Understanding of the concept of ‘uncertain risk’. A qualitative study among different societal groups

Abstract In environmental health science, the concept ‘uncertain risk’ refers to situations in which epistemic uncertainties prevent definitive statements about the presence or existence of risk. The concept is difficult for risk communication and may easily lead to miscommunications and misunderstanding. This research investigated how the concept of uncertain risk is understood by different societal groups. We conducted in-depth interviews with the general public (N = 40), and online open-ended questionnaires and online focus groups with scientists, policy makers and communication experts (N = 49). Results show that the understanding of ‘uncertain risk’ varied within and between groups. While at least some representatives of all groups described epistemic uncertainties, groups focused on different aspects of risk. That is, scientists based their descriptions only on uncertainties in risk assessment. Policy makers commonly indicated that these uncertainties caused uncertainty in risk management. Communication experts mainly gave general descriptions, but recognized aspects that were important for other groups. Representatives of the general public often described aspects that deviate from the scientific understanding, such as a lack of personal control. In risk communication, stating that ‘a risk cannot be excluded’ is not clear. Different associations and understanding have to be taken into account in risk communication. Explicating uncertainties may improve consistent understanding and foster a more balanced dialogue between different societal groups. Further (empirical) research is needed to demonstrate if and how explicating evidence and uncertainties in specific cases of uncertain risks improves understanding within and between groups.


Introduction
The concept of 'uncertain risk' is used to indicate there is uncertainty about the presence or existence of risk (Tosun 2013;Van Asselt and Vos 2008;Zander 2010). In environmental health may infer that there is a risk while the communication intends to convey it is unclear whether a risk exists. The concept, however, is often loosely defined and used in the policy and risk governance literature (Renn 2008a,b;Tosun 2013;Van Asselt and Renn 2011;Van Asselt and Vos 2008;Vos and Everson 2009;Zander 2010). For example, Van Asselt and Vos (2008) define uncertain risks as 'new, imaginable hazards, with which society has no or limited experience'. Tosun (2013) defines uncertain risks as ' [ … ] situations in which the relationship between activities and their potential hazard cannot be established'. In order to further clarify the scientific meaning of the concept 'uncertain risk' we conducted a conceptual review of the environmental health literature (Jansen et al. 2017) and tried to answer about what aspects of risk uncertainty exists (location of uncertainty) when scientists refer to 'uncertain risk'. The review showed that the uncertainty about the presence or existence of risk may be the result of six different epistemic uncertainties ( Figure 1). Epistemic uncertainty may exist about (1) whether an agent has hazardous properties; (2) what adverse effects the agent causes; (3) at what levels of exposure these effects occur; and (4) whether a suggested relationship exists between an agent and certain effects; (5) the source of adverse effects; or (6) actual exposure levels. Thus, 'uncertain' in 'uncertain risk' does not refer to the realization of outcomes as it does in the concept of 'risk' in the context of environmental health (US EPA 2015) but to the causes or 'initiating events' (Aven 2010).
When one or more of these uncertainties are present in scientific risk assessment, quantitative statements about probabilities of occurrence are arguably not be the best way to adequately express the uncertainties present in risk assessment (see e.g. Zander 2010). Moreover, depending on the specific 'uncertain risk case', the level of evidence and uncertainty about the presence of risk differs. For example, communicating uncertainty about the intrinsic properties of an agent (1), for example in newly engineered nanomaterials, implies a high level of uncertainty because there also is uncertainty about nature of effects (2) and the nature of the hazard (3). In contrast, uncertainty about the nature of the hazard, for example in chemicals with unclear toxicity, implies a lower level of uncertainty because there has to be sufficient evidence for an agent's hazardous intrinsic properties and a certain effect must be associated (Jansen et al. 2017). These differences are not clear in the concept of uncertain risk, in verbal presentations of uncertain risks, such as 'a risk cannot be excluded' or 'there possibly is a risk', or in numerical expressions of probabilistic risk. This underspecificity (Carey and Burgman 2008) presents a first issue for risk communication about uncertain risks.
Furthermore, many different groups take part in public debates about uncertain risks. Research has shown that non-experts as compared to experts have different knowledge and beliefs about environmental risks such as health risks due to electromagnetic fields (Claassen, Bostrom, and Timmermans 2016), wildland fires (Zaksek and Arvai 2004) or radon (Bostrom, Fischhoff, and Morgan 1992). These existing knowledge and beliefs form the mental framework in which new information is evaluated. This has implications for risk communication because different societal groups need information that is in line with their existing knowledge and beliefs (Morgan et al. 2002) to support decisions. For risk assessors, information about the uncertainties associated with scientific risk assessment may be important to be able to weigh the evidence underlying their scientific advice about uncertain risks. For policy makers, it may be important to convey how these uncertainties influence risk management options (Thompson and Bloom 2000). Non-experts, not familiar with the risk assessment process, do not clearly differentiate between different types of uncertainties in communications, and, therefore, may draw mistaken conclusions (e.g. about the presence or severity of a risk) (Miles and Frewer 2003;Wiedemann, Sch€ utz, and Thalmann 2008). Non-experts may also be interested in information other than uncertainties in risk assessment, such as information about how someone can be exposed or whether the risk can be controlled (Lion, Meertens, and Bot 2002). When such topics are not covered in risk communication, the odds of misinterpretation are high and a balanced dialogue between different societal groups may be impeded.
With the many ways 'risk' and 'uncertainty' can be understood; the specific situations in which the environmental health literature refers to 'uncertain risk'; and the different mental frameworks in which information is evaluated it is likely that communications about 'uncertain risks' may lead to miscommunications and misunderstanding. A better understanding of how the concept is understood within and between various societal groups is important for improving the match between risk assessment, communication practices, and risk perceptions of different societal groups. This study aims to explore how the general public, scientists, policy makers, and communication experts involved in the assessment and communication about environmental health risks understand the concept of 'uncertain risk'.

