*3.2. Simulation Experiment Description*

In this study, the independence and interaction of individual pig farmers as an actor were simulated in a computer-generated environment when considering the influences of their knowledge and actor-to-actor information exchange on their VA use. The simulation was performed using Matlab based on Wu's and Zhou's research [22,24], and is described as follows:

(1) The simulation area is a 20 × 20 square area. At the start of the simulation, 100 farmers were randomly distributed in this area. Specific parameters are listed in Table 1 below.


**Table 1.** Settings of experimental parameters.


penalties), take a value in [1,5] in the simulation, respectively, where 1 means no knowledge and 5 means complete knowledge. Based on the behavior probability model, *ϕi*1, *ϕi*1, and *ϕi*<sup>3</sup> are the coefficient part of proper use, and *ϕi*4, *ϕi*5, and *ϕi*<sup>6</sup> are the coefficient part of improper use. As proper and improper VA uses are two opposite behaviors, when a farmer has a high willingness to perform one behavior, the willingness to perform the other behavior will be relatively low. Therefore, it is assumed that the relationship between the two sets of coefficients is as follows:

$$\begin{cases} \begin{aligned} \varphi\_{i1} + \varphi\_{i4} &= 5 \\ \varphi\_{i2} + \varphi\_{i5} &= 5 \\ \varphi\_{i3} + \varphi\_{i6} &= 5 \end{aligned} \tag{4} \end{aligned} \tag{5}$$

To ensure scientific rigor and practical relevance of the simulation, the 100 farmers were assumed to have a lower-middle level of knowledge at the start of the simulation. It was assumed that *ϕi*<sup>0</sup> = 2 and *ϕi*<sup>1</sup> = *ϕi*<sup>2</sup> = *ϕi*<sup>3</sup> = 3, that is, the farmers' three categories of knowledge fluctuated in the range of [1,3].

(4) Farmers' expected return. The farmers' expected return can be calculated by Equations (1) and (2). Farmers' normal return, *G*, follows uniform distribution in [5,9] (in ten thousand yuan). *θ* is the ratio of farmers' increased return from improper VA use to that from proper usage. In general, the higher the knowledge level regarding VA use specification, the lower the probability of an improper return. Therefore, *θ* is correlated with *ϕi*1. To ensure that *θ* is nonnegative, it was assumed that *θ* + *ϕi*<sup>1</sup> = 5. Based on the finding of field interview regarding spot checks for pig farmers that were conducted by government regulators each year, the initial value of *q* was set to 0.3. According to the Regulations on Administration of Veterinary Drugs in China, the penalty for improper VA use was set to 30,000 yuan considering the various forms of improper use. Hence, *C*<sup>1</sup> = 3. The higher the farming return, the higher the pressure from public opinion and moral pressure when the misconduct is disclosed and sanctioned. Hence, it is assumed that *C*<sup>2</sup> = 2 × *G*.

#### **4. Results and Discussion**
