李祎惟，上海交通大学媒体与传播学院助理教授。

郭羽（通讯作者），澳门科技大学人文艺术学院助理教授，博士生导师。

I

Introduction

Risk communication is often conducted to increase public awareness and knowledge of risk issues, which may help them make better decisions in response to potential threats.The media are probably the most effcient platforms available to those involved in managing and communicating risks, particularly considering the rapid development of information and communications technologies (ICTs), which makes information more accessible to the general public (Harris & Sanborn, 2014). The online environment plays an increasingly important role in public opinion regarding important issues (Han, Brazeal & Pennington, 2018; Papacharissi,2004). In the meantime, it also serves as a convenient instrument, allowing users to actively search information, which may further facilitate their understanding.

Information seeking is considered a purposeful knowledge acquisition process (Johnson & Johnson, 1997) by individuals who engage in such a process when they intend to change their state of knowledge (Marchionini, 1997). As a proactive learning behavior, potential predictors of information seeking attract sustained attention in risk communication research (Yang, Aloe & Feeley, 2014). The present study attempts to explore the potential of the online information environment to promote risk information seeking.

A social-cognitive model of risk information seeking has been developed with the quality of online information, risk perception, knowledge, and information seeking self-effcacy being the most critical predictors. The model is examined in the context of nuclear radiation risk, which brings about concerns over harmful effects on both human health and environment. Therefore, effective communication and strategies for promoting public understanding of nuclear risks could become particularly important for reducing the potential social impacts. Findings of this study may not only shed light on the cognitive process by which individuals engage in proactive learning about risks but also have practical implications for risk communication strategies that aim to guide people in judgment and decision making.

II

Literature Review

A.The Shifting Focus of Risk Communication

Risk is about the probability and magnitude of potential damages or losses that are associated with a hazard (Fischhoff, Slovic, Lichtenstein, Read & Combs, 1978). Individualand societal-level reactions to risks from environmental and health issues are frequently studied (Brauch et al., 2011; Cox, 2013). Given the uncertainty of a risk situation, hazards or hazardous events often draw the attention of policy makers, risk managers and communicators on the need to take right steps toward preventing and alleviating the negative consequences (Coombs, 2012; Witte, Meyer & Martell, 2001).

Reynolds and Seeger (2005) stated that, regardless of the type of risk, action plans have always included communication strategies targeting the public. In the early development stage of risk communication, communications tended to focus on the delivery of scientifc facts and results of professional assessment (Covello & Sandman, 2001; Leiss, 1996). However, many cases demonstrated that although the numbers are scientifcally reliable, such as the potential death toll of a disaster and the prevalence rate of a disease, ordinary people quickly became desensitized to the numbers, which consequently resulted in the failure of risk communication (Anderson, 1997; Covello & Sandman, 2001).

Based on the lessons learned from the pedagogical means of communication, risk communication began to make an effort to interpret scientifc data (Covello & Sandman, 2001). Those involved in managing and communicating risks designed better information systems and cooperate with the media, which greatly facilitated the public' s understanding of a particular issue. However, merely knowing the risk seems to be not enough. The sense of control over a certain hazard, which is likely to be associated with knowledge, can lead to optimism bias in response (Covello & Sandman, 2001). That is, people are very likely to develop unrealistic beliefs that they are less at risk of being affected by a negative event (Weinstein, 1989). Accordingly, beyond only satisfying the need to be informed, strategies that can motivate people to take precautions and encourage them to make better judgments and decisions are badly needed.

Insights from behavioral theories may contribute to further development of risk communication strategies, for instance, the social cognitive theory (Bandura, 1986), which has demonstrated broad applicability to human behaviors in diverse contexts. According to the theory, human beings can actively learn from their immediate environments through direct observation or from the symbolic environment created by the media through vicarious observation (Bandura, 1977, 1986). Empirical research based on the social cognitive framework suggested that people tend to develop their behavior patterns through such observational learning, and there are several cognitive and psychological factors, including affect and effcacy beliefs, playing critical role in this process (e.g., Bandura, 2009; Namkoong, Fung & Scheufele, 2012; Prati, Pietrantoni & Zani, 2011; Stacey, James, Chapman, Courneya & Lubans, 2015).

