Using Narrative to Explain Uncertainty in Climate Change
By Grace Freeman1, Alina Rousseau1, Michelle Brunton1, Luke Kramer1, Stephanie Miller2, and Laura Kate Corlew1
1. University of Maine at Augusta, Augusta, Maine 04430
2. University of Maine, Orono, Maine 04469
Abstract
This study presents the findings on the use of narrative (i.e., storytelling) when communicating complex scientific uncertainties. A growing body of evidence in narrative cognition and communication has shown promise for the use of narrative in science communication. Prior research has shown uncertainty is difficult to communicate and raises ethical concerns since scientists do not want audiences to be either underconfident in projections with epistemic uncertainty (i.e., where unknowns remain) or overconfident in projections with aleatory uncertainty (i.e., based on random factors). This project researched the use of narrative in communicating the scientific uncertainty of data projection tools built to support agricultural and conservation decision makers with decisions related to climate change. Participants completed an empirical survey reviewing the underpinnings of four data projection tools. For each tool, participants were randomly assigned to read either a technical or narrative description of the tool. Results indicate that narrative descriptions increased understanding, though measurements for emotional response and behavior change were non-significant. Results additionally indicate that the use of narrative did not reduce participants’ confidence in projections of epistemic uncertainty. However, participants did show a statistically significant overconfidence in the models’ ability to predict completely random factors, indicating that caution should be applied and further research conducted on the use of narrative in communicating scientific uncertainty.
Keywords: communication, climate, description, narrative, uncertainty, epistemic, aleatory
Climate change and data scientists have an imminent need to identify and implement effective communication strategies for their technical work. This need is imperative as agricultural and conservation decision-makers need evidence to inform their decisions. Previous research has been promising for the use of narrative communication to facilitate the understanding of complex science. This study expands upon that research by exploring whether narrative communication is an effective tool for communicating uncertainty data, since decision-makers may hesitate to act if they misunderstand or lose confidence in projection models.
The BARRACUDA Project (Biodiversity And RuRal Adaptation to Climate change Using Data Analysis) is an interdisciplinary, multi-state, and multi-university research project in Northern New England seeking to create high-quality data models and tools to support agriculture and conservation sectors regarding climate change decision-making (RII Track-2 FEC; NSF Award #2019470). Within Barracuda, Team CHASM (Communication Has A Special Meaning) is a mentored undergraduate and graduate research team at the University of Maine at Augusta (UMA) seeking to gain a better understanding of effective climate change communication strategies. This study uses data projection tools created within the Barracuda Project as examples of uncertain information and empirically test participants’ understanding, confidence, emotional response, and self-reported behavior change according to the use of either narrative or technical descriptions.
Literature Review
Communication fuels scientific dissemination by conveying sentiment, knowledge, evidence, and value (Filiz 2020; Hansson et al. 2020). This paper focuses on two communication styles: technical communication, which has been shown to limit comprehension among non-scientists (Bromme, Rainer, and Jucks 2018; Schwingel 2018), and narrative communication, which is associated with increased recall, ease of understanding and comprehension, and shorter reading time (Dahlstrom 2014). Bullock et al. (2019) illustrate how the use of jargon can hinder the processing of scientific information and amplify resistance to the message. Narrative formats emotionally transport audiences, making scientific concepts more relatable and actionable (Appel et al. 2015; Downs 2014). Thus, narratives can influence behavior (Kim et al. 2012).
However, to our knowledge there are limited examples of narratives used to communicate uncertainty-inherent projections related to climate change. Audiences frequently misinterpret uncertainty, conflating variability with ignorance (Rydmark, Kuylenstierna, and Tehler 2020). This miscommunication can cause decisionmakers to overestimate uncertainty, therefore delaying actions (Horne, De Urioste-Stone, and Daigle 2021) or misallocating resources (Ward et al. 2019). Distinguishing epistemic uncertainty (i.e., known evidence with remaining gaps) from aleatory uncertainty (i.e., intrinsic randomness) is essential (van der Bles 2019). The application of narrative science communication in this context offers a potential resolution. Narrative science communication may help the general population more fully grasp the nuances of uncertainty to make decisions in their lives.
Methodology
Participants in this study were adult university students in Psychology and Communications courses, aged 18 to 65 (n = 81). Sixty‑five participants completed the demographic questionnaire. Of these, 69 % were female; 84 % identified as White/Caucasian, 6 % Native, 5 % Hispanic, 2 % Black, and 2 % multiracial. Informed consent was obtained; the survey blocked anyone under 18. This was intended as a pilot study; hence, the population comprises college students rather than agricultural stakeholders.
