key: cord-0944568-5vqj0ez5 authors: Wu, Jiang; Zuo, Renxian; He, Chaocheng; Xiong, Hang; Zhao, Kang; Hu, Zhongyi title: The effect of information literacy heterogeneity on epidemic spreading in information and epidemic coupled multiplex networks date: 2022-06-15 journal: Physica A DOI: 10.1016/j.physa.2022.127119 sha: 86a6f02f4bc468e4129ac4034b148ee05905852c doc_id: 944568 cord_uid: 5vqj0ez5 With the COVID-19 pandemic, better understanding of the co-evolution of information and epidemic diffusion networks is important for pandemic-related policies. Using the microscopic Markov chain method, this study proposed an aware–susceptible–infected model (ASI) to explore the effect of information literacy on the spreading process in such multiplex networks. We first introduced a parameter that adjusts the self-protection related execution ability of aware individuals in order to emphasis the importance of protective behaviors compared to awareness in decreasing the infection probability. The model also captures individuals’ heterogeneity in their information literacy. Simulation experiments found that the high information-literate individuals are more sensitive to information adoption. In addition, epidemic information can help to suppress the epidemic diffusion only when individuals’ abilities of transforming awareness into actual protective behaviors attain a threshold. In communities dominated by highly literate individuals, a larger information literacy gap can improve awareness acquisition and thus help to suppress the epidemic among the whole group. By contrast, in communities dominated by low information-literate individuals, a smaller information literacy gap can better prevent the epidemic diffusion. This study contributes to the literature by revealing the importance of individuals’ heterogeneity of information literacy on epidemic spreading in different communities and has implications for how to inform people when a new epidemic disease emerges. In 2020, COVID-19 became a global pandemic. Although the propagation of the virus has been studied and many governments are aware of the severity of the pandemic, the public's willingness or abilities to take self-protective behaviors vary greatly, leading to serious consequences in many countries. Several researchers argued that the spread of epidemic-related information, including knowledge of the virus (such as fatality rate and infection rate of the virus), preventive measures (such as wearing a face mask, social distancing) and current situation of the epidemic (such as regional risk level), can help to suppress the spread of epidemics from the perspective of information-epidemic coupled complex networks [1] [2] [3] [4] [5] . In reality, the information received comes from two sources: social media and aware neighbors. In addition, the information-adoption ability, or the awareness-acquisition ability that transforms information into self-awareness, varies from person to person. Most studies assume that all unaware individuals have an equal probability of acquiring awareness only from their aware neighbors, and an aware individual directly has a lower likelihood of being infected than an unaware individual [1, 6, 7] . However, there remain gaps between information and awareness as well as awareness and behavior, which are not equivalent. First, we argue that the assumptions are excessively arbitrary regardless of the social media information source. Next, awareness should be transformed into actual protective behavior (or the self-protection-related execution ability), which can reduce the probability of being infected. Tangcharoensathien et al. [8] pointed out that knowledge should be translated into actionable behavior-change messages for infodemic management. Information literacy of a person plays a vital role in transforming external knowledge, namely information on epidemic prevention, into awareness [9] [10] [11] [12] [13] . Then with a certain probability, the person is assumed able to take more positive action to prevent the epidemic. This study attempts to fill the research gap in exploring the effect of information literacy heterogeneity on suppressing the epidemic. Consequently, the core research question of this paper is: How would the information literacy heterogeneity of individuals affect information propagation and epidemic diffusion in an information-epidemic coupled multiplex networks considering the execution ability of awareness? In the model, information literacy heterogeneity is classified into high-and low information-literate individuals as simple parameters. It is assumed that the high information-literate individuals have higher efficiency (θ , 0.5 < θ ≤ 1) in adopting and transforming external epidemic-related information into self-awareness than the low information-literate individuals (1−θ ) , and the concept of information literacy gap (2θ −1) in which the gap of information adoption efficiency between the high-and low-information-literate individuals, is proposed to reflect the degree of the heterogeneity. To explore the effect of information literacy heterogeneity on suppressing the diffusion of the epidemic, this study proposes an aware-susceptible-infected model (ASI) by considering the self-protection-related execution ability (dominating by a parameter α) of the aware individuals. Finally, the present study focuses on understanding and controlling the epidemic diffusion from the perspective of the sub-divided high-and low-information-literate communities. It is found that the high information-literate individuals are more sensitive to social media information and the re-propagated information of neighbors. The self-protection-related execution ability does not help to acquire awareness, but helps to control epidemic diffusion. In communities dominated by high-information-literate individuals, a larger information literacy gap can improve the entire awareness acquisition and thus help to suppress the epidemic. In communities dominated by low information-literate individuals, a smaller information literacy gap benefits the epidemic prevention. In addition, only when the self-protection related execution efficiency attains a certain value can social media information or the re-propagated information of the neighbors help to suppress the diffusion of the epidemic. The remainder of this paper is organized as follows: In Section 