By: Castle Williams and Paul Miller
What is “warning fatigue” and when does it occur?
These questions were brought to my attention from a manuscript that I recently published in Weather and Forecasting with several coauthors at the University of Georgia entitled: Maximum Wind Gusts Associated with Human-Reported Nonconvective Wind Events and a Comparison to Current Warning Issuance Criteria. In this manuscript we argue that the wind speed criteria used to issue NWS wind products (i.e. high wind warnings and wind advisories) was much greater than wind speeds where we see damaging impacts occur. After this was published an individual reached out to us and said, “If you lower the threshold for the wind advisory and high wind warning, eventually you will get to where the public has wind warning fatigue.” This comment left myself and a fellow UGA graduate student (Paul Miller) asking, how is warning fatigue defined and is there an optimum warning frequency that maximizes awareness, but minimizes fatigue?
As we set off to answer these questions, we started with terms that have received considerable attention in weather risk communication: “the cry wolf effect” and “the false alarm effect.” The weather enterprise is constantly aware of the number of false alarms that occur with forecasts, especially with severe weather warnings. By definition, a false alarm occurs when a warning was issued, but the event never took place (Barnes et al. 2007). Along the same lines, the cry wolf effect (or false alarm effect) describes the reduced likelihood that an individual will take preventative action following a warning when previous warnings resulted in false alarms (Breznitz 1984). So after thinking long and hard about these definitions, we were struggling to find their connection to the overall frequency of warnings issued. Instead of stopping the investigation, we took a break from these questions and explored other disciplines for answers.
Surprisingly, we found the most answers in the field of cybersecurity, where user-oriented warnings and alerts are common place (Egelman et al. 2008; Sunshine et al. 2009; Felt et al. 2012; Akhawe and Felt, 2013; Anderson et al. 2014), and consequently the effects of alert over-issuance have received considerable attention. In addition to this discipline, other research emerged within the fields of advertising and health care, which shared a common concern with our original interpretation of warning fatigue: does the frequency with which an individual receives a message influence his or her likelihood to take action? The results of these studies point to a delicate balance: too few exposures, and individuals cannot form an attitude about a message; too many exposures, and their attitudes trend increasingly negative. This effect is termed the “inverted U” trend in advertising research (Schmidt and Eisend 2015) with the optimal number of exposures considered the “holy grail of effective frequency” (Tellis 1997, pg. 75). From these disciplines, we discovered three new terms that play a key role in the discussion of warning fatigue: habituation, alert/alarm fatigue, and effective frequency (Table 1). Unlike the terms “cry-wolf effect” and “false alarm effect,” these new terms specifically mention the frequency of alert issuance. Now we were on to something!
After exhausting the literature of various disciplines, we stumbled upon the best resource for warning fatigue: Brenda Mackie’s Ph.D. thesis entitled Warning fatigue: Insights from the Australian bushfire context (2013). This was the training manual for the term warning fatigue. After reading the entire document (which I highly suggest, it’s a great read!), I finally understood why it was incredibly difficult to define warning fatigue. Instead of thinking that this was an entirely new term, that differed from both the “cry-wolf effect” and “the false-alarm effect,” instead I should have been thinking that warning fatigue could be an all-encompassing term. Mackie (2013) clarifies this concept in her thesis by stating the different factors that combine to create “warning fatigue”: “Trust/credibility, over-warning (frequency), false alarms, skepticism, and helplessness are not new factors in public warning response to disaster communication. However, this research demonstrates that they combine in a unique way to produce the phenomenon called ‘warning fatigue.” This conclusion solidified our suspicions, the term “warning fatigue” is indeed different from the “cry-wolf effect” and “the false-alarm effect.”
Now that we have discovered a semantic difference between these terms, what exactly does this mean for the future of the weather enterprise? With several projects redefining the weather forecasting process (FACETS) and many watch, warning, and advisory products in the United States (HazSimp), we are at a revolutionary turning point in the way that weather information is communicated to the many publics. Because of these changes, there is a potential for the alteration of current product issuance thresholds or criteria. These changes would likely impact the number and/or frequency of product issuance, which in turn has the potential to affect the behavioral response of the individuals receiving these alerts. Therefore, future studies should attempt to better understand the effects associated with increasing/decreasing the amount of warnings seen by the general publics. Does warning frequency affect their behavior? If so, then we should ensure that the frequency of product issuance is optimized (i.e., the inverted U trend) to reduce the desensitization of warning messages by the public. However, because “warning fatigue” is made up of many moving parts, it will be difficult to solely isolate the frequency of product issuance. The perception of false alarms must also be considered. By increasing the number of warnings issued, you are also increasing the number of false alarms and/or perceived false alarms by the general publics. Even if they are all true alerts, if our audience perceives the alerts to be false alarms then we must treat them as such.