The survey was developed and hosted within Qualtrics (Qualtrics, Provo, UT). Participants were provided images from four data projection tools developed by the greater Barracuda Project research team. These data projection tools covered the following topics: Spotted-Wing Drosophila, the Soybean Crop Model, Crop Switching, and Data Visualization (Barracuda 2023). Participants were randomly assigned to read either a narrative or a technical description of the tool.
Participants randomly assigned to the narrative read a comedic short story about a hapless farmer whose neighbor helps him understand why he would be better off growing soybeans as opposed to pineapples based on the current and projected growing season changes. Participants randomly assigned to the technical description read the same information about the model inputs and outputs, though in a technical and data rich format.
Predictive models have inherent uncertainty. The epistemic uncertainty stems from the inability of any model to encompass the infinite variables that can affect the target outcome; decisions must always be made as to which variables to include or exclude in the simplified models. The narrative and technical descriptions include information about the specific data included in the model and describe the prediction outputs. The aleatory uncertainty stems from many other unpredictable factors that can also affect crop outcomes, such as deer infestations, wildfires, or bad seed stock. These factors are essentially random events that cannot be predicted in a crop yield model.
After reviewing the uncertainty information, all participants were given the same questions to measure their understanding, regardless of whether they received the information in the narrative or technical format. Participants were then asked additional questions designed to measure their emotional response/transportation (immersion into a narrative) and to evaluate their judgments of confidence in the uncertainty information presented. Each of these question sets were developed by the team according to each specific data projection tool. The emotional response/transportation questions were adapted from a short form transportation scale (Appel et al. 2015) which measures self-reported emotional, imaginative, and cognitive engagement with the descriptions. The content-based questions were presented as true/false to check respondents’ understanding of the material and Likert-scaled questions were used to assess the confidence level of respondents after they completed the readings.
Analysis of the data collected from this survey was conducted through Qualtrics and PSPP software (GNU Project 2007). The survey experienced significant attrition, such that later portions of the survey had lower completion rates. It was checked if there were any significant differences between scenarios, but attrition greatly inhibited the data sets. Therefore, the team combined all scenarios to create a more robust sample size when analyzing the difference in outcomes between narrative and technical reporting.
The overall aim of this study was to discover if narrative descriptions were more effective than technical descriptions at communicating uncertain scientific data. It was hypothesized that participants receiving the narrative versus technical description would 1) better understand, 2) have a stronger emotional response, and 3) a stronger likelihood of belief or behavior change, without 4) undermining their confidence in the uncertain data.
Results
Participant Comprehension
Regarding hypothesis 1 (that participants will better understand the uncertain scientific data when communicated in narrative versus technical format), the one-way ANOVA revealed a modest but significantly better understanding of the science when participants read a narrative reporting formats (n = 153) versus technical reporting formats (n = 150) (F(1, 301) = 6.22, p = .013*). Each data projection tool scenario had three questions that were checked for understanding of the reported science, leading to a correct understanding score between 0 (no correct answers) and 3 (all correct answers). Participants who read the narrative scenarios scored more 2s and 3s, showing increased understanding of the information (Table 1).
Table 1. Number of correct answers for narrative versus technical descriptions. Row % shows the distribution of the number of correct answers within each specific group. Column % shows the composition of each number of correct answers category across the two groups. Total % shows the distribution of every cell relative to the entire dataset.
Additionally, when participants were asked if the descriptions they read were easy or hard to understand, participants reported perceiving that the narrative explanations were easier to understand (Table 2). A Chi Square analysis revealed a significant difference in the perception of ease versus difficulty of the readings between participants who read narrative versus technical descriptions (χ² (df = 1, n = 258) = 9.84, p = .002**).
Table 2. Participant perception of description difficulty. Row % shows the split between “Easy” and “Hard” ratings within each group. Column % shows the breakdown of groups within each difficulty rating. Total % shows the percentage of the entire sample that falls into each specific cell.
Participants’ Emotional Responses
Hypothesis 2 (that participants would have a stronger emotional response to narrative versus technical writing) found no significant differences (F(1, 209) = 2.46, p = .118). These results indicate that participants who read the narrative descriptions were no more emotionally transported (i.e., invested and immersed) than those who read the technical descriptions.