2, this paper proposes the definition of information literacy and reviews the relevant studies on information-epidemic coupled multiplex networks. Section 3 summarizes the model structure and coupled methods. Section 4 illustrates the results of the simulation experiments with implications. Section 5 discusses the results from information sensitivity and three controllers. Finally, conclusions are presented in Section 6. Information literacy originally refers to the ability to recognize information needs and to identify, evaluate and use information effectively [14] . An information-literate individual is supposed to be able to recognize when information is needed and can locate, evaluate, and use effectively the information [15] . The concept of ''information literacy'' has been extended to many specific scenes, including computer literacy, library literacy, digital literacy, etc. In the internet era. information literacy is regarded as the essential framework that informs and unifies additional literacy types. While the medium and type of information may vary, the standard information literacy characteristics (determine, access, evaluate, incorporate, use, understand) are still common considerations [16] . Information literacy is closely related to human behavior. Earlier, information literacy has been mostly a practical and strategic concept guiding the library field's efforts in teaching information seeking and using skills. Information literacy evolves in the course of realizing specific work-related tasks and mundane activities [17] . Individuals with different information literacy will have different information behaviors. For example, the information literacy of young people has a huge impact on their information evaluation awareness, skills, and practices. Suboptimal online information literacy among youth today not only leads to potential shortcomings in young people's information consumption behaviors [18] but also their frequent sharing of misinformation and rumors on social media [19] . In addition, information literacy has also been proven to play a moderating role in information and communication technology adoption behavior [20] . Healthy behavior is one of the most important behaviors related to information literacy. Insufficient information literacy is regarded as an obstacle to the use of health information in online health information acquisition [21] . Obtaining health knowledge through social media and other channels can improve health information literacy and thus promote healthy behaviors [22] . In this paper, information literacy is defined as the ability of individuals to seek, find, understand and appraise epidemic information from online resources or their neighbors, and apply such knowledge into selfprotective behaviors. Whether an individual will benefit from the information from online resources and his neighbors largely depended on his information literacy. The spreading of information and that of the epidemic are not isolated but they interact with and affect each other [23] . And the information literacy plays a significant role in this interaction process. The epidemics can facilitate the spreading of information. Apart from the traditional way of disseminating information by word of mouth, individuals can post and obtain information from various social media platforms, which has exerted a profound impact on human society. Herein, information literacy could be reflected in individuals' posting and obtaining information from social media platforms. Normally, the outbreaks of epidemics can trigger and further be accompanied by the dissemination of epidemic-related information [24] . It is found that the epidemics facilitate the information spreading in the transient process, instead of the steady state [25] . Meanwhile, the information spreading can inhibit epidemics. On the one hand, the spreading of information holds back the spreading of the epidemic. Once gaining information about the epidemic, individuals often make some responses to reduce their infection possibility, such as washing hands with sanitizer, wearing a mask, or staying at home [26] . Herein, individuals' response to the epidemic could be influenced by information literacy. Individuals with high information literacy can effectively apply their awareness into self-protective behaviors. On the other hand, the infected individuals are likely to inform friends about the existence of the epidemic via social networks or word of mouth, thus, generating more new aware individuals, where information literacy could also play a significant role. Consequently, the spreading of epidemic-related information may exert a significant impact on the epidemic spreading in the population. It is found that the inhibition effect of information on epidemics can be reflected by both the transient process and steady state [25] . It is noted that information literacy can directly influence the process of information spreading and the interaction process of the spreading of information and epidemic, and thus exert a profound influence on the epidemic spreading. To comprehensively model the dynamic interactions between the spreading of information and epidemics, multi-layer networks have been widely adopted. The spreading of information and epidemic is represented by multiple network layers, while nodes represent the same entities in all layers. Granell originally applied the multi-layer network and further proposed the susceptible-infected-susceptible unaware-aware-unaware model to capture the interrelation between the process of epidemic spreading and information spreading [1] . Many researchers followed this line and extended many variants to address a wide range of specific problems. For instance, Guo modeled the information spreading layer as a time-varying network generated by the activity driven model while the contagion layer as a static network, and found that the spreading of information can not only enhance the epidemic threshold but also reduce the prevalence of epidemics [3] . Ye extended a behavior layer to the model and studied the influence of individual differences in risk perception and behavior change in people's responses to infectious disease outbreaks [26] . However, information literacy is seldom taken into consideration when modeling the dynamic interactions between the spreading of information and epidemics. This