1. 1 Social Science Research Findings and Evidence Based Applications for Practice (Rev 13)

2.  Dennis S. Mileti  Professor Emeritus  University of Colorado at Boulder &  START Center, University of Maryland, College Park  Erica Kuligowski  Research Associate  University of Colorado at Boulder &  START Center, University of Maryland, College Park 2

3.  Contributors:  John H. Sorensen, Oak Ridge Nat’l Lab  Barbara Vogt-Sorensen, Oak Ridge Nat’l Lab  Reviewers:  Susan Cutter, University of South Carolina  David Gillespie, Washington University  Kathleen Tierney, University of Colorado 3

4.  Supported by:  U.S. Department of Homeland Security, Grant Number N00140510629 to the START Center, University of Maryland at College Park  However:  Opinions, findings & conclusions are the author’s and do not necessarily reflect the views of the U.S. Department of Homeland Security 4

5.  SYNTHESIZE:  Findings from social science public warning research for hazards  Do it in plain language  PRESENT:  Evidence-based applications for practice  Regarding the…. 5

6. How do you Help People in Danger to: -STOP….. -HEAR…. & -TAKE PROTECTIVE ACTIONS FOR…. 6

7. 7

8. 8

9. 9

10. 10

11. 11

12. 12

13. 13

14. 14

15. 15

16.  50+ Years “Warning Response” Research  Emphasis on U.S. publics  Researched Hazards Include:  Natural: Hurricane Camille, Mt. St. Helens  Terrorism: World Trade Center 1993 & 9/11  Hazardous Materials: Mississauga, Nanticoke  Technology: Three Mile Island  Building Fires: MGM Grand, Cook County Hospital  We Know:  What works, what doesn’t & why, & how to apply it 16

17.  350 Page Annotated Bibliography (1 page/publication with key findings) is Available at:  http://www.colorado.edu/hazards/publications/informer/infrmr2/pubh azbibann.pdf 17

18.  Mileti, Dennis S., and John H. Sorensen. 1990. Communication of Emergency Public Warnings: A Social Science Perspective and State-of-the-Art Assessment. Report #ORNL-6609. Oak Ridge, TN: Report #ORNL-6609 for the Federal Emergency Management Agency. (Available on line as a pdf file)  Can be found at: http://emc.ornl.gov/EMCWeb/EMC/PDF/Communication Final.pdf 18

19.  150 Entry Bibliography is Available at:  http://www.colorado.edu/hazards/library/BuildingsEvacBib2007.doc 19

20.  PEOPLE STAY PEOPLE:  Across hazards  AND THE SAME FACTORS:  Determine public behavior across hazards & events  Mathematically modeled & same equations apply  BUT OUTCOMES VARY ACROSS EVENTS:  Because of different quantities for the factors that determine public response occur in different events, e.g., warning message quality, gender & age 20

21.  1. Warning System Definition  2. Warning System Preparedness  3 Myths  4 Alert  5. Diffusion  6. Mobilization  7. Notification, Messaging & Public Response  8. Compliance  9. Monitoring, Evaluation, & Feedback  10. Destinations  11. Modernizing Warning Systems  12. Next Steps 21

22.  Warnings Systems Mean:  Different things to different people  All Warning System Have:  The same subsystems  Integrating the Subsystems:  Reduces warning system failures 22

23.  SUBSYSTEM 1 = Risk Environment:  Natural, technological, civil  SUBSYSTEM 2 = Detection:  Monitoring, detection, data assessment, data analysis, prediction, & informing  SUBSYSTEM 3 = Management:  Interpretation, decision to warn, method of warning, content of warning, channel of warning, monitoring response for feedback & warning revision  SUBSYSTEM 4 = Public Response:  Interpretation, confirmation, response, & warning others 23

24. 24 Risk Environment Subsystem Detection Subsystem Management Subsystem Public Response Subsystem WARNING Cues Monitor Inform Monitor Inform & Interpret

