The paper’s remainder is presented as follows: Section 2 discusses this paper’s theoretical background and literature review. Section 3 develops the proposed decision-making methodology. In Section 4, we present the results, discussion on the behavioral factors, and the implications of the findings. Section 5 offers the conclusion, limitations, and future studies.

2. Theoretical Background and Literature Review

2.1. Farmer Decision-Making and Behavioral Factors

Several studies have surveyed the factors that stimulate or hinder farmers’ adoption of sustainable agricultural [12,39,40,41,42,43]. These works have predominantly focused on how farmer demographics and socio-economic factors, including firm size and access to finance, influence the adoption of eco-friendly practices. For instance, Agula et al. [12] found that the intensity of adopting sustainable practices is significantly influenced by farmers’ age, distance to farms, perception of soil fertility, number of extension visits, and the type of irrigation scheme available to farmers. Furthermore, Mishra et al. [40] revealed that farmers who cultivated row crops, had access to irrigation facilities, and agreed with crop diversification were significantly more likely than their respective counterparts to adopt more sustainable agricultural practices. Tu et al. [42] modeled the factors driving the adoption of eco-friendly rice production in the Vietnamese Melong Delta. The authors concluded that membership in agricultural cooperatives or clubs, farmer experience, and the perceived difference in selling price positively affected adoption. At the same time, risk awareness and the number of paddy plots negatively influenced the adoption.

Current studies recognizing the significance of behavioral factors have started considering how these behavioral factors inspire the adoption of eco-friendly practices. Previous research has investigated the specifics of individual adoptions using a variety of adoption theories. The behavioral study on the adoption and spread of innovations, which concentrates on the technical elements of adoption and diffusion, differs fundamentally from Rogers’ theory [44]. According to this perspective, the most popular theoretical approaches are the theory of reasoned action (TRA) and the extended TRA, which is connected to the theory of planned behavior (TPB) [44,45]. The TPB, developed by Ajzen [46,47], considers the impact of a subjective norm, attitudes, and perceived behavioral control adopting an innovation’s intentional and observable characteristics. Although TPB [48,49,50,51,52] is widely used, there are many problems with applying it and elucidating behavioral drivers that have led researchers to develop different theories or to use alternative models to comprehend how people perceive and react to changes in agricultural and food systems. Whatever the underlying theory, current research provides mounting evidence that farmers’ perspectives are crucial to the acceptance and spread of sustainable agricultural technology, such as green farming, and their satisfaction with system utilization [53]. However, the vast diversity of research on farmers’ views also suggests that it is still difficult to fully comprehend the idea of innovation adoption. Since many farmers have been labeled as small adopters, this may be much more complicated in developing and emerging nations [54].

As a consequence of the literature mentioned above, we conclude that there is still a considerable research gap regarding the role of farmers’ attitudes and intentions in adopting green agriculture, particularly in non-developed countries. Therefore, the present study focused on investigating farmers’ attitudes and intentions, particularly to determine factors contributing to the adoption and continuance of the implementation of green farming.

Recently, Dessart et al. [55] reviewed how behavioral factors affect farmers’ decision-making. The authors provided a comprehensive framework of behavioral factors for farmers’ adoption decisions. The behavioral factors are conceptually organized based on their “distance” from the decision-making in question. Behavioral factors are considered distal when they are higher-order, general “macro” principles, relatively remote from specific decision-making situations. At the other end of the spectrum, factors are proximal when they consist of lower-order, “micro” variables directly or almost directly related to the focus of the decision-making. On this distal–proximal spectrum, Dessart et al. [55] distinguished three broad dimensions of behavioral factors: dispositional, social, and cognitive. Dispositional factors predict farmers’ tendency to behave in a specific manner [56]. These most distal factors are comparatively steady, core variables linked to personality, motives, ideals, principles, general desires, and goals. Several studies have found a significant effect of dispositional factors on farmers’ approval of eco-friendly practices [57,58,59,60]. Social factors define farmers’ relations with other persons, such as fellow farmers or supervisors, and social norms and signaling motivations. Social norms broadly denote the joint representation of tolerable behavior and individual approval of certain comportment by others [61]. Dessart et al. [55] argued that interpersonal relations impact farmers’ decisions to accept eco-friendly practices. Social factors can be proximal or distal. For example, injunctive norms, i.e., what farmers anticipate people think of them, might influence farmers to implement a specific practice or more eco-friendly practices overall. Many authors have indicated the significant relationship between social factors and farmers’ adoption of sustainable farming practices [62,63]. Cognitive factors are considered proximal and often dwell on learning and thinking. These factors relate to farmers’ insights on the relative costs, benefits, and risks linked with a specific eco-friendly practice. For instance, while a farmer may consider it a costly or risky approach to implement, another farmer may see it as less expensive or dangerous. Previous studies have found a significant impact from cognitive factors on farmers’ adoption of eco-friendly farming practices [63,64,65].

This study fills the literature gap by adopting the comprehensive framework of Dessart et al. [55] (

Table 1. Literature on behavioral factors and eco-friendly farming practices adoption.

Constructs	Factors	References
Dispositional	Personality	[66,67]
	Resistance to change	[57]
	Risk tolerance	[68,69]
	Moral and environmental concerns	[58,59,60]
	Farming objective	[68,70]
Social	Descriptive norm	[41,62]
	Injunctive norm	[63,68]
	Signaling motives	[62,63]
Cognitive	Knowledge	[65]
	Perceived control	[63,71]
	Perceived costs and benefits	[58,64,72]
	Perceived risk	[65,72]

) to examine the behavioral factors promoting the adoption and implementation of AGPTs in rice production. In doing this, the study contributes to the existing literature by extending research on the behavioral drivers for AGPT adoption in developing and emerging countries.

2.2. Related Works on AGPTs

Environmentally friendly developments, such as AGPTs, are being promoted in countries all over the globe. The application of AGPTs targets several agricultural activities with environmentally friendly innovation to help promote sustainable agriculture and boost output [73]. Nevertheless, popular wisdom holds that output may increase without regard to unsustainable activities’ negative repercussions. While studies have determined the traditional approach to be harmful to the soil and crops, it may also pose a health risk to farmers [74]. Consequently, these techniques negatively impact the farm’s output and the farmers. Sustainable agricultural growth is the best path forward in this dilemma, and AGPTs are the possible answers [53]. Despite this, the challenge is to raise the adoption rate in developing nations like Ghana, where it has been low [75]. Studies on adopting green technology in industries such as manufacturing and energy have been published multiple times during the last few years. The use of green technology in the agriculture sector is, however, relatively scarce.