paper studies the heterogeneity of information literacy for the individuals to suppress the epidemic, and a basic Unaware-Aware-Unaware-susceptible-Infected-Susceptible model (UAU-SIS) which is first proposed by Granell et al. [1] is adopted. However, Granell et al. [1] did not consider the mass media information, which is actually the public's fundamental information source. Xia et al. [27] have improved the model by considering the mass media information, and the process of the mass media information is accepted. Nevertheless, they neglect the intermediate transition function of behavior between awareness formation and epidemic diffusion. Ye et al. [28] have proposed a three-layer model with a behavior change layer. However, their model is somewhat complicated, and a simple structure is possible. Moreover, little previous research has paid an eye on the heterogeneity of information literacy for the individuals. As a result, this study proposes an ILH-UAU-SIS model to explore the effect of information literacy for the whole individuals on the information propagation and epidemic diffusion in the coupled multiplex networks, which takes the self-protection related execution ability of the aware individuals into consideration as well. In the proposed ILH-UAU-SIS model, information literacy is reflected in information transformation into awareness. In this stage, the information-literate individuals are divided into high-information-literate individuals who have a higher transition probability (θ , 0.5 < θ ≤ 1) of transforming information into awareness and low information-literate individuals with a lower transition probability (1 − θ). The Microscopic Markov Chain Approach(MMCA) method [1] is then used to derive the variation of aware or infected individuals' proportions. Finally, by adjusting the proportion of information-literate individuals, some interesting results are found and concluded. There are four constraints in the proposed model. First, this model simplifies the process of information transforming into awareness. Some researchers point out that an information-literate individual should accept information with a critical spirit [12, 29, 30] . If so, the information should be further divided, such as useful, fake or false ones, to reflect the selection process. To simplify the process, the information is regarded as useful in this study. Second, the formation of awareness may be affected by the individual's degree in the multiplex networks and the number of its aware or infected neighbors [31] [32] [33] [34] , while it is not considered because it is not related with information literacy. Third, the information re-propagation is a kind of altruistic behavior for the aware by default, which is uncertain in reality. After all, Pan and Yan [35] found that the altruistic behaviors of infected nodes have a significant effect on suppressing the epidemic's spread. Fourth, the gap of information literacy is just a part of individual heterogeneity. Consequently, many factors that the individual heterogeneity reflected may not be considered, such as the heterogeneity of infection rates [36] , the influence heterogeneity in the epidemic layer [34] . Epidemic-related information comes from two sources: mass media and the aware individual re-propagated as shown in Fig. 1 . Mass media, such as TV, newspaper, Weibo, Facebook and so on, are information sources that the public can access. The aware individual also conveys information. The unaware must have aware neighbors, and then they adopt information from the aware. There are two kinds of information literacy for individuals: high information literacy and low information literacy. The high information-literate individuals have higher information adoption efficiency than the low ones. Normally, two parameters should be used to define these two types of information literacy. For example, 0 < θ 1 < θ 2 < 1. However, one more parameter can greatly increase the model's complexity, and just one parameter θ (0.5 < θ ≤ 1) for the high information-literate individuals is proposed to simplify the complexity of the model. Therefore, the parameter θ 1 for the low information-literate individuals is simplified as 1 − θ. There are two reasons for such parameter design. First, θ 1 and θ 2 sum equaling one is the most simple equation to reflect the heterogeneity of information literacy for individuals. Second, such a structure can demonstrate the information literacy gap for the high-and low information-literate communities through 2θ −1. Consequently, θ and 1−θ are designed to reflect the core point of this paper: the heterogeneity of information literacy. Besides acquiring from the external information, self-awareness has an additional source approach that is assumed to be acquired automatically if the individual is already infected. The heterogeneity of information literacy functions on the process efficiency of information transforming into awareness (I2 A). In specific, high information-literate individuals own a high transformation efficiency, while low information-literate individuals own a low one. In the process of I2 A, the awareness comes from two sources: being infected, external information. Except for being infected, acquiring awareness from external information is a complex process to consider both the information literacy and the information subject. There are two cases to acquire awareness from the social media (with a probability of λ 1 ) or the aware re-propagated information (with a probability of λ 2 ), where λ 1 (0 ≤ λ 1 ≤ 1) and λ 2 (0 ≤ λ 2 ≤ 1) are the probabilities of acquiring awareness from external information for a normal individual. It is assumed that the high information-literate individuals acquire awareness from the social media information with probability θλ 1 , and acquire awareness from the aware neighbors' re-propagated information with probability θλ 2 . The low information-literate individual acquires awareness with probability (1−θ) λ 1 and (1−θ) λ 2 , respectively. Self-protective behavior rather than self-awareness can change the probability of being infected, because there is a gap between awareness transforming into behavior. Actually, only protective behavior (such as wearing a face mask) α Probability for the aware to take protective actions; 0.8 by default. ratio Proportion of the high information-literate in the entire