25.  Warning System Preparedness:  Elaborate the subsystems  Develop linkages between them  Major Goals:  The rarely used system will work when needed  Weave together organizations, agencies, levels of government & disciplines that rarely interact  No communication link in the chain breaks 25

26.  Ready the Warning Players:  Warning plans, operating procedures & protocols  Warning training & exercises  Warning human factors issues addressed  Warning standards of performance  Evidence based:  Warning messages & dissemination channels  Prime the Audience:  Public warning & response education  “Grow” the Warning System Rationally:  Development priorities based on risk assessment 26

27.  A MYTH EXISTS WHEN PEOPLE:  BELIEVE its true (but its not)  Think they have EVIDENCE for it (but they don’t)  WON’T STOP BELIEVING it (no matter what)  Here Are Three of Them….. 27

28.  Non-problem:  Never occurred after a warning  Actual Problem:  “We didn’t issue a warning so we wouldn’t cause a panic”  Panic Occurs When:  Spaces are confined  Escape routes ARE available, but  People think: not enough time for everyone to use them, resulting in  People must: “compete to live”  Even then, Panic is Rare 28

29.  Definition:  “Keep it simple stupid”  Myth:  Applies to public warning messages  Reality:  Applies to advertising, not public warnings  Warned people become “information starved”  If those who warn don’t say enough, the public will find it elsewhere & confusion can result 29

30.  Myth:  People don’t respond after false alarms  Reality:  They do (perhaps differently)  False Alarms:  Can be productive for future response “if explained”  Anger local governments because they cost money  Exception:  People ignore sirens, especially if sounded frequently, e.g., for siren tests 30

31. STOP Ongoing Life Get People’s ATTENTION CAPTURE Your Audience But Keep in Mind that…. 31

32.  People:  Don’t remember meaning of:  Siren signals (wails, whoops, tones)  Color codes  Don’t distinguish between:  Advisories, watches & warnings  Except:  When signals/codes are “drilled into them”, e.g., weekly fire drills in schools 32

33.  Many Isolate “Themselves” from Information  Some are isolated by circumstance, e.g., poor  And Even when Signals Blare, Many:  Think they’re “safe” &  Disasters happen to other people  Some Sub-populations Need Unique Alerts, e.g.,  Hospitals in communities  Hearing impaired in buildings  Visitors & “out-of-towners”  Different language speakers 33

34.  Get People’s Attention, e.g.,  “Lights on” in theaters  Piercing sounds with TV crawlers  Wake People Up, e.g.,  Sleeping children & older adults  Hearing loss & under the influence  Outside Devices Loose Effectiveness if:  Windows shut & air/heat is on  3 minute sounding 10 decibels over ambient outdoor siren has a 62% chance of waking someone up  Need Indoor Devices for Alert at Night:  Fast moving community event  Fire in a hotel 34

35.  Warning Diffusion “Among those Warned”  Always happens, count on it, & use it  9/11 Example:  Most in country learned about attack in 1 hour  Many in towers found out a plane hit from friends/relatives  Rule of Thumb:  1 informal first warning for every 2 formal first warnings  Informal Alerting Increasing with New Technologies 35

36.  Diffusion = Getting the Word Out  A social process regardless of technology used  No “SILVER BULLET” Technology:  Different technologies = different effectiveness  USE ALL OF THEM (relying on one won’t work)  Reach sub-populations in different ways:  And using diverse technologies (channels) helps “confirm” the message which facilitates human response  Effectiveness impacted by time of day/night 36

37. 37 0.1

38.  Time between 1 st Warning & Starting a Protective Action  P eople don’t all act at once  Getting ready to respond delays response  People Delay in Order to:  Locate family & gather possessions  Confirm warning & need to take action  Talk things over with others  And a Few People Don’t Respond at All 38