Participants’ Likelihood of Belief/Behavior Change
Hypothesis 3 (that participants would show a stronger likelihood of belief or behavior change) was measured by asking participants two yes or no questions: “Do you perceive this type of data projection tool would be useful to you or someone you know?” and “If you had access to this tool would you use it to inform your decision-making?” A Chi Square analysis on perceived usefulness of each respective data projection tool yielded significant results, χ² (df = 1, n = 277) = 15.06, p = .000** (Table 3), such that participants receiving the narrative descriptions found the tools to be more useful than those reading the technical descriptions.
Table 3. Perceived usefulness of data uncertainty tools. Row % shows the split between “Yes” and “No” responses within each group. Column % shows the breakdown of groups within each opinion category. Total % shows the percentage of the entire sample that falls into each specific cell.
However, believing that the tools would be useful did not lead to participants reporting a likeliness to utilize the tools. A Chi Square analysis did not reveal evidence of perceived likelihood of behavior change to use the tools χ² (df = 1, n = 282) = 1.82, p = .178 (Table 4).
Table 4. Perceived likelihood of behavior changes to use tools. Row % shows the split between “Yes” and “No” responses within each group. Column % shows the breakdown of groups within each opinion category. Total % shows the percentage of the entire sample that falls into each specific cell.
Participants’ Confidence in Uncertain Data
Hypothesis 4 (that the use of narrative descriptions would not undermine participants’ confidence in uncertain data) was this study’s new application of the use of narrative in scientific communication. The survey included confidence questions for both epistemic and aleatory uncertainty related to each data projection tool. A check in confidence for the Soybean Crop Model despite epistemic uncertainty asked, “How confident are you that the model can predict harvestable biomass of soybeans under ideal conditions?” We hypothesized that the narrative descriptions would not undermine participants’ confidence in the scientific data. Indeed, a Chi Square analysis revealed no evidence of a difference in confidence level when participants read the narrative versus technical descriptions, χ² (df = 4, n = 258) = 6.16, p = .187 (Table 5).
Table 5. Participants reported confidence on questions of epistemic uncertainty. Row % shows the level of confidence responses within each group. Column % shows the breakdown of groups within each category. Total % shows the percentage of the entire sample that falls into each specific cell.
While the data indicated that narrative descriptions of science did not undermine the confidence of participants in uncertain data of an epistemic nature, the same was unfortunately true for uncertain data of an aleatory nature. The check for confidence (or more specifically, overconfidence) in the Soybean Crop Model’s ability to predict unpredictable outcomes asked, “How confident are you that the model can predict harvestable biomass of soybeans under unforeseen circumstances like deer attacking the crop?” As with all the data projection tool scenario descriptions, the Soybean Crop Model’s narrative and technical descriptions both included explicit information about what data was used in the model to create the projections, in this case the temperature and solar radiation of the plot each day, as well as the soil moisture content. Deer infestations would therefore be outside of the scope of the model.
Participants who read the technical description were more likely to recognize that the models could not account for random factors. Participants who read the narrative descriptions were less likely to recognize that random variables were beyond the scope of the science, χ² (df = 4, n = 231) = 10.70, p = .03* (See Table 6). This finding is of concern because it indicates that while audience confidence in uncertainty data may not be undermined by the use of narrative scientific description, the audience’s complex understanding of the scope and implications of the uncertainties may be adversely affected.
Table 6. Participants reported confidence on questions of aleatory uncertainty. Row % shows the level of confidence responses within each group. Column % shows the breakdown of groups within each category. Total % shows the percentage of the entire sample that falls into each specific cell.
Discussion
The study results indicate that narrative descriptions of scientific uncertainty data can facilitate better comprehension of information. The increased comprehension validates previous research (Dahlstrom 2014; Downs 2014) that narratives can be used to promote engagement with a wider audience, increasing the accessibility of science to the general population.
The study found no significant difference in participants’ emotional transportation (i.e., emotional investment). Participants who read the narrative descriptions were no more and no less transported than those who read the technical descriptions. While narratives do have an established capacity for transporting audiences to the extent that they are wholly absorbed in the story (Appel et al. 2015), the narratives used in this study did not establish that effect among participants. Future research can explore the impact of narratives that resonate more and less with participants’ personal experiences, values, or concerns.
The study did not replicate the increases in behavior change associated with the use of narrative descriptions that have been found in the literature (Downs 2014; Kim et al. 2012). While participants who read the narrative descriptions found the data projection tools to be more useful than those who read the technical descriptions, they did not indicate that they would change their behavior to make use of the tool. The lack of reported behavior change may be due to our participant pool being composed of college students that includes both individuals who are likely to use the data projection tool in their careers and those who are unlikely to have the need. Alternatively, the empirical survey format we used may not be able to capture the increase in behavior change found in the literature (Kim et al. 2012). Future research could particularly focus on populations who are more likely to have use for the tool in addition to exploring the use of more and less transporting narratives and adopting measures that more comprehensively capture behavior change.