nodes; initial is set as 0.5. can be observed, but awareness cannot; besides, only protective behavior can decrease the probability of being infected, but awareness cannot. Each aware individual has a certain probability of taking protective actions, and a parameter α (0 < α < 1) is applied to reflect the transition efficiency. Then, they can reduce the probability of being infected with protective behavior. Because of the duration of the epidemic, people may be numb or used to the existence of the epidemic and therefore the awareness may fade from their memory. There is a certain probability δ (0 ≤ δ ≤ 1) to lose awareness and become unaware. The infection spreads from certain infected individuals to their unaware neighbors with a probability , and it is assumed that the infected one will automatically acquire awareness. The infected nodes will recover with probability µ (0 ≤ µ ≤ 1) in each time step. In return, if an individual is aware in the information propagation layer and is susceptible in the epidemic diffusion layer, it reduces its infectivity by a factor γ (0 ≤ γ ≤ 1) if the awareness has been transformed into protective behavior. Considering so many parameters in the model, a parameters summary table is given below as seen in Table 1 . The initial values of λ 1 , λ 2 , δ, µ, γ , ρ AI , ρ AS , and ρ US refer to Granell et al. [1] ,Xia et al. [27] . Taking both awareness status and infection status into consideration, an individual can be in several states: unaware and susceptible (US), aware and susceptible (AS), or aware and infected (AI). In the coupled multiplex networks, a ij and b ij denote to be the adjacent matrices in information propagation layer and epidemic diffusion layer respectively. At the beginning, all nodes will be divided proportionally into i h (the high information-literate nodes) and i l (the low information-literate nodes). Every node i has probabilities P AI i (t), P AS i (t), and P US i (t) to be in the status of AI, AS, and US respectively. Supposing not existing the dynamical correlations [37] , the transition probabilities for node i not acquiring awareness from social media information can be divided into according to the I2 A ability of the node. The transition probabilities for node i not acquiring awareness from aware neighbors' re-propagated information can also be classified into r i h (t) and r i l (t). (1) where p A j = p AI j + p AS j , j is the number of neighbors, and a ji is the adjacent matrix in the information propagation layer. The transition probabilities of not being infected by any neighbors for the aware individuals should consider the protective behavior, as seen in Fig. 2 . As a result, the transition probability of not being infected by any neighbors for the aware individuals is: where b ji is the adjacent matrix in the epidemic diffusion layer, and β A = αγ β U The transition probabilities for the nodes which will not be infected by any neighbors if remaining unaware q U i (t) is: Two kinds of probability trees are proposed to reveal the possible status of the high and the low information literacy individuals and their transition probabilities, as seen in Fig. 3 . There are three steps in the transition process. i. The awareness acquiring from the re-propagated information of the aware neighbors or loss. Such as AI(h) losses awareness and becomes UI with probability δ. The gap of information literacy reflects on the unaware individuals: US(h) acquires awareness and becomes AS with probability 1 − r i h (t) in Fig. 3(c) , and US(l) not acquires awareness and remains US with probability r i l (t) in Fig. 3(d) . The gap of information literacy and the ratio effect of the two communities ii. The awareness acquiring from the social media output information. In this step, every node (actually the unaware) will accept information from the social media and has a certain probability of acquiring awareness. For example, UI(h) acquires awareness and becomes AI with probability θλ 1 in Fig. 3(a) , and UI(l) not acquires awareness and remains UI with probability 1 − (1 − θ)λ 1 in Fig. 3(b) . iii. To be infected or recovered. This step reflects the epidemic diffusion. The recovery situation does not differ from the information literacy, and AI(h)&AI(l) thereby have the same probability µ to be recovered and become AS(h) or AS(l) in Figs. 3(a) and 3(b) respectively. The infection process does not need to consider the gap of information literacy. Therefore, US(h) has a probability of 1 − q U i (t) to be infected in Fig. 3 (e), and US(l) has the same probability in Fig. 3(f) . The MMCA equations for the coupled dynamics in the multiplex networks are derived according to the total probability of different status. There are two kinds of nodes in the coupled multiplex networks, the microscopic Markov chains for each node of i h are: The microscopic Markov chains for each node of i l are: Five experiments are designed to explore their functions step by step, as summarized in Table 2 . First, the information adoption efficiency of social media and aware neighbors is investigated in Experiments 1 and 2, respectively. Second, their execution abilities are compared in the experiment 3 in Section 4.3. Then, the combination of information adoption and execution ability experiment are conducted in Experiment 4 in Section 4.4. Moreover, a validation of Experiment 4 in which θ is re-designed to follow a power-law distribution is conducted in Section 4.5. Finally, the effect of information literacy gap degree and their dominated proportion are conducted in the Experiment 5 in Section 4.6. The information and epidemic coupled multiplex networks is built in our experiments. As shown in Fig. 4 , an example where each network has 1000 nodes and the same structure is given. Massaro and Bagnoli [38] pointed out that the structure's similarity rather than the difference between the awareness layer and epidemic layer networks makes a sufficiently high precaution level possible to stop the infection. Therefore, the network structures in the awareness layer and epidemic layer are set the same; the scale free network is specified in this paper considering the practice. In the epidemic diffusion layer, a power-low degree distribution network is generated with the configuration model with exponent 2.5 of 1000 nodes. In the information propagation layer another power-low degree distribution network with exponent 2.5 as well and 400 extra random links. As for the realistic scenario, it is supposed that the individuals have and have more connections in information exchange than in off-line contacts. 