39.  Can Vary By:  Urgency of event  Severity of threat  Time of day/night  Time increases as message quality decreases  Non-linear (curved) Relationship between Time & Starting a Protective Action:  Typically an “S” shaped relationship  Here’s an example.... 39

40. 40

41. 41

42.  Predictions Based on SCIENCE Work:  “A” causes “B”  Predictions Based on NON-SCIENCE Don’t:  What people did in past events:  Using a past “B” to predict a future “B”  Behavioral intention surveys:  Intentions (opinions) & behavior (actions) are different  Key determinants of public warning response don’t operate in pre-event surveys & aren’t known by respondents 42

43.  Many “Statistically Significant” Factors Documented by Research but….  They Vary in Importance:  Strong vs. weak relationships  Real vs. spurious effects  High vs. low research evidence  Strong Evidence Exists for What Follows 43

44. “About the Warning Message” 44

45.  Five Dimensions:  Channel  Frequency  Content  Style  Source 45

46.  Number of Communication Channels:  The “more the better”  Type of Communication Channel:  Personal channels work best  The “more the better”  Communication Frequency:  The “more” its repeated & heard the better:  Repetition fosters confirmation  Confirmation fosters belief  Belief fosters taking action 46

47.  CONTENT (What to Say):  WHAT: Tell them what to do  WHEN: Tell them when (time) to do it  WHERE: Say who should & shouldn’t do it  WHY: Tell about the hazard’s consequences  WHO: Say who‘s talking (source):  There is NO single credible source, so use multiple sources for the same message 47

48.  STYLE (How to Say It):  CLEAR: Simply worded is best  SPECIFIC: Precise & non-ambiguous  ACCURATE: Errors cause problems  CERTAIN: Authoritative and confident  CONSISTENT:  Externally: Explain changes from past messages & differences from what others are saying  Internally: Never say “attack will occur soon, don’t worry” 48

49.  Social Cues Help:  “Monkey see, monkey do”  What neighbors, friends, & relatives are doing  What organizations are doing  Physical Cues Help too:  If confirm the risk (rain in flood warnings) 49

50. 50

51. “About the Audience” 51

52.  Everyone May “Receive” the Same Warning Message, but:  Differences in the people who hear it result in it “meaning” different things to different people  Overcoming These Human Receiver “Biases” is Possible, but Requires:  Well-crafted warning messages  Well-designed warning delivery systems  The Human Filter Includes…. 52

53.  Socio-economic Status:  Having little money, education, employment  Age:  Being young or old  Gender:  Being male  Ethnicity:  Being non-Anglo  Acculturation:  Not speaking English, born in another country 53

54.  Roles of Responsibility for Others:  Having children  Larger family size  Having pets  More kin relationships  Family united  Greater community involvement 54

55.  People “Normalize” Warning Information Based on Their Experience:  Think disasters faced will be like those experienced  Inclined to do what was appropriate in past events experienced 55

56.  How Message & People Factors Interact 56

57.  There is NO Single Credible Spokesperson:  STOP LOOKING FOR ONE  Why? People have different ideas about who’s credible  “Who’s Credible?” = Wrong Question:  Many “think” spokesperson credibility = message belief  They’re different & belief is what’s important  How to Achieve Warning Belief:  ONE MESSAGE OVER DIVERSE CHANNELS  FROM A “PANEL” OF SPOKESPERSONS :  e.g., officials, Red Cross, scientists, familiar newscaster, & others  REPEATED MULTIPLE TIMES  Here’s as Good as Single Spokespersons Get…. 57

58. 58

59.  Multi-faceted Concept Including:  PAST: What people “import” into the event  PRESENT: What people “think” based on the information/cues they get during the event  NATURAL INCLINATION: “I’m safe, don’t tell me I’m not”  Not Static & Can Change  Manage it in Warning Messages:  Provide warning information that “overcomes” differences in people’s past, present, & natural inclinations 59