Information seeking behavior is increasingly documented in the contexts of environmental and health issues as a key coping strategy in self-protective activities and psychosocial adjustment to uncertainty and anxiety (e.g., Kellens, Zaalberg & Maeyer, 2012; Kievik & Gutteling, 2011; Mileti & Darlington, 1997; Moldovan-Johnson, Martinez, Lewis, Freres & Hornik, 2014; Paton & Johnston, 2001; Yang, Aloe & Feeley, 2014). In this sense, risk communication may be more effective if it is individual-focused and able to promote information seeking. Accordingly, studies from the perspective of well-developed behavioral theories such as the social cognitive theory may yield new insights on the prediction of risk information seeking behavior.

B.Previous Risk Information Seeking Models

Previous studies have provided valuable findings on the prediction of individuals' information seeking behavior. The most insightful models are the Risk Information Seeking and Processing (RISP) Model (Griffn, Dunwoody & Neuwirth, 1999) and the Planned Risk Information Seeking Model (PRISM) (Kahlor, 2007, 2010), which have been repeatedly employed in varied contexts (see fndings of a meta-analysis by Yang, Aloe & Feeley, 2014). Previous fndings based on the two models have enriched our knowledge about determinants of information seeking. Particularly, the individual-level determinants come to the fore.

The RISP model was developed by Griffn et al. (1999) to predict individuals’ seeking and processing of health risk information. The model was built upon the adaptation and synthesis of the Heuristic Systematic Model (Chaiken, Liberman & Eagly, 1989) and the theory of planned behavior (Ajzen, 1991), striving to construct an audience-based model that could be helpful for the development and maintenance of preventive health behavior among the public. Focusing on the individual-level predictors, the RISP model proposed seven factors that have potential impacts on information seeking and processing, including individual characteristics, risk perception, affective response, information subjective norms, information insuffciency, outcome expectation, and perceived information gathering capacity. Results of model testing showed that outcome expectation, perceived information gathering capacity, affective response to hazard, informational subjective norms, and perceived information suffciency were direct predictors of information seeking and processing behavior. The model also suggested the importance of current knowledge, which was a signifcant predictor of perceived information suffciency and perceived information gathering capacity.

Following the primary assumptions of the RISP model, Kahlor (2007, 2010) developed the Planned Risk Information Seeking Model. The model was frst introduced as an augmented version of the RISP model (Kahlor, 2007), which treated risk information seeking as a deliberate behavior. Based on a thorough review of main concepts of previous models, the PRISM had been built by underlining the role of individual-level factors in order to predict information seeking across contexts (Kahlor, 2007; Moldovan-Johnson, Martinez, Lewis, Freres & Hornik, 2014). The resulting integrated model indicated that attitude toward seeking and perceived seeking control were signifcant direct predictors of the intention to seek risk information. However, different from the RISP model, the PRISM demonstrated no signifcant association between perceived knowledge and the outcome variable. These fndings suggested that, regardless of the current state of knowledge, as long as people have positive outcome expectancy toward information seeking and confdence in their information gathering capacity, they will likely engage in information seeking behavior.

C.The Present Study

a.Symbolic Environment, Knowledge, and Perception of Risk

As afore-reviewed, risk is about the probability and the magnitude of potential negative consequences that are associated with a hazard or hazardous event (Fischhoff, Slovic, Lichtenstein, Read & Combs, 1978). According to the theory of risk society (Beck, 1992), the extent to which those future threats are known to the general public largely determines the policies and measures for dealing with the risks (Bhme, 1997; Bhme & Stehr, 1986; Stehr, 1994). In this sense, the risk society is also described as a knowledge society, which features the process of conceptualizing risks and identifying the characteristics of risks (Strydom, 2002).In the modern society, compared to other communication channels, the media, particularly Internet-based media, play an outstanding role in producing, interpreting, and spreading riskrelated knowledge (e.g., Allan, Adam & Carter, 2000; Kasperson et al., 1988; Kitzinger, 1999; Kull, Ramsay & Lewis, 2003; Stern & Fineberg, 1996; Dahlstrom, Dudo & Brossard, 2012; Miles & Morse, 2007; Wahlberg & Sjberg, 2000), which makes the risk society also to be seen as a communication society (Delanty, 1999).