Table 2. Prior studies on AGPT adoption.

Author	Country	Study Objective	Method/ Approach	Study Results
[76]	Malaysia	To measure the intention level of farmers to adopt GFT based on the psychological aspects based on the theory of planned behavior (TPB)	Structural Equation Modeling	The results have shown that direct and indirect attitude and indirect subjective norm, as well as direct–indirect perceived behavioral control, have a positive and significant influence on the intention of adopting.
[77]	China	To examine the logical relationship of farmers’ willingness to adopt green fertilization technology	Structural Equation Modeling	The results showed that adoption motivation, adoption opportunity, technical operation ability, and anti-risk ability had significant positive direct effects on adoption willingness, which were 0.610, 0.381, 0.491, and 0.297, respectively. Trust had an indirect effect, which was 0.191.
[43]	China	To perform social network analysis to identify critical stakeholders and barriers in agriculture green technology diffusion	Network Analysis	The results show that agricultural research institutes, universities, agribusiness, agencies of township promotion, the government, and farmers’ relatives are key stakeholders and that the limited market demand for green technology and the high cost of its diffusion are two main barriers.
[78]	China	To evaluate the contribution of time preferences to farmers’ technology adoption behavior	Econometric Modeling	The results show that time preferences significantly reduce technology adoption; in particular, farmers who are more present-biased have a lower proportion of technology adoption.
[74]	China	To study the factors influencing tea farmers’ adoption of PGCT for tea plant pest control	Structural Equation Modeling	The results indicated that subjective norms and perceived behavioral control have a positive and significant effect on intention, while attitude has no significant effect on the intention
[79]	China	To analyze the key factors affecting the development of green agriculture	Econometric Modeling	The results show that a farmer’s age, land type, compensation for land transfer, technical service organization, related training, and economic and technological subsidies had significant effects on their green agricultural production willingness.
[73]	China	To analyze the formation mechanism of the peer effects	Econometric Modeling	The authors found peer effects in the adoption behavior of green control techniques among farmers, which only existed in core and intermediate farmers and not in marginal farmers

shows a summary of previous research on the adoption of AGPTs.

A lot of the prior research on the adoption of AGPTs focused on a wide variety of topics, such as the social motives of farmers as behavioral triggers [76]. Extant studies have used a variety of approaches, including structural equation modeling, econometrics, and network analysis, to study the adoption of AGPTs. Because AGPT adoption is a lengthy and interconnected process, it is impacted by several competing attributes. Farmers require a synchronized review of the AGPTs and the motivating factors for their acceptance to implement AGPTs successfully. Additionally, farmers confront difficulties obtaining the knowledge they need to make informed decisions. Consequently, farmers often depend on inaccurate and fuzzy information while making judgments. Consequently, prior research methodologies are ineffective in this regard. As a result, in this study, we model farmers’ choices on using AGPTs as a MADA issue.

MADA entails concurrently considering several conflicting factors, constraints, and objectives, necessitating compromises and trade-offs. Analyzing the determinants of AGPT adoption is an important aspect of sustainable decision-making. It is a multi-dimensional challenge with several “boxes checked” simultaneously. MADA evaluates factors to see if each is a good or unfavorable option for a certain application, as opposed to qualitative and survey approaches, which generally test correlations, explain, and study cause and effect linkages. It also attempts to evaluate this variable, depending on the specified variables, against every other possible choice to aid the decision maker in picking an alternative with the least amount of sacrifice and the most benefits [80]. In addition, fuzzy MADA aids in modeling the uncertainties in farmers’ evaluation information. Deviating from prior studies, we employ the fuzzy MADA in this study to examine the behavioral factors motivating farmers’ adoption of AGPTs.

2.3. MADA and Green Technology Adoption

Multi-attribute decision analysis (MADA) methods often solve real-world problems with many alternatives and conflicting evaluation attributes [23,24,25]. A broad application of MADA approaches for evaluating green technologies has provided an ideal answer for complex decision-making challenges. Green technology adoption decision issues may be solved using a variety of MADA approaches, both conventional and fuzzy. A review of prior studies on green technology development using MADA approaches is shown in

Table 3. Some studies on the application of MADA in green technology adoption.

Author	Country	Application Area	Study Objective	Method/ Approach
[81]	China	Emerging industries	To evaluate green technology innovation on ecological economic efficiency of strategic emerging industries	Entropy-weighted TOPSIS
[82]	China	Emerging industries	To assess the barriers of green technology adoption for enterprises	Fuzzy AHP
[83]	Pakistan	Energy	To assess the barriers to the implementation of cleaner energy technologies	Modify Delphi and fuzzy AHP
[84]	United Kingdom (UK)	Building construction	To select retrofit for non-domestic building by using green technology	AHP
[25]	Ghana	Energy	To identify and rank the barriers to renewable energy development	MULTIMOORA-EDAS
[85]	Turkey	Energy	To evaluate solar projects	AHP and fuzzy VIKOR
[86]	Egypt	Transportation	To select an electric bus for green transportation	AHP and TOPSIS
[7]	Pakistan	General	To develop an integrated green technology framework to fill a gap in the literature by prioritizing green technologies’ most critical attributes	Fuzzy AHP
[87]	Ghana	Human-resource management	To investigate critical barriers to green human resource management (GHRM) implementation	PSI
[8]	Pakistan	General	To develop an integrated strategic framework based on strengths, weaknesses, opportunities, and threats (SWOT) for effective green technology planning	Gray AHP and gray TOPSIS
[88]	Ghana	Energy	To assess renewable energy barriers and prioritize their adoption and development strategies	CRITIC and fuzzy TOPSIS

.