This study explored how the use of narrative in science communication might impact an audience’s understanding of and confidence in uncertain projections. Scientific uncertainty has been shown to be a specific area of struggle for audiences (Horne, De Urioste-Stone, and Daigle 2021; Ward et al. 2019). Epistemic uncertainty needs to be better understood so the public can have confidence in robust projections. This is because understanding epistemic uncertainty may help people see why a projection looks the way it does and how it can become more trustworthy over time. However, each predictive model adds to a body of evidence that need not be discounted in its entirety due to ongoing epistemic uncertainty. The results indicate that audience confidence in epistemic uncertainty data was not impacted using narrative descriptions. This is promising for advances in science communication.
The results also included that participants who read narrative descriptions were less likely to recognize the limitations that the data projection models possess in predictions concerning random factors beyond the model scope. While there was little difference between participants reporting they were confident or very confident in the models’ ability to predict truly random factors, participants who read the technical descriptions were more likely to be unconfident and very unconfident than those who read narrative descriptions. It is possible that this disparity resulted from narrative descriptions invoking a sense of fictionality in participants’ understanding, such that the possibilities and limitations of science were perceived to change. The very use of narrative may lead audiences to suspend disbelief and entertain applications inconsistent with reality (e.g., a model predicting yield in ideal growing conditions is able to predict a deer infestation). As such, future research should explore whether suspension of disbelief can impact an audience’s cognitive adherence to scientific laws in the context of fictional or non-fictional scientific narratives
Conclusion
This study examined whether narrative communication could alleviate the difficulty science communicators face when presenting uncertain projection models. We hypothesized that, compared with a technical description, a narrative format would (1) improve comprehension, (2) increase emotional transportation, (3) raise belief or behavioral intention, and (4) not reduce confidence in the uncertain data. We found that narrative descriptions improved participants’ understanding of scientific information. In the same vein, participants also perceived narratives to be easier to understand. However, our data does not provide clear support that narrative communication can be successfully applied to the communication of uncertainty data. Participants’ confidence in uncertain data did not lessen with narrative descriptions, though participants did show an overconfidence in predictions beyond the scope of the models. Similarly, our data does not support the idea that participants will alter their behavior, nor will they have a strong emotional response when presented with narratives about data projection tools. Considering these findings, using narrative communication styles should be approached with an awareness of its potential pitfalls in explaining uncertainty. While we could not prove that the narrative communication style was abundantly superior across all hypotheses, effective communication in the climate and agricultural conservation field should be seen as imperative for bolstering audience engagement. Using caution, and with additional research as to its appropriate use with uncertain data, narrative science communication can serve as an instrument to foster scientific literacy in broader communities.
Declarations
Author Contribution Statements
All authors contributed to the study conception and design, including material preparation, data collection, analysis, and commented on previous versions of the manuscript. All authors read and approved of the final manuscript.
K.C., G.F., and A.R. wrote the main manuscript text.
K.C. provided academic support, professional advising, manuscript writing, editing and data analysis.
S. M. provided academic support and developed technical descriptions.
L.K., M.B., G.F., A.R., and K.C. developed narrative descriptions.
G.F. performed data analysis of the survey results, wrote the literature review, results and conclusion portions of the manuscript, and assisted with edits to the main manuscript.
A.R. wrote the methodology section, drafted the manuscript, and created an Institutional Review Board (IRB) application for ethics board approval of study, and assisted with edits to the main manuscript.
G. F, M.B. and L.K. performed the literature review for this study.
M.B. assisted with drafting and developing the survey questions.
L.K. assisted with statistical data analysis of survey results, and the digital creation of the survey in Qualtrics, and drafting the literature review portion of the manuscript.
Ethics approval
The questionnaire and methodology for this study was approved by the Institutional Review Board committee of the University of Maine at Augusta (IR 30195) which recommend this proposal be exempted from further review pursuant to 45 CFR 46.104(d) (2) “[r]esearch only including interactions involving educational tests, survey procedures, interview procedures, or observation of public behavior”.
Funding
This work was supported by a National Science Foundation Grant (RII Track-2 FEC; NSF Award #2019470).
Competing Interest
The authors have no relevant financial or non-financial interests to disclose.
Consent to participate
Informed consent was obtained from all individual participants included in the study.