Method
A qualitative public-centered approach was used. We included four societal groups: the general public, and three expert groups that play a formal role in the assessment of and communication about uncertain risks in the Netherlands: scientists, policy makers, and communication experts. 1 Representatives of the expert groups had to be experts in their respective fields.
First, we performed in-depth, open-ended semistructured interviews with 40 representatives of the general public. After this, 49 representatives of the expert groups completed an online open-ended questionnaire based on the same questions. Finally, after completing the questionnaire, 45 of the expert representatives participated in online focus groups. Below we discuss these three steps in further detail.

Interviews with members of the general public
Members of the general public were recruited through a snowball sampling method. We aimed to achieve a balanced male-female ratio, including individuals with a non-native Dutch cultural background, of all levels of education (low, intermediate, high; according to the International Standard Classification of Education [ISCED] 2011 (UNESCO Institute for Statistics (2012)), and from different parts of the Netherlands. See Table 1 for an overview of participant characteristics. Participants did not receive any compensation for their participation.
Participants were interviewed in their own homes or, when more convenient for the respondent, at a public location offering some privacy. The interviews generally lasted between 30 to 45 min and were audio recorded with the participant's permission. All interviews were conducted between August and November 2016 by the first author. The interview followed a semistructured protocol and started by introducing the goal and general topic of the research.
This brief introduction was followed by open-ended questions addressing the participant's understanding of the concept of 'risk' and of 'uncertain risk'. After these questions, the interview addressed a specific case of an uncertain risk. See Table 3 for the questions discussed in this article.

Open-ended questionnaires and focus groups with experts
Experts from public authorities, universities, hospitals and ministries involved in the assessment of and/or communication about environmental health risks in the Netherlands were approached to participate in our study. Participants were invited based on expertise in their respective fields (science, communication and policy). Scientists were selected based on their expertise on: (1) antibiotic resistance; (2) chemical substances in food packaging materials; or (3) nanomaterials in food. In total, 45 experts divided across three expert groups agreed to participate in the study. See Table 2 for an overview.
First, experts received an online questionnaire. Table 3 presents the questions discussed in the current article. Questionnaires were completed in February and March 2017. Next, the experts participated in a focus group with 2-4 other participants within their field of expertise through video conferencing. 2 Focus groups were chaired by scientific experts from the same field. The chairs were instructed to intervene as little as possible and to focus on moderating the process. The focus groups lasted 1 hour and were audio-and video-recorded with the participants' permission. All focus groups were conducted in March 2017. Table 3 provides a complete overview of methods, questions and statements addressed in the current article.

Qualitative coding and analysis
The audio recordings of the interviews and focus groups were transcribed verbatim. Data from the online open-ended questionnaire and the transcriptions were analyzed using Atlas.ti 7 (Friese 2014). Qualitative data analysis comprised two phases. In the first phase, the first author applied descriptive level coding to text fragments that were identified to be relevant. We decomposed participants' descriptions of their understanding of uncertain risks in specific aspects that provide meaningful information about their understanding of the concept of 'uncertain risk'. Descriptions of individual participants could consist of one or several aspects. Judgment of relevance of text fragments was guided by the three dimensions of uncertainty proposed by Walker et al. (2003). The second author independently coded part of the data to ascertain that no relevant fragments had been missed and to validate the descriptive level coding. The second coder identified a few additional text fragments to be relevant. The two coders discussed and resolved minor differences in the coding and in the second phase categorized the descriptive level codes to create higher-level, conceptual coding. The third authorwho did not participate in the first phase of the coding processreviewed the categorization of all descriptive level codes. Categorization was subsequently discussed by the first and third authors and minor adjustments were made. Simultaneously, the second author coded a part of the data using the conceptual codes. No additional text fragments were identified. In total, 10% of the data was coded by the first two authors.