Table 3 shows that there are numerous applications of MADA regarding the adoption of green technologies. A careful observation shows a limited use of MADA for analyzing green technologies for agricultural production. For instance, Singh and Mallick [89] used the fuzzy TOPSIS method to examine sustainable green chamber farming practices. Despite the inadequate employment of MADA in examining AGPT adoption, prior studies failed to consider the importance of the linguistic expressions provided by the decision-makers during the evaluation process. As a result, it is prudent that we develop a novel decision support model that can overcome the identified gaps in the extant literature. Among the MADA methods, PSI has recently received attention from scholars [27,28,29,90]. The PSI method can model a conflict in determining the relative importance between attributes [28,90]. As a result of its advantage over the other MADA methods, it has been applied to solve numerous problems in diverse disciplines. For example, Kumar et al. [90] applied the PSI method in investigating the performance of mechanical properties and sliding wear. Again, Sari [28] used PSI to appraise the recovery alternatives of electrical and electronic wastes. Ulutas et al. [27] designed a fuzzy integrated model based on PSI for selecting a green supplier.

To solve the problem of linguistic importance and the fuzziness of farmers’ evaluation of AGPTs, we employ the concept of PLTSs. PLTSs can model the vagueness and uncertainty in farmers’ adoption decisions while considering the importance of the linguistic evaluations provided by the farmers. Since the introduction of PLTSs, many scholars have extended its application to resolve many decision-making issues [32,34,35,36,37,38]. For example, Agbodah and Darko [32] developed some probabilistic linguistic aggregation operators based on Einstein’s t-Norm and t-Conorm and applied them in multi-criteria group decision-making. [34] created interactive multi-attribute decision-making using PLTS. Furthermore, Li et al. [35] employed PLTS for medical scheme selection regarding the referral system. Liang et al. [38] utilized PLTS to analyze web celebrity shops and developed improvement pathways using sentiment analysis and a fuzzy cognitive map. Lu et al. [37] proposed the TOPSIS method under probabilistic linguistic MAGDM through entropy weight and applied it to select a supplier for agricultural machinery products. Wen and Liang [36] relied on PLTS to capture decision-makers’ attitudinal characteristics in large-group decision-making. The literature shows that the application of PLTS in agricultural decision-making is limited. Hence, this study fills the literature gap by extending the PSI method into a probabilistic linguistic fuzzy environment and developing an improved MADA approach to prioritize the behavioral factors that stimulate farmers’ AGPT adoption.

2.4. Probabilistic Linguistic Term Set

Pang et al. [33] developed the probabilistic linguistic term set (PLTS) as an improvement of the hesitant fuzzy linguistic term set (HFLTS). The PLTS can depict the importance of linguistic terms based on the probabilities associated with them. Pang et al. [33] defined the PLTS as follows:

Definition 1.

Let$\mathrm{H}=\left\{{\mathrm{H}}_{\vartheta }|\vartheta =0,1,\cdots ,\eta \right\}$be a linguistic term set. Hence, a probabilistic term set (PLTS) is defined as follows:

$$L(p)=\left\{L^{(k)}(p^{(k)})|L^{(k)}\in \mathrm{H},r^{(k)}\in \eta ,p^{(k)}\ge 0,k=1,2,\cdots ,\#L(p),{\displaystyle \sum _{k=1}^{\#L(p)}{p}^{(k)}\le 1}\right\}$$

(1)

where $L^{(k)}(p^{(k)})$denotes the linguistic term $L^{(k)}$corresponding to the probability $p^{(k)}$, $r^{(k)}$denotes the subscript of $L^{(k)}$, and $\#L(p)$represents the number of all linguistic terms in $L(p)$.

To compare and rank two PLTSs, Pang et al. [33] proposed the following score and deviation degree functions:

Definition 2.

Let $L(p)=\left\{L^{(k)}(p^{(k)})|k=1,2,\cdots ,\#L(p)\right\}$ be a PLTS, and $r^{(k)}$ is the subscript of the linguistic term $L^{(k)}$. Hence, the score $E(L(p))$and deviation degree $\sigma (L(p))$of $L(p)$is given as follows:

$$E(L(p))=s_{\overline{\alpha }},\overline{\alpha }=\frac{{\displaystyle \sum _{k=1}^{\#L(p)}{r}^{(k)}{p}^{(k)}}}{{\displaystyle \sum _{k=1}^{\#L(p)}{p}^{(k)}}},$$

(2)

$$\sigma (L(p))=\frac{{\left({\displaystyle \sum _{k=1}^{\#L(p)}{(p^{(k)}(r^{(k)}-\overline{\alpha }))}^2}\right)}^{\frac{1}{2}}}{{\displaystyle \sum _{k=1}^{\#L(p)}{p}^{(k)}}}$$

(3)

According to (2) and (3), the ranking method for two PLTSs can be given as:

(i)

If $E(L_1(p))>E(L_2(p))$, then $L_1(p)$ is more significant than $L_2(p)$, denoted by $L_1(p)>L_2(p)$;

(ii)

If $E(L_1(p))=E(L_2(p))$, then

(a)

if $\sigma (L_1(p))<\sigma (L_1(p))$, then $L_1(p)>L_2(p)$;

(b)

if $\sigma (L_1(p))=\sigma (L_1(p))$, then $L_1(p)$ is indifferent to $L_2(p)$, indicated as $L_1(p)~L_2(p)$.

3. Methodology

This study explores the behavioral factors that inspire agricultural green production technologies (AGPTs) adoption among rice farmers in Ghana. We argue that examining these behavioral factors concerning the identified AGPTs can be regarded as a MADA problem. Therefore, we design a MADA technique grounded on the preference selection index (PSI) under a probabilistic linguistic (PL) environment and name it the PL-PSI method.

