Galbraith2 jfs. Rader jfs. Folk jfs. Blackburn jfs. Adams jfs. Front St. Flick2 jfs. Jackson jfs. Geib jfs. Minor jfs. Osborne-Smith jfs. Jefferson Jefferson Co. Ferron jfs. Kurtz jfs. Boone jfs. Battiato jfs. OBrien jfs.
Porter jfs. Fisher jfs. Wilhelm jfs. Engle jfs. Reese jfs. Madison Madison County Dept. Dodge-Dorsey jfs. Muth, JD - Randall.
Muth jfs. Ringer jfs. Felton jfs. Shank - chris. Nickell - Angela. Nickell jfs. Brubaker jfs. Schwall jfs.
Swisher jfs. Edwin Moses Blvd. Niedermier2 jfs. Good jfs. Main Street Dayton, OH Burns jfs. Brown jfs. Peake jfs. Seventh St. Boyer - David. Boyer jfs. Pike Pike County Children Services Wischmeyer jfs. Kohls jfs. Harless jfs. Allen jfs.
Oliver jfs. Pulfer jfs. Jenkins jfs. Ed - Deborah. Forkas jfs. Myers jfs. Arlington St. Schaffner - Timothy. However, Twitter data does not magically provide insights about people for a specific topic such as uncertainty. Using this platform for this kind of research requires a thoughtful methodological approach.
To understand how people make sense of uncertain information around a particular event, the research does not just collect and analyze every tweet posted in relation to that event.
For one, this would return hundreds of millions of tweets, a scale at which only quantitative analysis could be applied, limiting meaningful sociobehavioral insights. Additionally, many if not most of these tweets would have nothing to do with uncertainty and risk, as this is only one topic out of many that people experience and then report on Twitter.
To gain meaningful insights about how people experience and make sense of uncertainty around hurricanes, I use a human-centered data science approach. This involves 1 collecting contextual data, 2 applying context-sensitive methods, and 3 iterating between the micro scale of individual activity and the macro scale of social dynamics. University of Colorado, Boulder, Motivated by this approach, I implemented a research agenda based on a dataset of tweets during the Atlantic hurricane season.
Specifically, these tweets contain visual representations images portraying hurricane forecasts and risks—e.
The data collection also includes associated contextual data, including all the retweets, quote tweets, and replies attached to these authoritative-sourced risk image tweets. Using a variety of context-sensitive methods, the research investigates how and why various risk representations diffuse differently on Twitter and how people make sense of a particular type of risk representation portraying uncertainty in hurricanes, the spaghetti plot.
Figure 1. Cone of uncertainty Cone of uncertainty forecast graphic showing the probable track of Hurricane Maria. Source: ABC7. Figure 2. Radar image Radar imagery of Hurricane Irma.
PDF Street Stories NYC Special Edition Hurricane Sandy (Street Stories Worldwide Book 1)
Source: NOAA. Figure 3. Spaghetti plot Spaghetti plot showing projected paths from various computer models for Hurricane Irma. Source: 10News.
First, we used a quantitative approach to study the data at a macro scale to measure the diffusion of various types of hurricane risk image tweets, or how much people engage with them. We thought this would be an indicator of how much people use each type of tweet in their decision-making and risk processing during hurricanes. We wanted to measure more active engagement with tweets that could also provide more context for why someone engaged, so we chose instead to use replies, retweets, 8 Though retweets reflect a more passive form of engagement than replying or quote-tweeting—both of which entail the user writing additional content—they are still the most common way of measuring diffusion on Twitter.
We hope that by including reply and quote tweet diffusion in this research, these forms of engagement will become just as normalized as retweets for future diffusion work, especially because they provide context about the diffusing user beyond the retweet. We also were interested in temporal patterns of diffusion, i.
These novel diffusion metrics consider the diffusion of information on Twitter contextualized within the rapid and evolving nature of hurricanes themselves. The diffusion results from this analysis alone do not tell the full story, but instead inform how to conduct subsequent qualitative analyses on the content of the contextual data surrounding different kinds of images, iterating back to the micro scale of individual activity. In particular, this involved analysis of replies and quote tweets.
For instance, we found that cone of uncertainty tweets continued to be replied to for a significantly longer amount of time than other categories, so we investigated the long-term replies to these tweets. Many people reply to cone of uncertainty image tweets with questions grounded in managing uncertainty, i. What does a normal hurricane look like?
Asking for those of us with no expertise on the subject. A deeper analysis into one type of forecast image, the spaghetti plot , elaborates upon these high-level findings across many image categories. In a study that is currently underway, we use a technique called discourse analysis to analyze conversations i.
Spaghetti plot images similar to figure 3 were chosen for this analysis because they so explicitly portray meteorological uncertainty, and preliminary analyses also revealed that the audience for these images expressed uncertainty about them. That still seem pretty reasonable even given the continued eastward shift? However, there are other responses to spaghetti plot tweets that reflect unique ways of managing and expressing uncertainty as compared to other risk image types. Spaghetti plots are themselves unique since they display a range of hurricane track forecasts produced from various computer models rather than a single forecast.
You got this Pulling4Purple. These analyses into how people engage with and make sense of hurricane forecast imagery reveal a variety of forms of uncertainty. In addition to the meteorological uncertainty inherent in the forecast information, it is important to address the uncertainty that arises when people contextualize the information to their own situations.