4 . Structure of the proposed multiplex networks. hA stands for the high information-literate nodes with awareness; hU represents the high information-literate nodes without awareness; lA and lU indicates the low information-literate nodes with and without awareness; hI is the infected high information-literate nodes, and hS is the susceptible high information-literate nodes; and lI&lS are corresponding the infected and the susceptible low information-literate nodes. Aware proportion varies according to social media information adoption efficiency. β U in x-axis is the infected probability for the unaware, ρ A , ρ Ah and ρ Al in y-axis are corresponding aware individuals' proportion. Initial information literacy parameter θ is set as 0.8, and parameter α is also set as 0.8 by default. The awareness gaining probability from social media λ 1 is adjusted from 0.1 to 0.4, and awareness gaining probability from aware neighbors λ 2 is set as 0.15. The initial fraction of high-information-literate individuals, ratio, is set as 0.5. Other fixed parameters such as forgetting probability δ is fixed at 0.6, natural recovery probability of µ is fixed at 0.4. A fundamental hypothesis is that an aware node with protective behavior has a lower probability of being infected. The factor γ is set as 0.2 to reflect the function of protective behavior in reducing the infectivity. A simple equation β A = αγ β U + (1 − α)β U can display the relationship directly. The initial fraction of aware-infected nodes ρ AI is 0.2, the fraction of aware-susceptible nodes ρ AS is 0.4 and the fraction of unaware-susceptible nodes ρ US is also set as 0.4. The topologies of the networks and the fixed parameters of processes are the same below. Social media is a vital information source for the public. The transfer of social media information into awareness efficiency reflects the self-learning ability of a person and heterogeneity of information literacy. To explore the adoption efficiency of social media information for high-and low information-literate individuals on epidemic suppression, the first experiment is designed to adjust λ 1 . Fig. 5 depicts the change in the proportion of the aware individuals of the entire node ρ A , high information-literate nodes ρ Ah , and low information-literate nodes ρ Al in terms of the infected probability β U without awareness and the social media information transferring into awareness efficiency λ 1 . The parameters not mentioned are set by default in the upward section. Overall, the higher the social media information adoption efficiency λ 1 , the more aware individuals become. When β U is sufficiently large in the stationary state, ρ Al is approximately 0.4, without any visual increase, whereas ρ Ah fluctuates from 0.25 to 0.45. Therefore, the awareness increment acquired from social media in Fig. 5(a) is mainly due to the high information-literate individuals rather than by low information-literate individuals when the information literacy is significantly different, such as 0.8 by default. In contrast to the awareness proportion, the infected ratio is not sensitive to λ 1 as observed in Fig. 6 . Although social media is a crucial awareness acquisition source, it is not the only influential factor in reducing infection probability, which makes social media transition into awareness efficiency not important in controlling epidemic diffusion in the experiment 1. Therefore, the subsequent experiments are designed to explore the independent effect of λ 2 and α, and the joint effect of λ 1 &α, λ 2 &α on suppressing the epidemic. In addition to social media, aware neighbors are another crucial information source. To explore the information adoption efficiency awareness of neighbors for the high-and low information-literate individuals on epidemic suppression, an experiment that adjusts λ 2 is designed. Fig. 7(a) , the high information-literate individuals in Fig. 7(b) , and the low information-literate individuals in Fig. 7(c) . Parameter β U reflects the severity of the epidemic. The higher β U is, the more serious the epidemic. When the epidemic is sufficiently serious, such as β U is larger than 0.5, λ 2 does not function at all. When the epidemic is relatively mild, the proportion of awareness is significantly different according to λ 2 . The benefit of awareness increment gradually decreases with an increase in λ 2 . The aware ratios of high information- In comparison with the sensitive awareness of λ 2 , the infected ratio is not sensitive, as observed in Fig. 8 . Perhaps there are several factors that affect an individual being infected, and the information of neighbors is just one of the sources used to acquire awareness. Moreover, only transforming the awareness into actual protective behavior can reduce the probability of the individual being infected in the epidemic diffusion layer. Awareness should be executed into actual protective behavior because this will decrease the probability of being infected. Thus, awareness of the execution ability is crucial in the co-evolution of an information-epidemic coupled multiplex networks. To explore the awareness of the execution ability of individuals for the high-and low information-literate individuals on epidemic suppression, an experiment that adjusts α is designed. The strong executive ability of the aware individuals is not observed in improving the entire aware proportion for the high-or low information-literate individuals, as shown in Fig. 9 . However, there are significant benefits for the awareness to decrease the probability of being infected. The larger the parameter α, the stronger the executive ability and the lower the final ratio of being infected, as observed in Fig. 10 . In particular, the high information-literate individuals exhibit a clearer difference in terms of α, which is the reason for the information literacy gap. To decrease the probability of being infected, one should first acquire awareness from social media information or re-propagated information of the neighbors, which is adjusted by λ 1 and λ 2 , respectively; the aware individual should take the actual protective behavior, which is controlled by α. Both procedures