3.1. Problem Description

Let $G=\{g_1,g_2,\dots ,g_n\}$ be a set of AGPTs adopted by the rice farmers and $F=\left\{f_1,f_2,\dots ,f_m\right\}$ be a collection of behavioral factors that influence the adoption of AGPTs. Based on the linguistic scale: $s_1= \mathrm{Not} \mathrm{all} \mathrm{important}$, $s_2= \mathrm{Slightly} \mathrm{important}$, $s_3= \mathrm{Moderately} \mathrm{important}$, $s_4= \mathrm{Very} \mathrm{important}$, $s_5= \mathrm{Extremely} \mathrm{important}$, the rice farmers provide their evaluation to the extent by which the factor $f_i$ promotes the adoption of the AGPTs $g_j$. Then, we analyze each linguistic term’s frequency and build a comprehensive PLTS $L_{ij}(p_{ij})$ using the PLTS concept and statistical analysis. Finally, the probabilistic linguistic decision matrix $D={\left(L_{ij}(p)\right)}_{m\times n}$ in

Table 4. Probabilistic linguistic decision matrix $D={\left(L_{ij}(p)\right)}_{m\times n}$.

	$$g_1$$	$$g_2$$	$$\cdots$$	$$g_n$$
$$f_1$$	$$L_{11}(p_{11})$$	$$L_{12}(p_{12})$$	$$\cdots$$	$$L_{1n}(p_{1n})$$
$$f_2$$	$$L_{21}(p_{21})$$	$$L_{22}(p_{22})$$	$$\cdots$$	$$L_{2n}(p_{2n})$$
$$⋮$$	$$⋮$$	$$⋮$$	$$\cdots$$	$$⋮$$
$$f_m$$	$$L_{m1}(p_{m1})$$	$$L_{m2}(p_{m2})$$	$$\cdots$$	$$L_{mn}(p_{mn})$$

can be created according to the PLEs obtained for each factor $f_i$ concerning the AGPTs $g_j$.

The $L_{ij}(p_{ij})$ represents the evaluation score of the factor $f_i$ concerning the AGPTs $g_j$$(i=1,2,\cdots ,m;j=1,2,\cdots n)$.

3.2. Probabilistic Linguistic Preference Selection Index

The algorithm of the proposed methodology is as follows:

• Step 1: Build the probabilistic linguistic decision matrix $D={\left(L_{ij}(p)\right)}_{m\times n}$ as presented in Table 4.
• Step 2: Compute the mean of the evaluation values of every AGPT using the probabilistic linguistic average (PLA) operator [33] as follows:

$$L_j(p_j)=\frac{1}{m}\left({\displaystyle \underset{L_{1j}^{(k)}\in L_{1j}(p),L_{2j}^{(k)}\in L_{2j}(p),\cdots ,L_{mj}^{(k)}\in L_{mj}(p)}{\cup }\left\{p_{1j}^{(k)}{L}_{1j}^{(k)}\oplus p_{2j}^{(k)}{L}_{2j}^{(k)}\oplus \cdots \oplus p_{mj}^{(k)}{L}_{mj}^{(k)}\right\}}\right)$$

(4)

• Step 3: Calculate the preference variation value regarding each AGPT as follows:

$$\begin{array}{ll}{\mathsf{\Upsilon }}_j& ={\displaystyle \sum _{i=1}^m\left(d\left(L_{ij}(p_{ij}),L_j(p_j)\right)\right)}\\ & ={\displaystyle \sum _{i=1}^m\sqrt{\frac{{\displaystyle \sum _{k=1}^{\#L_{ij}(p_{ij})}{\left(p_{ij}^{(k)}{r}_{ij}^{(k)}-p_j^{(k)}{r}_j^{(k)}\right)}^2}}{\#L_{ij}(p_{ij})}}}\end{array}$$

(5)

where $r_{ij}^{(k)}$ and $r_j^{(k)}$ are the subscripts of the linguistic term $L_{ij}$ and $L_j$, respectively.

• Step 4: Obtain the deviation in the preference variation value for every AGPT as follows:

$${\Delta }_j=\left[1-\frac{{\mathsf{\Upsilon }}_j}{m-1}\right]$$

(6)

• Step 5: Determine the total preference value for each AGPT as follows:

$${\varpi }_j=\frac{{\Delta }_j}{{\displaystyle \sum _{j=1}^n{\Delta }_j}}$$

(7)

• Step 6: Ascertain the preference selection index ${\Pi }_i$ of each factor $f_i$ based on the probabilistic linguistic weighted average (PLWA) operator [33] as follows:

$${\Pi }_i={\displaystyle \underset{L_{i1}^{(k)}\in L_{i1}(\tilde{p}),L_{i2}^{(k)}\in L_{i2}(p),\cdots ,L_{in}^{(k)}\in L_{in}(p)}{\cup }\left\{\left({\varpi }_1{p}_{i1}^{(k)}{L}_{i1}^{(k)}\right)\oplus \left({\varpi }_2{p}_{i2}^{(k)}{L}_{i2}^{(k)}\right)\oplus \cdots \oplus \left({\varpi }_n{p}_{in}^{(k)}{L}_{in}^{(k)}\right)\right\}}$$

(8)

• Step 7: Calculate the score $s({\Pi }_i)$ according to Definition 2.
• Step 8: Rank the factors in descending order of $s({\Pi }_i)$. The larger $s({\Pi }_i)$, the higher the extent to which the factor $f_i$ influences the adoption of the AGPTs.

4. Decision Analysis

4.1. Research Approach

In light of the first step of the PL-PSI MADA method, we identify the behavioral factors that influence farmers’ adoption decisions. Based on the framework of Dessart et al. [55], we present the behavioral factors to four experts from academia and industry to examine the suitability and appropriateness of the factors in promoting AGPT adoption in Ghana. Through a brainstorming online discussion among the experts, nine behavioral factors $\left(f_1-f_9\right)$ were considered suitable for our case study. Following Dessart et al. [55] the nine factors were grouped into three dimensions: dispositional, social, and cognitive as depicted in Figure 1.

According to the 2020 annual report of Ghana’s Ministry of Food and Agriculture (MoFA), we identified four AGPTs adopted by rice farmers. These include green fertilizer ($g_1$), soil testing formula for fertilization ($g_2$), biological pesticides ($g_3$), and insecticidal lamps ($g_4$). Based on the AGPTs and the behavioral factors in Figure 1, we designed a questionnaire. A five-point Likert scale was adopted to evaluate the extent to which these nine behavioral factors were essential in adopting AGPTs. The questionnaire comprised information on household demographics and factors that influence AGPT adoption. To ensure the reliability of the questionnaire, we performed a pilot test. We selected fifteen farmers for the pilot test, and the feedback obtained was used to improve on the questionnaire and its reliability.