Data Availability
Data is available to view at: https://www.openicpsr.org/openicpsr/workspace?goToPath=/openicpsr/228762&goToLevel=project
References
Appel, Markus, Timo Gnambs, Tobias Richter, and Melanie C Green. 2015. “The Transportation Scale-Short Form (TS-SF).” Media Psychology (Philadelphia) 18 (2): 243–66. https://doi.org/10.1080/15213269.2014.987400.
Barracuda – Biodiversity and Rural Response to Climate Change Using Data Analysis, 2023. URL https://biobarracuda.org/ (accessed 7.8.25).
Bromme, Rainer, and Regina Jucks. 2018. “Discourse and Expertise: The Challenge of Mutual Understanding between Experts and Laypeople 1.” In The Routledge Handbook of Discourse Processes, 2nd ed., edited by David N. Rapp, Michael F. Schober, and M. Anne Britt. Routledge. https://doi.org/10.4324/9781315687384-13.
Bullock, Olivia M., Daniel Colón Amill, Hillary C. Shulman, and Graham N. Dixon. 2019. “Jargon as a barrier to effective science communication: Evidence from metacognition.” Public Understanding of Science 28, no. 7: 845-853. https://doi.org/10.1177/0963662519865687
Code of Federal Regulations. 2025. Retrieved from Title 45 Subtitle A Subchapter A Part 46 Subpart A § 46.104: https://www.ecfr.gov/current/title-45/subtitle-A/subchapter-A/part-46/subpart-A/section-46.104
Dahlstrom, Michael F. 2014. “Using Narratives and Storytelling to Communicate Science with Nonexpert Audiences.” Proceedings of the National Academy of Sciences – PNAS (United States) 111 (Supplement 4): 13614–20. https://doi.org/10.1073/pnas.1320645111.
Downs, Julie S. 2014. “Prescriptive Scientific Narratives for Communicating Usable Science.” Proceedings of the National Academy of Sciences – PNAS (United States) 111 (Supplement 4): 13627–33. https://doi.org/10.1073/pnas.1317502111.
Filiz, Bijen. 2020. “The Relationship between Effective Communication Skills and Verbal Intelligence Levels of Faculty of Sport Sciences Students.” International Journal of Educational Methodology 6, no. 3: 603-612. https://doi.org/10.12973/ijem.6.3.603
GNU Project. 2007. GNU PSPP (Version 3) [Computer Software]. Free Software Foundation. Boston, MA. Available from: https://www.gnu.org/software/pspp/”
Hansson, Sten, Kati Orru, Andra Siibak, et al. 2020. “Communication-Related Vulnerability to Disasters: A Heuristic Framework.” International Journal of Disaster Risk Reduction 51: 101931. https://doi.org/10.1016/j.ijdrr.2020.101931.
Horne, Lydia, Sandra De Urioste-Stone, and John Daigle. 2021. “Climate Change Adaptation and Mitigation in the Face of Local Uncertainty: A Phenomenological Study.” Northeastern Naturalist 28 (sp11): 108–28. https://doi.org/10.1656/045.028.s1107.
Kim, Hyun Suk, Cabral A. Bigman, Amy E. Leader, Caryn Lerman, and Joseph N. Cappella. 2012. “Narrative health communication and behavior change: The influence of exemplars in the news on intention to quit smoking.” Journal of Communication 62, no. 3: 473-492. https://doi.org/10.1111/J.1460-2466.2012.01644.X
Rydmark, Joacim, Jan Kuylenstierna, and Henrik Tehler. 2021. “Communicating Uncertainty in Risk Descriptions: The Consequences of Presenting Imprecise Probabilities in Time Critical Decision-Making Situations.” Journal of Risk Research (Abingdon) 24 (5): 629–44. https://doi.org/10.1080/13669877.2020.1801807.
Schwingel, Johanna M. 2018. “Enhancing Scientific Communication Through an Undergraduate Biology and Journalism Partnership.” Journal of Microbiology & Biology Education (United States) 19 (1). https://doi.org/10.1128/jmbe.v19i1.1445.
van der Bles, Anne Marthe, Sander van der Linden, Alexandra L. J Freeman, et al. 2019. “Communicating Uncertainty about Facts, Numbers and Science.” Royal Society Open Science (England) 6 (5): 181870. https://doi.org/10.1098/rsos.181870.
Ward, Alastair I, Suzanne Richardson, Roy Macarthur, and Aileen C Mill. 2020. “Using and Communicating Uncertainty for the Effective Control of Invasive Non‐native Species.” Mammal Review (Oxford) 50 (2): 211–20. https://doi.org/10.1111/mam.12188.