60.  About Perceived Risk “During the Event”:  Different from “pre-event” risk perception  Roadblock to Taking Action:  People don’t “perceive they’re at risk”  People “perceive that they’re safe”:  And search for information to confirm that they are, and that’s what they’ll believe if they find it, not the warning  Moreover:  People dichotomize risk into:  Do something/do nothing  They don’t act in proportion to probability estimates  And they’re inclined to:  “Normalize” the risk information they receive 60

61.  Milling/Confirmation:  The KEY to how warnings work  Few People Do Something because They’re Told to:  People need to think it’s their own idea  People Think It’s Their Idea & then Act from:  MILLING AROUND: talking about it with others and confirming the risk and what they could do about it  Before Taking Protective Actions, People Need to:  Have confirmation (additional information)  Talk it over with others 61

62. 62

63.  Perceived Risk Determined by:  Multiple communications  Multiple channels  Milling Determined by:  Multiple communications:  Multiple channels  Plus Perceived risk  Warning Response Behavior Determined by:  Multiple communications  Multiple channels  Perceived risk  Plus Milling 63

64. 64 STATUS ROLES EXPERI- ENCE CUES INFO RECEIVED MILLING KNOW- LEDGE PERCEIVED RISK ACTION INFO BELIEF

65. 65 STATUS ROLES EXPERI- ENCE CUES INFO RECEIVED MILLING KNOW- LEDGE PERCEIVED RISK ACTION IINFO BELIEF

66.  Models Are Represented by Equations:  Called “simultaneous multiple regression equations”  Equations Enable Us to Determine:  Effect of every factor on other factors while controlling for the effects of everything else (good science)  The Result is:  Can distinguish between what’s really important & what isn’t  When to Get Excited:  When different studies reach the same conclusions  That’s where we are with research on public response to warnings for hazardous events 66

67. X4 = β 41X1 + β 42X2 + β 43X3 + e4 X5 = β 51X1 + β 52X2 + β 53X3 + β 54X4 + e5 X6 = β 61X1 + β 62X2 + β 63X3 + β 64X4 + β 65X5 + e6 X7 = β 71X1 + β 72X2 + β 73X3 + β 74X4 + β 75X5 + β 76X6 + e7 *Averill, J. D., D.S. Mileti, R.D. Peacock, E.D. Kuligowski, N. Groner, G. Proulx, P.A. Reneke, and H.E. Nelson. 2005. Federal Building and Fire Safety Investigation of the World Trade Center Disaster: Occupant Behavior, Egress, and Emergency Communications. Report NCSTAR 1-7, National Institute of Standards and Technology, Gaithersburg, MD. Available at: http://wtc.nist.gov/NISTNCSTAR1-7.pdf 67

68.  All Factors AREN’T EQUAL  Some Factors are REALLY Important:  CONTENT: what the message says:  Especially what actions to take  REPETITION: Hearing same warning many times  CUES: Seeing things that confirm the message  MILLING: Confirming it with others  Other Factors are LESS Important:  Demographics (unless information is poor) 68

69.  Message Factors:  Largest impact of all on public response  If “High Quality” Message Factors:  Influence of other factors decrease  Ability to manage public response can be high  Example: Nanticoke  If “Low Quality” Message Factors:  Influence of other factors “increases”  Ability to manage public response can be lost  Example: Three Mile Island 69

70.  Good Public Warning Response Doesn’t Happen Naturally:  Due to differences between the people being warned  Influence of People Differences:  Can be overcome by providing good messages  But Good Warning Messages Don’t Happen Naturally Either:  Require warning training & warning preparedness 70