The sample for this study comprised 200 rice farmers from three ecological regions in Ghana. We employed a multi-stage sampling method for selecting the study participants. During the first stage, three regions, namely Northern, Eastern, and Volta, were nominated. These regions were chosen due to their high rice production during the 2020 rice season. With rice farmers’ association leaders’ aid, we picked 70 rice growers from each of the three regions. As a result, 210 questionnaires were administered. However, 10 of the returned questionnaires were inaccurate and not usable for further analysis and were thus withdrawn. The data was edited and coded into an Excel spreadsheet for the PL-PSI method.

Before collecting data, this study obtained ethical clearance from the Sichuan Agricultural University (SICAU) College of Management. The study ensured the confidentiality, privacy, and anonymity of the participants. Both written and verbal consent were obtained from each participant before data collection. The study’s results are current and reliable. Any text collected from peer-reviewed academic papers was incorporated with suitable references to prevent copyright concerns. The authors were properly acknowledged via in-text citations and a comprehensive reference list.

4.2. Results

Of the 200 rice farmers sampled for the study, 67% were males, while the remaining were females. A total of 34% of the rice growers were involved in an off-farm job. The average year of formal education for the rice farmer was 11.90%, and the mean age of rice farmers was 42.12 years. The average farming experience of rice farmers was 12.74 years. The decision results based on the PL-PSI model are outlined below:

• Step 1: Based on the questionnaire responses, we employed the PLTS described in Section 2.3 and constructed the initial probabilistic linguistic decision matrix of

  Table 5. The probabilistic linguistic decision matrix.

  Factors	${\mathit{g}}_1$	${\mathit{g}}_2$	${\mathit{g}}_3$	${\mathit{g}}_4$
  $f_1$	$\left\{\begin{array}{l}{s}_1(0.110),s_2(0.110),\\ s_3(0.140),s_4(0.110),\\ s_5(0.540)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.115),\\ s_3(0.160),s_4(0.100),\\ s_5(0.600)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.030),s_2(0.120),\\ s_3(0.160),s_4(0.110),\\ s_5(0.580)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.190),s_2(0.050),\\ s_3(0.150),s_4(0.110),\\ s_5(0.520)\end{array}\right\}$
  $f_2$	$\left\{\begin{array}{l}{s}_1(0.020),s_2(0.115),\\ s_3(0.160),s_4(0.120),\\ s_5(0.590)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.060),s_2(0.115),\\ s_3(0.170),s_4(0.100),\\ s_5(0.560)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.080),s_2(0.120),\\ s_3(0.130),s_4(0.090),\\ s_5(0.600)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.040),s_2(0.125),\\ s_3(0.115),s_4(0.285),\\ s_5(0.435)\end{array}\right\}$
  $f_3$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.120),\\ s_3(0.155),s_4(0.235),\\ s_5(0.465)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.110),\\ s_3(0.155),s_4(0.235),\\ s_5(0.475)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.095),s_2(0.125),\\ s_3(0.160),s_4(0.230),\\ s_5(0.390)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.125),s_2(0.030),\\ s_3(0.280),s_4(0.115),\\ s_5(0.450)\end{array}\right\}$
  $f_4$	$\left\{\begin{array}{l}{s}_1(0.020),s_2(0.120),\\ s_3(0.155),s_4(0.115),\\ s_5(0.590)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.035),s_2(0.110),\\ s_3(0.160),s_4(0.095),\\ s_5(0.600)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.120),\\ s_3(0.155),s_4(0.105),\\ s_5(0.595)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.040),s_2(0.165),\\ s_3(0.220),s_4(0.145),\\ s_5(0.430)\end{array}\right\}$
  $f_5$	$\left\{\begin{array}{l}{s}_1(0.040),s_2(0.120),\\ s_3(0.160),s_4(0.120),\\ s_5(0.570)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.020),s_2(0.100),\\ s_3(0.170),s_4(0.100),\\ s_5(0.630)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.060),s_2(0.115),\\ s_3(0.150),s_4(0.090),\\ s_5(0.590)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.150),s_2(0.055),\\ s_3(0.110),s_4(0.250),\\ s_5(0.435)\end{array}\right\}$
  $f_6$	$\left\{\begin{array}{l}{s}_1(0.030),s_2(0.115),\\ s_3(0.155),s_4(0.120),\\ s_5(0.580)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.100),s_2(0.120),\\ s_3(0.170),s_4(0.100),\\ s_5(0.560)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.080),s_2(0.110),\\ s_3(0.120),s_4(0.100),\\ s_5(0.610)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.060),s_2(0.140),\\ s_3(0.110),s_4(0.265),\\ s_5(0.425)\end{array}\right\}$
  $f_7$	$\left\{\begin{array}{l}{s}_1(0.300),s_2(0.120),\\ s_3(0.160),s_4(0.120),\\ s_5(0.590)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.040),s_2(0.110),\\ s_3(0.150),s_4(0.100),\\ s_5(0.600)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.110),s_2(0.125),\\ s_3(0.155),s_4(0.105),\\ s_5(0.505)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.040),s_2(0.130),\\ s_3(0.120),s_4(0.240),\\ s_5(0.480)\end{array}\right\}$
  $f_8$	$\left\{\begin{array}{l}{s}_1(0.020),s_2(0.090),\\ s_3(0.150),s_4(0.120),\\ s_5(0.640)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.030),s_2(0.110),\\ s_3(0.120),s_4(0.100),\\ s_5(0.650)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.120),\\ s_3(0.155),s_4(0.100),\\ s_5(0.610)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.145),s_2(0.055),\\ s_3(0.105),s_4(0.275),\\ s_5(0.420)\end{array}\right\}$
  $f_9$	$\left\{\begin{array}{l}{s}_1(0.025),s_2(0.115),\\ s_3(0.155),s_4(0.115),\\ s_5(0.590)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.075),s_2(0.115),\\ s_3(0.155),s_4(0.105),\\ s_5(0.550)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.055),s_2(0.125),\\ s_3(0.160),s_4(0.110),\\ s_5(0.550)\end{array}\right\}$	$\left\{\begin{array}{l}{s}_1(0.130),s_2(0.040),\\ s_3(0.235),s_4(0.155),\\ s_5(0.440)\end{array}\right\}$
  Mean	$\left\{\begin{array}{l}{s}_{0.066},s_{0.228},s_{0.463},\\ s_{0.522},s_{2.864}\end{array}\right\}$	$\left\{\begin{array}{l}{s}_{0.046},s_{0.223},s_{0.470},\\ s_{0.460},s_{2.903}\end{array}\right\}$	$\left\{\begin{array}{l}{s}_{0.062},s_{0.240},s_{0.448},\\ s_{0.462},s_{2.794}\end{array}\right\}$	$\left\{\begin{array}{l}{s}_{0.102},s_{0.176},s_{0.482},\\ s_{0.818},s_{2.242}\end{array}\right\}$