Results
Within the studied sample, the aspects used to describe their understanding of the concept of 'uncertain risk' varied within and between groups. Some individuals described multiple epistemic uncertainties found in Jansen et al. (2017), while others (including experts) described only one. Additionally many different associations were found in the participants' descriptions of the concept of 'uncertain risk' (Appendix A). Some participants of the general public stated that the distinction between 'risk' and 'uncertain risk' was not clear. Many participants had negative associations with risk and specifically put forward that 'risk' is directly related to 'danger', inherently uncertain, or represented a deliberate choice (e.g. 'taking a risk'). Some participants stated that 'a risk' represents something that could either go wrong or right; that risks are part of our daily lives; and that risk also has positive aspects.  Some organization did not have a communication advisor for the specific case addressed in the remainder of the interview. In these cases, professionals were put forward who would normally answer questions regarding this subject. c One toxicologist was put forward as policy expert. In our results we follow (Neale, Miller, and West 2014) and make a distinction between commonly mentioned aspects, and rarely or occasionally mentioned aspects with the aim to clarify patterns in the qualitative data. Because of the explorative and qualitative nature of this study the results are not meant to convey generalizability beyond the study population (Neale, Miller, and West 2014). Table 4 presents all aspects that were commonly used to describe the understanding of the concept of uncertain risk by at least one group. Table 5 presents example quotes for each of these aspects. A single quotation can describe several aspects and could, therefore, be allocated to multiple categories.
The commonly described aspects were categorized in six overarching themes and two detached aspects. 1. Participants often described causes of uncertainty that define the concept of 'uncertain risk'.
For example, participants mentioned that scientific knowledge can be insufficient, inconsistent or contested. Additionally, participants from the general public put forward a lack of personal awareness and knowledge. 2. Many participants identified specific uncertainties about qualitative properties of risk, such as uncertainty about the nature of (adverse) effects, and uncertainty about, or due to, delayed effects (i.e. effects that occur in the long term). Specific uncertainties were often described in the context of a specific example or in abstract terms (see e.g. the quotations referring to delayed effects and actual exposure levels in Table 5). Level of uncertainty is reported here as a cause of uncertainty because each of the uncertainty aspects not related to a cause of uncertainty implicitly indicates a (different) level of uncertainty. 3. Participants from all groups suggested that the concept of 'uncertain risk' indicated that there was uncertainty about quantitative properties of risk. For example, uncertainty about the probability of occurrence (i.e. one in a hundred or one in a thousand) was one of the commonly mentioned elements for all groups. A few members of the general public understood the concept of 'uncertain risk' to refer to a risk with a small chance, or extra small chance (relative to the probability that the concept of 'risk' referred to). 4. Scientists in particular made general statements about the risk assessment process by stating that uncertain risks are difficult to assess or quantify. Participants from the general public stated that uncertain risks are difficult to assess by the public. 5. Several uncertainties about risk management were put forward. For example, uncertainty about the effectiveness of policy options, but also uncertainties about the action perspectives for the general public. Uncertain risks were commonly associated with a lack of personal control by participants from the general public. 6. When no specific uncertainties were put forward, participants typically made general uncertainty statements about risk. For example, participants in each group suggested that the concept of 'uncertain risk' refers to uncertainty about the presence of the risk and/or that 'uncertain risks' were defined by uncertainty about the nature of the risk. Finally, and not classifiable in any of the themes above, 'uncertain risk' evoked statements that referred to uncertainty about individual risk (i.e. will an individual experience adverse effects or not) and participants of the general public expressed that 'uncertain risks' referred to situations that are unpredictable. Figure 2 shows the overlap between commonly mentioned aspects by the different groups involved in this study. It shows that all groups commonly mentioned uncertainties associated with risk assessment and commonly made general statements about uncertain risks (e.g. 'uncertainty about the presence of risk'). Scientists and representatives of the general public commonly put forward aspects that were not shared with any other group (e.g. actual exposure levels and a lack of personal [public] knowledge). Communication experts put forward one general statement (uncertainty about the severity of risk) that was not (commonly) mentioned by other groups. All groups mentioned unique topics that are relevant for their own (professional) group. For example, policy makers put forward that 'uncertain risks' also signified uncertainties in risk management. Members of the general public specified uncertainties that pertained to personal relevance (e.g. lack of control). Communication professionals (shaping communication between different groups) identified uncertainties that were considered relevant for specific groups: uncertainties in risk assessment for scientists, uncertainty in risk management for policy makers, and uncertainty about action perspectives for the public. Scientist relied on uncertainties associated with risk assessment in their descriptions of the concept of uncertain risk.