71. 71

72.  This is a MANDATORY EVACUATION ORDER from the Yellow County Sheriff’s Department AND Fire Authority. After consulting with the U.S. Geological Survey and the National Weather Service, we issue a mandatory evacuation order for the following people in Yellow County:  If you LIVE IN or ARE IN an area BELOW or NEAR burnt slopes, evacuate now. Do not delay. This is a MANDATORY EVACUATION ORDER. Evacuate. Evacuate NOW.  What we mean when we say evacuate is: GET OUT OF ALL CANYONS, and get out of them NOW.  If you don’t live in or aren’t in an area below or near burnt slopes, you don’t need to do anything. 72

73.  There’s high risk of catastrophic mudslides and debris flows due to rain on burnt slopes:  Mudslides and debris flows could occur NOW, and they could be large enough to completely bury homes, roads, and lives.  They can occur without notice.  The amount of rain needed to start a catastrophic mudslide or debris flow is small. Don’t think you’re safe because the rainfall you see is slight.  The risk of catastrophic mudslides and debris flows below all burnt slopes in all Yellow Country is real. 73

74.  If you LIVE IN or ARE IN an area BELOW or NEAR burnt slopes evacuate NOW.  Evacuate without delay.  This is a mandatory evacuation order.  Plan on being away for several days and bring everything that you need with you.  There’s a Red Cross shelter at Monroe High School in the town of Yellow. Large animals can be brought to the Yellow County Fairgrounds.  If you have questions or require assistance, please call 123- 456-7890. 74

75.  THE MESSAGE IS:  1. CLEAR (simply worded)  2. SPECIFIC (precise and non-ambiguous)  3. ACCURATE (no error)  4. CERTAIN (authoritative and confident)  5. CONSISTENT (within and between messages)  ABOUT:  6. WHAT (what to do)  7. WHEN (when to do it)  8. WHERE (who should & shouldn’t do it)  9. WHY (hazard & consequences)  10. WHO (who’s giving the message)  AND IS CONFIRMED:  11. REPEATED frequently  12. over MULTIPLE CHANNELS 75

76.  Influenced by Information During the Event:  MESSAGE QUALITY & QUANTITY  “How people respond is more the result of the messages they’re provided during an event than it is anything else”  Has Been Observed to Be:  HIGH IN: Haz-mat events, building fires, hurricane surge zones  LOW IN: Slow-term river floods  Likely Inclined to be Higher with Increased:  Severity of event & shortness of time to impact 76

77.  When Safe People Evacuate  Occurs in all evacuations (& can be high)  Wide Variance in Observed Rates, e.g.,  Hurricane Floyd: 12%-49% coastal; 26% mean  Graniteville, SC (chlorine): 59% buffer zone mean  Depends on Messages “During” the Event:  Most don’t include safe people in warning messages  Simply Not Predictable from Pre-event:  Risk perceptions or behavioral intentions 77

78.  Monitor Public Response in Events  Find out what the public is doing  Range of Ways to Monitor, e.g.,  Traffic guides (can tell you)  Police & fire (can call in reports)  Video monitors (you can see it)  Adjust Next Messages Accordingly  What you say impacts what the public does 78

79.  Most Don’t go to Official Shelters  Friends/Relative Play Major Role:  Echo shelter needs in protracted events as:  Relatives & friends get “tired of company”  Secondary hazards occur  Many Stay in Hotels & Motels 79

80.  Evacuee Surveys  Shelter Records 80

81. 81

82.  1. Publicized  2. Older Evacuees  3. Lower Income  4. Urban Area  5. Few Friends/Relatives  6. Large Area Involved  7. Night Evacuation 82

83.  Warning Preparedness is Different from Response Preparedness Because:  Warning = pre-impact  Response = post-impact  Warning Preparedness is Out of Date:  WHY?: Society changed, preparedness for public warnings hasn’t  Here’s What Changed…… 83

84.  Most Warning Systems:  Based on a “linear” model of communication  Communication has Become “Non-linear”:  Innovations in communication technologies  Changes in communication practices  Requires Warning Systems Change to:  Preparing for “non-linear” communication  Here’s What it Looks Like…………… 84