  .
• Step 2: Following the second step, we computed the mean of each AGPT evaluation value using the decision matrix of Table 3 and Equation (4). The result is given as follows:

$$L_1(p_1)=\left\{\begin{array}{l}{s}_{0.066},s_{0.228},s_{0.463},\\ s_{0.522},s_{2.864}\end{array}\right\}, L_2(p_2)=\left\{\begin{array}{l}{s}_{0.046},s_{0.223},s_{0.470},\\ s_{0.460},s_{2.903}\end{array}\right\}, L_3(p_3)=\left\{\begin{array}{l}{s}_{0.062},s_{0.240},s_{0.448},\\ s_{0.462},s_{2.794}\end{array}\right\} \mathrm{and} L_4(p_4)=\left\{\begin{array}{l}{s}_{0.102},s_{0.176},s_{0.482},\\ s_{0.818},s_{2.242}\end{array}\right\}$$

• Step 3: Per Equation (5), the preference variation value regarding each AGPT was calculated, and the result is given as follows.

$${\mathsf{\Upsilon }}_1=0.734, {\mathsf{\Upsilon }}_2=0.818, {\mathsf{\Upsilon }}_3=1.295 and {\mathsf{\Upsilon }}_4=1.297$$

• Step 4: Furthermore, we obtained the deviation in the preference variation value for every AGPT based on Equation (6), and we present the outcome in column 3 of Table 4.

$${\Delta }_1=0.908, {\Delta }_2=0.898, {\Delta }_3=0.838 and {\mathsf{\Upsilon }}_4=0.838$$

• Step 5: The total preference value for each AGPT was determined according to Equation (7), and we show it in column 4 of Table 4.

$${\varpi }_1=0.261, {\varpi }_2=0.258, {\mathsf{\Upsilon }}_3=0.241 and {\mathsf{\Upsilon }}_4=0.241$$

• Step 6: Then, we computed the preference selection index ${\Pi }_i(i=1,2,3,4,5,6,7,8,9)$. based on Equation (8), and the outcome is given in

  Table 6. Preference selection index, overall scores, and ranking of the behavioral factors.

  Factors	PSI $\left({\Pi }_{\mathit{i}}\right)$
  $f_1$	$\left\{\begin{array}{l}{s}_{0.088},s_{0.199},s_{0.457},\\ s_{0.430},s_{2.801}\end{array}\right\}$
  $f_2$	$\left\{\begin{array}{l}{s}_{0.050},s_{0.237},s_{0.434},\\ s_{0.589},s_{2.737}\end{array}\right\}$
  $f_3$	$\left\{\begin{array}{l}{s}_{0.066},s_{0.194},s_{0.559},\\ s_{0.820},s_{2.230}\end{array}\right\}$
  $f_4$	$\left\{\begin{array}{l}{s}_{0.030},s_{0.256},s_{0.516},\\ s_{0.459},s_{2.776}\end{array}\right\}$
  $f_5$	$\left\{\begin{array}{l}{s}_{0.066},s_{0.196},s_{0.444},\\ s_{0.556},s_{2.789}\end{array}\right\}$
  $f_6$	$\left\{\begin{array}{l}{s}_{0.067},s_{0.242},s_{0.419},\\ s_{0.580},s_{2.724}\end{array}\right\}$
  $f_7$	$\left\{\begin{array}{l}{s}_{0.125},s_{0.242},s_{0.440},\\ s_{0.560},s_{2.728}\end{array}\right\}$
  $f_8$	$\left\{\begin{array}{l}{s}_{0.054},s_{0.188},s_{0.398},\\ s_{0.589},s_{2.912}\end{array}\right\}$
  $f_9$	$\left\{\begin{array}{l}{s}_{0.070},s_{0.199},s_{0.526},\\ s_{0.483},s_{2.670}\end{array}\right\}$

  .

• Step 7: Finally, we deduced the preference selection index scores since they were probabilistic linguistic elements using Definition 2. The nine behavioral factors’ overall scores are listed in column 5 of

  Table 7. Ranking of the behavioral dimensions and individual factors.

  Dimensions	Average Score	Dimension Ranking	Factors	$\mathbf{Overall} \mathbf{Scores} \mathit{s}\left({\Pi }_{\mathit{i}}\right)$	Global Ranking
  Dispositional context	$S_{0.793}$	$3rd$	$f_1$	$S_{0.795}$	$7th$
  			$f_2$	$S_{0.809}$	$4th$
  			$f_3$	$S_{0.774}$	$9th$
  Social context	$S_{0.809}$	$2nd$	$f_4$	$S_{0.807}$	$5th$
  			$f_5$	$S_{0.810}$	$3rd$
  Cognitive context	$S_{0.811}$	$1st$	$f_6$	$S_{0.806}$	$6th$
  			$f_7$	$S_{0.819}$	$2nd$
  			$f_8$	$S_{0.828}$	$1st$
  			$f_9$	$S_{0.790}$	$8th$

  . To ascertain the dimensions’ scores, we averaged the individual factors’ scores within a dimension, and we present the outcome in column 2 of Table 7.
• Step 8: The ranking positions of the behavioral factors are presented as follows:

$$f_8\succ f_7\succ f_5\succ f_2\succ f_4\succ f_6\succ f_1\succ f_9\succ f_3.$$

Thus, knowledge ($f_8$) is the most-important factor in promoting the adoption of AGPTs among rice farmers in Ghana.

4.3. Discussion

Table 7 shows the global ranking of each factor influencing AGPT adoption among rice farmers in Ghana. Out of the nine factors, the top-five ranked factors are knowledge, perceived cost or benefits, descriptive norm, moral or environmental factors, and injunctive norms. On the other hand, the least-important factors are farmers’ objective, perceived risk, personality traits, and perceived control.