Discussion
This research aimed to gain insights into how the scientific concept of 'uncertain risk' is understood by different societal groups. The concept intends to describe situations in which epistemic uncertainties prevent definitive statements about the presence or existence of risk. Results show that the understanding of 'uncertain risk' varied within and between groups. While at least some representatives of all groups described epistemic uncertainties, some recognized that the concept could refer to multiple epistemic uncertainties, while others only put forward one. Different groups also focused on different aspects of (uncertain) risk. Scientists based their descriptions only on uncertainties in risk assessment. Policy makers commonly indicated that these uncertainties in risk assessment caused uncertainty in risk management (options). Communication experts mainly gave general descriptions, but also recognized aspects that were considered important for other groups. Representatives of the general public often described associations with uncertain risks that deviate from the scientific understanding, such as a lack of personal control. Additionally, participants commonly mentioned uncertainties (e.g. 'uncertainty about the magnitude of the risk') that imply the understanding that a risk exists. The general public in particular attributed many aspects to 'uncertain risk' that deviate from the scientific-technical concept. Most of these attributions were not mentioned by scientists, policy makers, or communication experts. 3 For example, members of the general public commonly mentioned a low perceived likelihood and uncertainty due to delayed effects. These factors are identified to attenuate people's risk perceptions (Slovic 2016). In addition, representatives of the public stated that they 'do not concern themselves with' uncertain risks. This indicates that part of the public may not find uncertain risks an important issue and, as a consequence, are not worried about 'uncertain risks' (Powell et al. 2007).
However, some members of the general public expressed that 'uncertain risks' refers to 'risks that they are not aware of', that is, environmental health risks that are known to science or policy but are unknown to the public. An example is an incident in which tap water in contaminated, that is not communicated to the public because exposure levels are too low to cause adverse effects. If the public is informed by the media about these risks and not by the government, this may undermine their trust in the government, even though there was no risk to personal health (Powell et al. 2007). Additionally, a feeling of a lack of personal controlalso commonly mentionedmay amplify public perception of the seriousness of risk (Miles and Frewer 2003;Slovic 2016).
Some participants did not differentiate between the concept of 'uncertain risk' and 'risk', and associated 'risk' with 'danger' and something that could go either wrong or right. These attributions will influence how the general public responds to communications about uncertain risks. For example, just mentioning the word risk in the statement 'A risk cannot be excluded'commonly used in scientific reports to indicate there is no evidence for riskmay elicit unintentional associations with danger.
The concept of uncertain risk may be a useful catchall term to describe situations in which there is uncertainty about the existence or presence of risk. Current findings suggest that respondents of different groups may have partial, different or conflicting representation of what the concept entails. The ambiguity of the concept of 'uncertain risk' challenges risk communication efforts as much as the epistemic uncertainties in risk assessment that define the concept (Lundgren and McMakin 2013). For example, the statement 'Research is conducted into the risks of nanomaterials in food. Currently, these risks are uncertain.' intends to convey that it is uncertain whether (specific) nanomaterials in food are hazardous to human health. The results of the current study show that policy makers and scientists may understand there is uncertainty about the hazardous propertiesand thus uncertainty about whether a risk exists -, but that the general public may believe that the presence of risk is already determined. They may feel that they are unnecessarily exposed and that their concerns are not being heard, which may complicate a balanced dialogue. While previous research suggests that the general public does not clearly differentiate between specific types of uncertainties in communications (Miles and Frewer 2003;Wiedemann, Sch€ utz, and Thalmann 2008), clarifying the available evidence and remaining uncertainties about the presence of a risk may improve consistent understanding between different societal groups (Wiedemann et al. 2011). When 'uncertain risk' evokes the idea that a person has no control over a risk, this has to be accounted for in risk communication in order to fit an individual's existing knowledge and beliefs (Morgan et al. 2002). Otherwise, different terminology should be used to avoid associations that are unwanted or may have unintended consequences.
Overall, results of this explorative study show that the concept of 'uncertain risk' elicits different associations within and between groups. Contrary to the scientific-technical understanding of the concept, individuals may understand that the presence of risk is determined or feel they have no control over the situation. In risk communication, stating that 'a risk cannot be excluded' is not clear. Explicating uncertainties may improve a mutual understanding and foster a more balanced dialogue between different societal groups. Further (empirical) research is needed to demonstrate if and how explicating evidence and uncertainties in specific cases of uncertain risks improves consistent understanding within and between groups.