“Warning fatigue” also can help answer one of the burning questions in the weather enterprise: should we extend weather warning lead time? Back in April 2016 we saw week long warnings for an impending “Tornado Outbreak” that never materialized. That brings up the question: Do long-term lead times (e.g., a week prior to a severe weather outbreak) impact an individual’s willingness to adopt or actively take safety measures during a severe weather event? Additionally, those individuals who received that repeated messaging may be less likely to react for future warnings/outbreaks. More research is needed to better understand recovery time for weather warning messaging. When does an individual become receptive again to these messages? One day? One week? Individual differences and other factors more than likely play a role, but perhaps more information can be discovered in future studies. Lastly, the advertising literature warns that a balance must be kept when issuing messages to individuals. By extending the tornado warning lead time, we are also increasing the number of times that a person sees that message prior to taking action. At a certain point that message will start to have negative effects, which may actually discourage someone from taking action. These are the challenges that the weather enterprise will face in the future, and something that must take into consideration when redefining the way we present weather information. We have made great strides as a community toward perfecting the science, now it is time to work on effectively communicating the science to the publics.
Table 1. Definitions of terms similar to “warning fatigue.”
|Alarm/Alert fatigue||Alarm fatigue occurs when nurses become overwhelmed by the sheer number of alarm signals, which can result in alarm desensitization and in turn, can lead to missed alarms or a delayed response to alarms.||Sendelbach and Funk, 2013|
|Habituation||The diminishing of a physiological or emotional response to a frequently repeated stimulus||Thompson and Spencer, 1966|
|Cry-wolf effect (False-Alarm Effect)||A decline in attention to disaster warnings following a false alarm has been labeled the cry-wolf effect or false-alarm effect.||Atwood and Major, 1998|
|False alarm||A warning was issued, but the even never occurred||Barnes et al., 2007|
|Effective frequency||The precise number of exposures that maximizes consumer response to an advertisement||Tellius, 1997|
|Warning fatigue||People who are exposed to recurring messages about an event which does not eventuate become “tired” of hearing warnings. This includes the factors: Trust/credibility, over-warning (frequency), false alarms, skepticism, and helplessness.||Mackie, 2013|
|Over-warning||People encounter too many warnings in the world, and it is thought that people will be less likely to attend to warnings as a consequence||Wogalter and Leonard, 2005
Castle Williams is 1st year PhD Student at the University of Georgia in the Department of Geography. He holds degrees in Psychology (B.S.), Geography (B.S., M.S.), and Atmospheric Science (B.S.) from the University of Georgia. His overall research interests lie at the intersection of weather and society, but specifically include weather communication, weather messaging, societal impacts of weather, and risk communication.
Akhawe, D., and A.P. Felt, 2013: Alice in warningland: A large-scale field study of browser security warning effectiveness. In Presented as part of the 22nd USENIX Security Symposium (USENIX Security 13) (pp. 257-272).
Anderson, B. B., C.B. Kirwan, J.L. Jenkins, D. Eargle, S. Howard, and A. Vance, 2015: How polymorphic warnings reduce habituation in the brain: Insights from an fMRI study. In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems (pp. 2883-2892). ACM.
Barnes, L. R., E.C. Gruntfest, M.H. Hayden, D.M. Schultz, and C. Benight, 2007: False alarms and close calls: A conceptual model of warning accuracy. Weather and Forecasting, 22(5), 1140-1147.
Breznitz, S. and C. Wolf, 1984: The psychology of false alarms.
Egelman, S., L.F. Cranor, and J. Hong, 2008: You’ve been warned: An empirical study of the effectiveness of web browser phishing warnings. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 1065-1074). ACM.
National Severe Storms Laboratory, 2016: Forecasting a continuum of environmental threats. [Available online at: http://www.nssl.noaa.gov/projects/facets/%5D.
Felt, A. P., E. Ha, S. Egelman, A. Haney, E. Chin, and D. Wagner, 2012: Android permissions: User attention, comprehension, and behavior. In Proceedings of the Eighth Symposium on Usable Privacy and Security (p. 3). ACM.
National Oceanic and Atmospheric Administration, 2016: NWS hazard simplification project: Simplifying and clarifying the message. [Available online at: http://www.nws.noaa.gov/com/weatherreadynation/files/HazardSimplification.pdf%5D.
Mackie, B, 2014: Warning fatigue: Insights from the Australian bushfire context. Thesis.
Miller, P. W., A.W. Black, C.A. Williams, and J.A Knox, 2016: Maximum wind gusts associated with human-reported nonconvective wind events and a comparison to current warning issuance criteria. Weather and Forecasting, 31(2), 451-465.
Schmidt, S., and M. Eisend, 2015: Advertising repetition: A meta-analysis on effective frequency in advertising. Journal of Advertising, 44(4), 415-428.
Sunshine, J., S. Egelman, H. Almuhimedi, N. Atri, and L.F Cranor, 2009: Crying wolf: An empirical study of SSL warning effectiveness. In USENIX Security Symposium (pp. 399-416).
Tellis, G. J., 1997: Effective frequency: One exposure or three factors? Journal of Advertising Research, 75-80. [Available online at: http://www-bcf.usc.edu/~tellis/Effective%20Frequency.pdf%5D.