85. 85 DETECTGOV’TMEDIAPUBLIC

86. 86 PUBLIC FRIENDS & FAMILY WHITE HOUSE PAGER RADIO CELL PHONE CNN/FOX NEWS WEATHER RADIO CO- WORKER REVERSE 911 INTERNET TV

87.  Yesterday….We Prepared for:  Emergency alert system (EAS) messages  Joint information center (JIC) press briefings  Fire fighter’s messages in buildings  Today….We Must Prepare to:  Manage a public “warning” conversation:  In which everyone is giving & getting warning information to & from everyone else 87

88.  1. Modern Plans:  Eliminate sources of past warning failures  2. Evidence-Based Messages:  Worded in ways “documented” to yield sound public response  3. Modern Technologies for Today’s World:  Provide warning in our “non-linear” communication world using modern technologies 88

89.  Eliminate Warning Provider’s Personality:  Believe myths (e.g., panic) & withhold warnings  Downplay risk as communicate “up” an organization  Provide Adaptable “Canned” Messages:  Don’t know research findings on warning messages so say something else  Don’t think about “ending” a protective action  Guarantee Repetitive Messaging:  Few know to say it many times 89

90.  Eliminate Public Confusion Resulting from What Other Warning Providers are Saying:  Address wrong information given by others  Render inconsistent information consistent  Focus public on best warnings  Give “official warning” to other warning providers to upgrade what they say 90

91.  Reach Everyone at Risk:  Use diverse devices and channels:  Avoid missing audience segments  Reach all special populations  Communicate to People not at Risk:  Safe members of public who are near  Other warning providers (many are non-local)  Monitor Response & Change What’s Said Next Accordingly:  Warnings should not be “static” but an “adapted” conversation based on how people respond 91

92.  Solve Technological Communication Problems with Technological Solutions:  Compatibility: problems observed since 1950s  Overload: inevitable  Electrical supply: some warnings go out after impact  Fail safe: has to work when needed  Mutually exclusive/redundant: more than one  Dedicated: available when needed  Customized: for special populations 92

93.  Solve Societal Communication Problems with Social Solutions:  Between organizations:  Organizations who don’t communicate routinely are disinclined to do so during warnings  Staff who don’t communicate to centralized personnel routinely don’t do so during warnings  To the public:  Install ways to reach marginalized sub-populations 93

94.  Public:  Location-specific (customize)  Diverse publics = diverse warnings pathways  People at risk & people not at risk  Organizations & Special Facilities:  Location-specific (customize)  Diverse facilities = diverse warning pathways  Facilities at risk & host facilities  Warning Partners:  Location-specific (customize)  Divers partners = diverse communication pathways  All of them (whether you want them or not):  Local, regional, statewide, national, international 94

95.  Other Organizations:  What other say in their warnings  What emergency organizations are doing  Media & Call-Ins:  Rumor control  Special Facilities:  What they’re doing  Public Protective Response:  Are they doing it or not  What the Public is Saying to Each Other:  Warning “informatics” 95

96.  Centralized Approach:  All inputs & outputs available in one place  Configuration:  State-of-the-art “virtual” communication system  Dedicated, redundant, & mutually exclusive  Systems Architecture:  Capacity to communicate (send & receive) over all of today’s devices 96

97. 97

98.  National Public Response Data Repository  Meta-analysis of Existing Survey Data:  Within & across disciplines, hazards, & agencies  Studies of Public:  Non-evacuation protective actions  Response in large urban areas  Response to no notice & short notice events  Variation in mobilization times  Ending events & issuing all clears  Evacuation vs. migration vs. area abandonment  Penetration of New Warning Technologies 98

99.  “Evidence Based” Guidance:  How to write effective warning messages  Inter-organizational warning preparedness  Prototype (Canned) Warning Messages  Modernize Warning Preparedness:  New technologies  Societal changes since plan development  Evidence-based Behavior Assumptions in Protective Action Models 99

100. 100