According to the social cognitive theory, the symbolic environment created by the media, in particular the online environment, is the place where people can learn perceptions, attitudes, and behavior patterns more conveniently and efficiently than they do through direct observation in their immediate environments (Bandura, 1986, 2001). Therefore, the symbolic environment also tends to be one of the major sources of risk-related information for most people, which is likely to determine how they comprehend a particular risk. Besides knowledge, risk perception—the intuitive and subjective judgment of risk among nonexperts—receives great attention in risk communication research, and the media is among the wide array of factors that can influence risk perception (Kasperson et al., 1988; Wachinger, Renn, Begg & Kuhlicke, 2013; Wahlberg & Sjberg, 2000). Instead of investigating general patterns of media use, the present study takes the step to examine the potential influence of information characteristic, namely, the quality of information, which may have more specifc implications for communication practices. A suffcient amount of information with easy access, from credible sources, understandable, and showing the complete information is very likely to compensate for people's shortage of knowledge. With regard to the relationship between knowledge and risk perception, it was originally assumed that a large volume of risk-related knowledge could lead to a low level of risk perception because knowledge might ensure a more comprehensive evaluation of risks to reduce people’s sense of uncertainty. However, most of the existing evidence has rejected this assumption and suggested that knowledge tends to increase people's risk perception (Wachinger, Renn, Begg & Kuhlicke, 2013).

Overall, this study proposes the following hypotheses to explore the influence of information from social media on ordinary people’s knowledge and risk perception of nuclear radiation.

H1: Information quality is positively associated with knowledge of nuclear radiation.

H2: Information quality is positively associated with risk perception of nuclear radiation.

H3: Knowledge is positively associated with risk perception of nuclear radiation risk.

b.Self-effcacy and Outcome Expectation

Self-efficacy and outcome expectation are key concepts in the social cognitive theory (Bandura, 1986). The theory claims that people can control and regulate their behavior in a goal-directed way. Through the observing and learning process, people evaluate their own abilities and anticipate the outcomes of a certain behavior. The stronger people’s confdence in their ability to execute a behavior (self-efficacy), the more likely that they will perform the behavior; in the meantime, if the outcome expected (outcome expectation) is valued, the frequency of behavior will increase, and when unfavorable outcomes are expected people will reduce the frequency of behavior (Bandura, 1997). The present study then assumes that selfefficacy and outcome expectation regarding risk information seeking are closely associated with the actual behavior.

Mastery experience and psychological state are claimed to be two major individuallevel contributors of self-efficacy and outcome expectation (Bandura, 1997). In the context of information seeking, it is plausible to assume that high-quality information and knowledge people obtained can generate the sense of mastery and control; in the meantime, they also provide the most authentic evidence to assure people that they will succeed in future information seeking and learning. Concerning the psychological contributors of self-effcacy and outcome expectation, risk perception can serve as the stimulus, persuading people to instill more beliefs in their abilities and positive outcomes to boost precautionary behavior such as information seeking.

Therefore, based on Bandura’s (1997) explanation of the sources of effcacy beliefs, as well as the direct relationships between efficacy beliefs and the given behavior, this study proposes the following hypotheses to further explore how information from the social media environment can motivate people to seek additional information about nuclear radiation.

H4: Information quality is positively associated with perceived self-effcacy and outcome expectation regarding information seeking.

H5: Knowledge of nuclear radiation is positively associated with perceived self-effcacy and outcome expectation regarding information seeking.

H6: Risk perception of nuclear radiation is positively associated with perceived self-effcacy and outcome expectation regarding information seeking.

H7: Perceived self-efficacy and outcome expectation are positively associated with information seeking.