that acquire awareness and take protective behavior are indispensable for decreasing the probability of infectivity. To better elucidate the effect of social media information and re-propagated information of the neighbors on suppressing the epidemic diffusion, two experiments that combine the awareness execution were conducted. Fig. 11 shows the variation in the infected proportion according to λ 1 and α. Fig. 11 has a clear adjusting effect when α is greater than 0.25, which proves the effect of social media information propagation in suppressing the diffusion of the epidemic. However, there is insignificant infected case improvement for the high information-literate individuals' community, the low-information-literate individuals or even the entire community when λ 1 is increasing and α is smaller than 0.25. Moreover, the worst situation for ρ Ih or ρ Il is approximately 0.23. This demonstrates that the awareness execution ability parameter α may have a minimum to help to obtain a significant benefit in suppressing the diffusion of the epidemic. When α is sufficiently large and approaching 1, ρ Ih ranges from 0.23 to 0.18, whereas ρ Il ranges from 0.23 to 0.21. Clearly, the high information-literate community benefits more from a unit of λ 1 than the low information-literate community. Moreover, the best infection situation for the high information-literate-community (ρ Ih min ≈ 0.11) is much better than that of the low information-literate-community (ρ Il Consequently, to suppress the diffusion of the epidemic, some compulsive requirements are necessary to ensure that the awareness execution ability parameter α is larger than a certain value, or the information propagation work will be in vain and ineffective even for individuals with high information literacy. Unlike Fig. 8, λ 2 in Fig. 12 also has a clear adjusting effect when α is greater than 0.25, which confirms the effect of the re-propagated information of the neighbors in suppressing the diffusion of the epidemic. When α is smaller than 0.25, there are no improvements for the increase of λ 2 , which verifies the significance of the awareness execution ability. Though the execution ability is different in terms of individual heterogeneity, the government could force some policies to promote the actual self-protective behavior transition. Dividing the information-literate individuals into high-and low-information-literate individuals is a simple and effective way to take the information literacy heterogeneity into consideration, and is convenient to provide targeted suggestions. To simulate the practical situation, θ is re-designed to follow a power-law distribution with exponent 2, of which the value can attain that 20% individuals will obtain approximately 80% information literacy values. Where a series of random numbers following power-law distribution are generated at first, and then to apply 0-1 normalization method on them to ensure θ (0 < θ ≤ 1). Then, a validation Experiment 4 based on power-law distributed information literacy is conducted, in which all the parameters are in line with original Experiment 4 except θ and abandoned ratio. In accordance with Fig. 11 (a) and Fig. 12(a) , ρ I is decreasing with the increase of α, λ 1 , or λ 2 , which has verified the function of these three controllers in controlling the diffusion of the epidemic. By contrast, ρ I fluctuates small and just improves from 0.58 to 0.54 (compared with 0.46 to 0.3 in original Experiment 4. It is probable that power-law distribution has generated lots of small values of θ, which results in low awareness acquirement and therefore low improvement for ρ I from the whole perspective. Moreover, compared with uniform distributed θ in original Experiment 4, it is observed that the joint effect of λ 2 and α is slightly better than the joint effect of λ 1 and α. The minimum ρ I is 0.5504 in Fig. 13 (a) and is smaller than the minimum value (0.5490) in Fig. 13(b) , while the minimum ρ I in Fig. 11 (a) is larger than that in Fig. 12(a) . Actually, it is in great probability that the lots of small values of θ have resulted in lower efficiency of awareness acquirement from mass media information than from neighbors. Actually, it can also be divided into two classifications for power-law distributed information-literate individuals, that is, high-and low-information-literate individuals. It can be inferred that low-information-literate individuals in great ratio in original experiment may simulate the approximate effect as power-law distributed information literacy scenario. As a result, two classifications design may be more targeted. The information literacy gap reflects the degree of information literacy difference between the two communities, which might provide suggestions for the government to concentrate limited resources on narrowing the gap or strengthening the advantage of the high-information-literate community. This is meaningful in reality. To explore the effect of the information literacy gap degree and the dominated proportion of the two communities, the final experiment is designed. Fig. 14 depicts the proportion of the variation of the aware individuals according to the ratio of high-information-literate individuals and the gap in information literacy θ. In Fig. 14(b) and ρ Ah , the growth of θ exhibits a slight difference to improve ρ Ah and to depress ρ Al when the ratio is fixed. However, the entire awareness situation makes a significant difference in Fig. 14(a) . i. ρ A decreases with increasing θ when the ratio is fixed at a relatively low value, which implies that the community has a larger proportion of low information-literate individuals. θ reflects the gap degree of information literacy: the larger θ is, the more significant the gap in transforming information into awareness for the high-and low information-literate individuals. Actually, the low information-literate individuals, who are the majority, have a lower probability of 1 − θ to acquire awareness, resulting in a decrease in ρ A accompanying the growth of θ. ii. ρ A increases with the growth of θ when the ratio is fixed at a relatively high value, which implies that the highinformation-literate individuals dominate the entire community. Consequently, the attribute of the major community (the high-or low information-literate individuals) dominates the awareness proportion variation of the entire