Farmers’ knowledge is ranked first among the factors that promote AGPT adoption among rice farmers in Ghana. Thus, the most significant precondition for rice farmers to adopt AGPTs is knowing both the existing AGPTs and the current government policies associated with AGPTs. While the lack of green farming awareness decreases farmers’ urgency in adopting sustainable practices, farmers’ knowledge of these practices increases their desire to adopt such practices [55,65]. An advisory service to promote farmers’ awareness and an understanding of green practices and governments’ policies must be established to increase farmers’ knowledge. Our findings’ next factor that influences AGPT adoption is the perceived benefits. Adopting green technologies entails both costs and benefits [91]. Moreover, the cost of using green farming practices may result from procuring specialized equipment or adopting new ways of farming. These green practices may save time, reduce labor costs, and increase productivity. Our results imply that farmers within the rice industry consider if the intended benefits outweigh the cost of adopting green farming practices. The perceived cost and benefit can be handled via government-subsidized policies. The government can also provide some green farming inputs at subsidized costs to reduce these farmers’ financial burdens. Some previous studies have made similar recommendations [55,92].

The descriptive norms, which explain how one neighbor’s behavior influences his decision to adopt AGPTs, are the third factor that promotes the adoption of AGPTs. Farmers with little knowledge of their neighbors’ environmental strategies are less likely to implement AGPTs [93]. To motivate the larger community of Ghanaian rice farmers to adopt AGPTs, others who have adopted these sustainable practices may communicate and encourage their fellow farmers on the significance of adopting such policies. A similar method has been proven effective in encouraging consumers in various aspects [94]. The fourth factor that influences AGPT adoption is moral and environmental factors. These factors relate to how personal conscience, ethics, concern for others’ welfare, and ecological issues affect the adoption of green farming practices. While farmers with high moral and environmental values are more likely to accept some conservative practices, those with low moral and ecological concerns are less likely to implement green farming practices [95]. Moral and environmental factors ranking fourth suggest that rice farmers consider the environmental effects of specific sustainable practices before adopting AGPTs. Thus, policyholders should organize programs to explain to farmers the importance of ecological conservation.

Next, we have found injunctive norms to be the fifth factor that stimulates AGPTs. These injunctive norms explain what farmers think others expect from them. If rice farmers in Ghana believe that society expects them to commit to sustainable projects, they will focus more on agri-environmental projects. This suggests that rice farmers’ awareness of the societal demands regarding environmental concerns encourages adopting ecological practices. Policies intended to educate farmers concerning agriculture in general and the value of eco-friendly farming practices may upsurge the injunctive norm towards more green farming practices. The sixth-ranked factor in the AGPT adoption among rice farmers is perceived control. This factor relates to Ghana’s rice farmers’ perceptions that they have the necessary skills and time to take action. Implementing new green farming practices may require farmers to learn new skills and to have enough time to implement such practices. The more farmers anticipate they can quickly actualize the practices associated with green farming practices, the more likely they will participate. Efforts should be made to educate these rice farmers on adopting green farming practices without prolonging the training time.

Even though differences in thinking, feeling, and behavior influence farmers’ agri-environmental usage decisions, personality trait is the seventh factor influencing AGPTs. With personality traits ranking seven out of the nine factors, Ghanaian rice farmers are not motivated to seek new sustainable experiences out of curiosity. We have found perceived risk and farming objectives to be the two least-important factors influencing rice farmers’ adoption of AGPTs among Ghanaian rice farmers. Perceived risk shows how the usage of agri-environmental practice may contribute to higher production or otherwise. If Ghanaian rice farmers anticipate that adopting green farming practices is risky and may not bring many returns, this factor should have been among the top. Perceived risk ranking eighth suggests that rice farmers do not see adopting green farming practices as risky ventures. Perhaps they consider higher productivity from green farming practices compared to the traditional methods. Finally, our results show that the least-important factor promoting AGPT adoption among rice farmers is the farming objective.

Based on the average scores of the behavioral factors, cognitive-dimension factors are the most-important to consider in adopting AGPTs among rice farmers in Ghana. These cognitive factors are proximal to the decision-making process, and efforts should be made to ensure their proper integration to enhance the adoption of green farming practices in rice farming. This study’s cognitive factors include perceived risk, perceived control, perceived cost and benefit, and knowledge. Our findings suggest a need to raise the awareness of Ghanaian rice farmers’ understanding of the existence, policies, and perceived benefits obtained from the use of these practices. The significance of the cognitive dimensions in agri-environmental practices has been recognized by existing studies [55,64]. Following the cognitive dimension is the social dimension. These social factors explain how societal norms, including others’ perceptions, influence rice farmers’ decisions to adopt green farming practices. Thus, to quicken the adoption of AGPTs, others who have used these practices may serve as ambassadors to explain to those yet to use them the perceived benefits they stand to get after adoption. Previous studies have reported the effectiveness of using similar practices in the agriculture sector [41,62]. Finally, our results show that dispositional factors are the least-important in influencing the adoption of AGPTs.

4.4. Implications

This study makes a theoretical contribution by employing a comprehensive framework on behavioral factors to analyze farmers’ adoption decisions. Our findings reveal that cognitive-dimension factors significantly influence rice farmers to adopt AGPTs. These factors include perceived risk, control, cost and benefit, and knowledge. Cognitive-dimension factors are proximal to rice farmers’ decision-making, so properly integrating these factors is essential to promote green rice production. Following the cognitive dimension are the social and dispositional dimensions, respectively. Subsequently, we consider the global ranking of the individual behavioral factors to provide specific policy interventions. The results show that the five most-important behavioral factors promoting AGPT adoption are farmers’ knowledge about the practices, perceived benefit and cost, distinctive norms, moral and environmental concerns, and injunctive norms. These results imply that farmers’ knowledge about the current green-based farm management practices and the existing government policies associated with these practices is crucial.