III

Method

A.Procedures and Participants

Data used in this study were collected via an online survey among Chinese social media users who were living near nuclear power stations. Participants were recruited from a panel of a research institute by sending an invitation email with the URL link of the web-based questionnaire. A total of 393 valid responses with a response rate of 56.8% were gathered (male=201, female=192). Age was ranging from 18 to 63 years old (M=33.44). Regarding socioeconomic status, 20.4% of the respondents had a monthly income which was equal to or less than 3,000 RMB, 54.2% of the respondents had a monthly income at the level of 3,001 to 8,000 RMB, and 25.4% of the respondents reported they had a monthly income which was more than 8,000 RMB. A majority of the respondents (67.7%) had a bachelor’s degree, and 7.4% of the respondents had a master’s degree or doctorate.

B.Measures

Social media information quality was measured by five items modified from previous studies (Dutta, Pfster & Kosmoski, 2010; Wang & Strong, 1996). Participants were asked to evaluate the accessibility, credibility, understandability, suffciency, and comprehensiveness of information they obtained from social media. All the items were rated on 7-point scales (1=not at all, 7=extremely, α=.95).

Knowledge was measured by asking the participants how they thought they knew nuclear radiation. The item was rated on a 7-point scale (1=none, 7=a great deal).

Risk perception can be quantified through identifying individuals' evaluations on the attributes of a particular risk such as likelihood, dread, controllability, among others (see the psychometric paradigm in Slovic, 1987). This study selected four attributes from previous research that have been frequently employed in studies associated with environmental risk (e.g., Barnett & Breakwell, 2001; Duan, 2005; Fischhoff, 1995; Lai & Tao, 2003; Slovic,1987; Wachinger, Renn, Begg & Kuhlicke, 2013; Zhang, He, Mol & Lu, 2013). Participants were asked to indicate the extent to which they thought nuclear radiation was risky, dreadful, catastrophic, and worried about nuclear radiation. All the items were rated on 7-point scales(1=not at all, 7=extremely, α=.84).

Self-efficacy was assessed by three items adopted and modified from previous studies (Griffin, Dunwoody & Neuwirth, 1999; Kahlor, 2007, 2010; Kellens, Zaalberg & Maeyer, 2012; Yang, Aloe & Feeley, 2014). Participants were asked to indicate the extent to which they had confdence in their abilities to understand, search for, and evaluate the credibility of information they got. All the items were rated on 7-point scales (1=not at all, 7=extremely, α=.92).

Outcome expectation was assessed by three items designed especially for the current investigation by referring to the social cognitive theory. Participants were asked to indicate the extent to which they believed information seeking would increase their knowledge, make them alert to, and let them be well prepared for nuclear radiation risk. All the items were rated on 7-point scales (1=not at all, 7=extremely, α=.89).

Information seeking was measured by three items adopted from previous research (Kuhlthau, 1991). Participants were asked to indicate the extent to which they had intentions, attended to, and actually searched for information about nuclear radiation risk. All the items were rated on 7-point scales (1=not at all, 7=extremely, α=.92).

It is well to mention here that the psychometric literature suggests that seven-point scales are a little better than 5-points (e.g., Nunnally & Bernstein, 1994; Finstad, 2010). In general, comparing to fve-point scales, seven-point scales increase the variance in the measure, offer respondents wider range of stimuli, and tend to produce better distributions of data (see Finstad, 2010). In addition to that, using 7-point scales by following previous studies opens up possibilities for comparative analysis, which can help elucidate the generalizability of relationships among key variables and contribute to theory development.

IV

Results

To examine the hypotheses proposed in this study, structural equation modeling was performed with AMOS. The resulting model shown in Figure 1 (X2=454.21, df=141, X2/df=3.22<5) is signifcant at the 0.001 level. Other statistics suggested that the resulting model had an adequate ft for the data: the comparative ft index (CFI) is .95>.90, the Tucker-Lewis index (TLI) is .94>.90, the root mean square error of approximation (RMSEA) is .08 (90% confidence interval=.07~.08, based on 1000 bootstrap replicates), and the standardized root mean square residual (SRMR) is .04<.05. Overall, the resulting structural model explained 75.5% of the variance in risk information seeking.