community. iii. When ratio is fixed at 0.5, the entire awareness proportion ρ A varies from 0.7510 to 0.7393 in terms of θ, as shown in Fig. 15 . This implies that the difference in information literacy does not help to improve the entire community awareness when the two types of information-literate communities have equal ratios. iv. When θ is 0.5, ρ A is 0.7510, regardless of the ratio. When θ approaches 1, the range of ρ A becomes larger. ρ Ah is positively correlated with the ratio if θ is provided, and ρ Al displays a negative correlation with the ratio if θ is provided. Fig. 16 depicts the infected proportion variation in terms of θ and the ratio. This corresponds to the variation of the aware individuals in Fig. 14. Information Sensitivity The high-information-literate individuals are more sensitive to external information. Although trends of the high-and low-information-literate individuals are similar in Experiment 1 and Experiment 2, the highinformation-literate individuals are more sensitive and show a significant difference. The infected ratio is not sensitive to λ 1 as observed in Fig. 6 , which is not quite equivalent to the result of Xia et al. [27] . Their proposed mass media generation rate m is similar to λ 1 . They observed whatever the value of β is, the larger m will greatly reduce the disease prevalence. However, they did not consider the difference and importance of protective behavior compared with awareness. The proportion of aware individuals is significantly different for different information-literate individuals. The highinformation-literate individuals are more sensitive to the variation of λ 2 at the onset of the epidemic. Therefore, the authority should enhance information propagation for highly literate individuals such as intellectuals and medical care personnel, who may benefit more in comparison with those with no target information propagation. Considering the distinguish of information sensitivity for high-and low-information-literate individuals, practical implications are concluded for sub-divided communities. Namely, aiming at different places where different ratios of information-literate individuals may exist, different suggestions and epidemic control policies should be implemented to improve the entire awareness situation and thus help to suppress the diffusion of the epidemic. In communities where the majority are high information-literate individuals, such as hospitals or schools, the gap in information literacy should be enhanced. The finding is in accordance with Manzo and van de Rijt [39] who found that targeting hubs robustly improves containment through a population embedded network simulation, in which the targeting hubs can be extended to targeting communities. In practice, the key high information-literate individuals should be provided with more useful and valuable information to help to improve their awareness of acquiring probability. However, in those low income communities where the majority are low-information-literate individuals, such as factories, information sources should be spread to all of them rather than to a few individuals, to decrease the difference in information literacy [40] . Three Controllers λ 1 and λ 2 are two controllers for public propaganda, and α is a controller for coercive policy by the government. λ 1 contributes to the acquisition of awareness. Acquiring the social media information probability for the unaware. There are several ways to improve the convenience of social media in reality. For instance, promoting 5G network development will benefit people to browse information more quickly; the wide use of mobile networks makes it easier to access online information anywhere, putting up several public service advertisements online and offline to increase the chances of receiving the epidemic related information for the public; and updating the epidemic information on time to garner public attention. Social media information is observed more important than the re-propagated information of the neighbors in suppressing the epidemic. In Experiment 3, the best suppressing epidemic situation for λ 1 with ρ I approaching 0.293 is better than that of λ 2 with ρ I approaching 0.324, which has verified it. Ultimately, social media is authoritative and normally the first information source for the majority of the public. Therefore, the government should manage social media well, from epidemic content to information updated frequency, which is vital for epidemic suppression. Improving the information adoption efficiency of the neighbors benefits significantly in terms of acquiring awareness at the beginning of the epidemic where the diffusion speed is observed to be slow. Consequently, professionals should evaluate the degree of seriousness of the epidemic at its onset and propose serious suggestions for governments to persuade officials to start propagation work. Thereafter, the officials with authority will play important hub-point roles in the social network, thereby improving λ 2 and promoting information propagation to the public. Epidemic prevention information related to α can be emphasized. Epidemic related information can be divided into many aspects, such as disease research progress, the newest infection situation at home and abroad, the breakout and control in the local area, and the public service advertisements about the epidemic. Among the public service advertisements, some operable and practical measures such as wearing a face mask when going out, washing hands frequently, sterilizing the permanent address regularly, or rejecting public gatherings will help to increase the value of α and thereby help to control the diffusion of the epidemic. Some coercive policies to increase α value and ensure it to exceed a certain threshold are necessary for suppressing the epidemic. Despite the severity of the COVID-19, Donald Trump did not wear a face mask until July 12th, 2020, and he once mocked Joe Biden for doing so [41] . In such a situation, he had a low α value, and his negative act in controlling social distance and wearing a face mask caused a low α value for the public. Meanwhile, his behavior and saying had a negative propagation effect for the public, which results in low values of λ 1 and λ 2 . All of which results in American epidemic being uncontrollable during his governing. On the contrary, Joe Biden has been