These results provide insightful background for policymakers in addressing farmers’ adoption decisions regarding green farming practices, thereby reducing rice production’s environmental impact. First, the findings have discovered that providing advisory services and making information available to rice farmers are central to their adoption decisions. Therefore, policymakers should provide avenues to raise farmers’ awareness of AGPTs. Second, perceived benefit and cost are considered the second-most crucial factor, indicating that policymakers should provide information and education on the environmental benefits of adopting AGPTs. Furthermore, policymakers can design some government-subsidized environmental schemes for rice farmers. These policies will reduce rice production’s environmental impact and ensure sustained food production and security in the long run. The differences in the factors’ scores are minimal, indicating that all of the behavioral factors are crucial in adopting AGPTs. Hence, policymakers must consider multiple behavioral factors when designing green rice production policies. Third, this study’s proposed MADA method provides a novel way for researchers to investigate farmers’ adoption decisions.

4.5. Comparative and Sensitivity Analysis

4.5.1. Comparative Study

To verify the effectiveness of our proposed model, we conducted a comparison with other models such as PL-TOPSIS, PL-EDAS, and PL-WASPAS. Based on the initial matrix of Table 3, we rank the alternatives using the PL-TOPSIS, PL-EDAS, and PL-WASPAS methods and compare their results with the outcome of our proposed model. The weights of the evaluation attributes are the same as our model to ensure the consistency of comparison. The final ranking results of the methods are presented in

Table 8. Ranking results of the various MADA models.

No.	Methods	Rank
1	PL-TOPSIS	$\begin{array}{l}{f}_8\succ f_7\succ f_2\succ f_4\succ f_5\succ \\ f_6\succ f_1\succ f_9\succ f_3\end{array}$
2	PL-EDAS	$\begin{array}{l}{f}_7\succ f_8\succ f_5\succ f_2\succ f_6\succ \\ f_4\succ f_1\succ f_9\approx f_3\end{array}$
3	PL-WASPAS	$\begin{array}{l}{f}_8\succ f_7\succ f_5\succ f_2\succ f_4\succ \\ f_6\succ f_1\succ f_9\succ f_3\end{array}$
4	Our proposed method	$\begin{array}{l}{f}_8\succ f_7\succ f_5\succ f_2\succ f_4\succ \\ f_6\succ f_1\succ f_9\succ f_3\end{array}$

.

According to Table 8, our proposed method can select the same optimal factor as the PL-TOPSIS and PL-WASPAS methods. Thus, knowledge ($f_8$) is the most-important factor in promoting the adoption of AGPTs among rice farmers in Ghana based on three of the four methods. However, all four methods identified farming objective ($f_3$) as the least-important factor. This indicates that our proposed method is stable and valid. However, the proposed PL-PSI has some merits over the other methods. Unlike the other methods, the proposed method can simultaneously determine the weights of the attributes and compute the overall scores of the alternatives. In the other methods such as PL-TOPSIS, PL-EDAS, and PL-WASPAS, decision-makers must compute the weights of the attributes separately using other methods such as AHP, CRITIC, and BWM. This makes the PL-PSI method simple, logical, and systematic in unraveling complex decision-making issues.

4.5.2. Sensitivity Analysis

The outcomes of this research were subjected to sensitivity analysis to verify their consistency and robustness. By altering the weights of the AGPTs in different circumstances, this study seeks to disclose new rankings for the behavioral factors. This section considers five new scenarios to compare to the current case. In the 1st scenario, all the AGPTs were treated equally and allocated the same weight. From the 2nd to the 5th scenario, each AGPT was given a higher weight, while the rest were allocated a low weight. For instance, in scenario 2, AGPT $g_1$ was assigned a higher weight, 0.5, while the rest $\left(g_2-g_4\right)$ shared the same weight. All of the other AGPTs were processed following this rule throughout the scenarios. Figure 2 shows the final sensitivity analysis scores for the behavioral factors $\left(f_1-f_9\right)$.

It can be observed from Figure 2 that, though there were some fluctuations in the rankings of the factors across several cases, most of them were consistent with the current case. For instance, three of the five new scenarios examined ranked knowledge ($f_8$) as the most-important behavioral factor, the same as what we obtained in the current work (current case in Table 6). Furthermore, all five scenarios ranked farming objective ($f_3$) as the least-important factor, which is consistent with what was obtained in our current case. We can infer that the weights of the AGPTs significantly impact the PSI method’s final results. Hence, it is appropriate that a suitable model is employed to determine the importance of the decision attributes. Unlike the other MADA methods, the PSI has an inherent mechanism to determine the weights of the decision attributes and rank the alternatives. Therefore, we conclude that our findings are consistent, robust, and valuable.

5. Conclusions

Because of the increasing use of agricultural inputs such as fertilizer application and diesel-powered heavy-duty equipment, AGPT adoption has become necessary for sustainable farming. AGPTs boost agricultural output while having no detrimental influence on the environment. This study aims to investigate the importance of behavioral factors in adopting AGPTs. Deviating from the extant studies, we construct the AGPT adoption decisions as a MADA problem. Initially, we identify the necessary behavioral factors from literature and analyze them using the framework developed by Dessart et al. [55]. According to this framework, nine behavioral factors are classified into three main dimensions, namely, dispositional, social, and cognitive. We develop a novel decision support model known as PL-PSI to rank these factors. The PL-PSI can model the fuzziness and incomplete information during the decision-making process of adopting AGPTs. The PL-PSI method is applied to rice growers in Ghana to investigate their behavioral tendencies in AGPT adoption. The results reveal that the five most-important factors promoting the adoption of AGPTs include knowledge, perceived cost and benefit, descriptive norm, moral and environmental concern, and injunctive norm. This study’s findings may help policymakers understand the factors influencing farmers adopting environmentally friendly farming techniques. The result implies that rice farmers’ adoption choices are heavily influenced by the availability of advisory services and information. As a result, governments should make it possible for farmers to learn more about AGPTs. We conduct a comparative and sensitivity analysis to ensure the validity and robustness of the proposed PL-PSI method. The outcome shows that our proposed method is stable, logical, consistent, and robust for analyzing complex decision problems.

This study comes with some limitations. The research focuses only on three regions in Ghana. Moreover, the study concentrates on rice production’s environmental impact. Furthermore, the study failed to consider the simultaneities among the adoption of the AGPTs. Future studies can consider a large sample size, preferably all the regions in Ghana, while exploring other agricultural production, including livestock and aquaculture. The simultaneities among the adoption of the AGPTs can be examined in future research to understand the interrelationship between the adoption of AGPTs during the decision-making process.