H1 assumes a positive association between social media information quality and knowledge of nuclear radiation. Results shown in Figure 1 supported this hypothesis, suggesting that information quality was a signifcant positive predictor of knowledge (β=.36, p<.001). In the meantime, results demonstrated that information quality (β=.17, p<.01) and knowledge (β=.17, p<.01) were significant positive predictors of risk perception of nuclear radiation, which supported H2 and H3.

With regard to the prediction of efficacy beliefs regarding risk information seeking, results (see Figure 1) frst indicated that social media information quality had great potential to contribute to self-effcacy (β=.57, p<.001) and outcome expectation (β=.48, p<.001). Thus, H4 was supported by this study. Knowledge of nuclear radiation was a significant positive predictor of self-effcacy (β=.22, p<.001) but it had no predicted effect on outcome expectation (β=.09, p=.06). Therefore, H5 was partially supported. Risk perception of nuclear radiation was found to have no predictive power on effcacy beliefs (self-effcacy: β=-.03, p=.58; outcome expectation: β=.07, p=.20). Hence, H6 was rejected by this study.

H7 posits that efficacy beliefs regarding risk information seeking are directly and positively associated with the behavior. Results in Figure 1 demonstrated that perceived self-efficacy can significantly contribute to information seeking (β=.82, p<.001), however, outcome expectation lacks power to predict information seeking (β=.06, p=.41). Accordingly, H7 was partially supported by this study. It is worth noting that although risk perception had no signifcant association with effcacy beliefs, it was a signifcant positive predictor of risk information seeking (β=.16, p<.001). In addition to risk perception, knowledge was also found to have the potential to directly contribute to information seeking (β=.12, p<.001).

Furthermore, to identify important pathways from social media information to searching for additional risk-related information, the potential mediating roles of knowledge, risk perception, and self-efficacy as indicated in Figure 1 were examined. Results presented in Table 1 suggested that all indirect effects examined in this study were statistically signifcant.

V

Discussion

Drawing upon insights from existing literature on risk information seeking and the social cognitive theory, the present study focused on the role of the online information environment in public reactions to environmental issue and developed a social-cognitive model of information seeking in the context of nuclear radiation risk in China. Similar to previous risk information seeking models such as the RISP model (Griffn, Dunwoody & Neuwirth, 1999) and the PRISM (Kahlor, 2007, 2010), the social-cognitive model suggested by this study also highlighted the power of individual-level factors in predicting information seeking, including knowledge, risk perception, and self-efficacy regarding the behavior. More importantly, fndings of this study extended previous work by exploring and examining the motivating role of the environmental factor, namely, the symbolic environment created by the media, which have implications for risk communication practices that aim to promote proactive learning and precautionary behaviors among the general public.

Perceived self-effcacy was found to have the strongest predictive power on information seeking. This fnding supported Bandura’s assertion, which considered self-effcacy to be the most powerful one among numbers of factors affecting human behavior (Bandura, 1995). Our fnding is also consistent with previous literature (Kievik & Gutteling, 2011; Schwarzer & Fuchs, 1995), suggesting that self-efficacy tends to have the closest association with the behavioral variable and to be the most signifcant predictor of behavior change. As expected, knowledge that people believe they have about nuclear radiation seems to be a major source of self-efficacy for finding additional credible and useful information. Our resulting model (see Figure 1) also demonstrated a direct pathway from knowledge to information seeking. However, inconsistent with the hypothesis, instead of serving as the source of self-effcacy, risk perception—affective response and a mental shortcut to decision making (Slovic, Finucane, Peters, & MacGregor, 2010)—has the potential to directly lead to information seeking behavior. These direct associations between knowledge/risk perception and risk information seeking found by this study are generally consistent with the RISP model, further suggesting the importance of communication strategies for increasing public knowledge of risk and appropriately amplifying public risk perception.