propagandizing to wear a mask, which has increased λ 1 and λ 2 for his followers. While, things are not getting obvious better for his positive propagation. It is not until he was elected President of the United States that relative coercive policies were carried out to keep distance [42] to increase α value. It is turned out that, valid and big enough α is the base of λ 1 and λ 2 to be effective. To assist epidemic control in reality, this study proposed an aware-susceptible-infected model based on the microscopic Markov chain approach to explore the effect of information literacy. In the model, the self-protection related execution ability of the aware individuals which is adjusted by a parameter α is novelly introduced into the information propagation and epidemic diffusion coupled multiplex networks. Through simulation experiments, there are three main findings. Firstly, the high information-literate individuals are more sensitive to social media information and the repropagated information of the neighbors. In communities dominated by high information-literate individuals, a larger information literacy gap can improve the entire awareness acquisition and thus suppress the epidemic. Therefore in practice, the key high information-literate individuals should be provided with more useful and valuable information to improve their awareness of acquiring probability. Secondly, in communities dominated by low-information-literate individuals, even a smaller gap in information literacy can largely benefit epidemic regulation. Thirdly, the self-protection related execution ability is helpful in controling epidemic diffusion. Moreover, only the self-protection related execution efficiency attains a certain value, and social media information or the re-propagated information of the neighbors can help to suppress the diffusion of the epidemic. Practically, some coercive policies should be carried out by the government to increase α value and ensure that it exceeds a certain threshold, which is necessary in suppressing the epidemic. The findings of this study fill the gap in investigating individuals on information processing and self-protection-related execution abilities. Nevertheless, the gap in information literacy may include other information related factors, such as awareness transforming into self-protective ability, awareness transforming into information re-propagated information ability, and awareness transforming into unawareness ability, which will be covered in future work. Dynamical interplay be tween awareness and epidemic spreading in multiplex networks Two-stage effects of awareness cascade on epidemic spreading in multiplex networks Epidemic spreading with activity-driven awareness diffusion on multiplex network Coevolution spreading in complex networks Investigation of epidemic spreading process on multiplex networks by incorporating fatal properties Fundamentals of spreading processes in single and multilayer complex networks Dynamical processes and epidemic threshold on nonlinear coupled multiplex networks Framework for managing the COVID-19 infodemic: Methods and results of an online, crowdsourced WHO technical consultation What is information literacy? Survey of information literacy instructional practices in U.S. academic libraries Digital competences, computer skills and information literacy in secondary education: mapping and visualization of trends and concepts Information literacy and fake news: How the field of librarianship can help combat the epidemic of fake news Historical development of definitions of information literacy: A literature review of selected resources Workplace experiences of information literacy Conceptions of information literacy: new perspectives and implications Reframing information literacy as a metaliteracy Information literacy as a sociotechnical practice Believing the unbelievable: Understanding Young peoples information literacy beliefs and practices in the United States Why students share misinformation on social media: Motivation, gender, and study-level differences Understanding factors influencing information communication technology adoption behavior: The moderators of information literacy and digital skills Assessing internet access and use in a medically underserved population: implications for providing enhanced health information services, Health Inf Factors influencing people's health knowledge adoption in social media Spreading processes in multilayer networks Co-evolution spreading of multiple information and epidemics on two-layered networks under the influence of mass media 2016 12th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD) Effect of heterogeneous risk perception on information diffusion, behavior change, and disease transmission A new coupled disease-awareness spreading model with mass media on multiplex networks Effect of heterogeneous risk perception on information diffusion, behavior change, and disease transmission The importance of search as intertextual practice for undergraduate research Situating information literacy in the disciplines The impact of heterogeneous response on coupled spreading dynamics in multiplex networks Effects of awareness diffusion and self-initiated awareness behavior on epidemic spreading -An approach based on multiplex networks The impact of awareness diffusion on SIR-like epidemics in multiplex networks The impact of individual heterogeneity on the coupled awareness-epidemic dynamics in multiplex networks The impact of multiple information on coupled awareness-epidemic dynamics in multiplex networks Epidemic spreading in multiplex networks with heterogeneous infection rate Langevin approach for the dynamics of the contact process on annealed scale-free networks Epidemic spreading and risk perception in multiplex networks: A self-organized percolation method Halting SARS-CoV-2 by targeting high-contact individuals E-health literacy and associated factors among chronic patients in a low-income country: a cross-sectional survey Donald trump wears face mask for the first time get vaccinated or wear a mask until you do The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. No empirical data were generated during the current study. The R code for the simulation model, MMCA simula-tion results, and Monte Carlo simulation results are available on the github (https://github.com/ZuoRX/InformationLiteracy).