By focusing on the social media environment, this study found that information from the symbolic environment played a signifcant role in inuencing knowledge and risk perception of nuclear radiation and self-efficacy regarding information seeking. More specifically, the better the quality of information is, the higher the levels of knowledge, risk perception, and self-effcacy will likely be. It seems that information of good quality can not only facilitate a more comprehensive understanding of a risk issue but also make a positive impression, helping people develop confdence in future information seeking. The current exploration of the potential impacts of information characteristics offers practical insights to communication practitioners. Nuclear risk communication should be information-oriented, providing suffcient and easy to access information which is credible, understandable, and unbiased. Besides information that is in line with the government’s pro-nuclear policies, such as expansion plan and commercial nuclear power information, information explaining the potential adverse effects of nuclear energy should also be provided for increasing public awareness and motivating proactive learning. However, a decade-long content analysis of nuclear power related news articles in two state-owned media in China only found three articles that opposed nuclear power (Wang, N. Li & J. Li, 2014). Previous research also suggested that in general the Chinese public is poorly informed about the negative aspects of nuclear energy and the management of nuclear risks (He, Mol, Zhang & Lu, 2013).

In terms of the platform for risk communication and risk information seeking, the online environment, particularly the social media, provides assorted alternative information sources to the traditional ones and heterogenous information, which greatly increases people’s opportunity of cultivating a comprehensive view of an issue (Kietzmann, Hermkens, McCarthy & Silvestre, 2011; Yin, Lampert, Cameron, Robinson & Power, 2012). Moreover, the technological advantages of low cost and easy access of Internet-based media may also help users build confdence in information seeking. Previous research has suggested the promising role of Internetbased media in risk and crisis communication (Goolsby, 2010; Sutton, Palen & Shklovski, 2008). However, Internet-based media also raise concern over their lack of effectiveness in attracting sustained attention to a hazardous or risk issue. For instance, on social media, although an issue gets extensively covered and shared by verifed accounts to draw wider attention, it tends to be quickly overshadowed by other topics. As a result, the richness of information may counteract the effects of information quality. Future research is encouraged to further explore to what extent information spread via Internet-based media and social networks actually increase public awareness of risk. Overall, although almost nothing can compete with the effectiveness of the Internet and social media in disseminating information, the disadvantages of using it and potential barriers to its effectiveness are worthy of closer scrutiny.

As important as the findings and implications, several limitations of this study need to be acknowledged. The first limitation relates to the measurement of knowledge. This study assessed knowledge by asking the participants how much they thought they knew nuclear radiation. Results reected subjective knowledge, which might differ from the actual knowledge level among the sample. Future research should consider including quiz into the survey to evaluate the general public’s risk-related knowledge. The second limitation relates to the disadvantages of online survey. As an exploratory investigation of Chinese people’s reactions to nuclear radiation, this study focused on residents living near nuclear power stations. Web survey service offered us convenient access to this particular population. However, such non-probability sampling may limit generalization of findings. Future research should try to get a representative sample of the target population. Meanwhile, regarding the potential inuence of the media, future research should include examinations of different media formats to improve understanding of how traditional and new media function in risk communication. Moreover, this study used cross-sectional data for hypotheses testing. However, static analysis can indicate potential causal relationships but is not persuasive enough to determine the causation. Therefore, longitudinal studies are encouraged to provide more defnite confrmation of media inuence on the public’s cognitive, affective, and behavioral reactions to risk.

VI

Conclusion

Drawing upon insights from existing information seeking models, the present study provides additional evidence on the predictive power of individual-level factors. Moreover, by exploring the role of social media in the context of nuclear radiation risk, this study extends previous works and suggests potentials and barriers for communicating risk issues via online information environment. Overall, in response to the shifting focus of risk communication research, the social cognitive model developed by this study shed new light on how individuals engage in purposeful knowledge acquisition process; practically, findings of the current investigation have important implications for risk communication strategies to motivate the lay public to learn about the given issue, which can help them make better decisions to prevent the potentially unintended and undesirable consequences. To further examine the applicability of the model, future research may consider investigating whether the cognitive pathways to information seeking will vary according to types of risk.

本文系简写版，参考文献从略，原文刊载于《国际新闻界》2020年第4期。

封面图片来源网络

本期执编 / 小束