About CrowdSA - Problem Definition
Social Media and Mobile Access as Enablers for Crowdsourcing
With the rapid rise in the popularity of social media (800M+ Facebook users, 400M+ Twitter users)1 and the emergence of ubiquitous mobile access (5,6B+ mobile phone users)2, the sharing of observations and opinions has become common-place (2.7B+ Facebook likes/comments per day, 200M+ tweets)2. This enables unprecedented access to these social perceptions and the ability for analytics to support a variety of intelligent crowdsourcing applications where the “wisdom of the crowd” helps solving intricate problems [Naga11] The spectrum of applications is enormous, ranging from brand tracking and trend forecasting via competitor analysis, opinion mining, and image classification, to event detection and crisis management [Doa11].
Potential of Crowdsourcing in Crisis Management
Crowdsourcing has been recognized as having a tremendous potential in emergency and crisis situations caused by humans (e.g., mass panic), infrastructure (e.g., power failures), environment (e.g., floods), or a combination thereof, being often characterized by hard predictability, abrupt occurrence, long duration, and wide-area affects [Whit10, Reut12]. Basing on the “citizens as sensors” metaphor, the crowd is able to provide – at a micro-level, from various perspectives – either completely new crisis information (e.g., observations, estimations, advices, requests) or evidence for corroborating, extenuating, or disproving existing information [Meie11]. Thus, benefits arise during (i) the pre-crisis phase by detecting unexpected or unusual incidents, possibly ahead of official communications, (ii) the in-crisis phase, by providing rapid information on the evolution of the crisis (e.g., impacts and causes), not available elsewhere due to damaged or simply non-available infrastructures (e.g., sensors), and finally (iii) the post-crisis phase, by enabling some kind of feedback-loop, i.e., performing forensic analysis of crowdsourced crisis information [Came12]. Overall, there is a considerable opportunity in increasing the situational awareness of responsible authorities and the preparedness of affected citizens in turn, by sensing, analyzing and aggregating crowdsourced information [Reut12]. This is not least since citizens are an authentic source of crisis information, being intrinsically motivated to contribute, turning community knowledge into emergency intelligence [Pale07].
Massive use of Social Media in Crisis Situations
Social media have been used in crisis situations for at least 10 years for collection and dissemination of information, not only by the public but also by authorities in order to enhance crisis situation awareness [Reut12]. From a historical point of view, after the terrorist attacks of 9/11, wikis, created by citizens, were used for discovering missing people [Pale07]. In the meantime, a plethora of studies on social media in crisis situations is available, covering human crime and terroristic acts [View08], infrastructure failures [Sutt10] and natural disasters such as floods [Bruns12ab], wildfires [Lato11], hurricanes [View10], and earthquakes [Guy10] as well. It turned out that, within the first hour of a crisis, employing social media is essential since it takes up to 24 hours for mainstream media and authorities to catch up to the same information quality level [Mill09], thereby considerably enhancing awareness wrt. geo-location and situational update information [View10].
To exemplify the potential of crowdsourcing for enhancing situation awareness in crisis management, let’s consider a massive flooding scenario due to heavy rainfalls. Although conventional information sources of authorities such as weather forecasts or water-level sensors can give a first situational picture, additional crowdsourced information through social media channels – either sensed from citizens unintentionally contributing situational information or by explicit requests to a possibly closed group of intentionally contributing citizens (e.g., red cross volunteers1) – could enhance this picture in various directions. For example, detected observations on bulky material like trees in a river, could, e.g., allow a system to semi-automatically derive a potential critical situation, by forecasting that the water flow could be blocked at a bridge down the river, leading to a flooding of the surrounding infrastructure, maybe further aggravated by a dangerous “Seveso” area nearby. Further social media reports on flooded cellars and underpassings, power blackouts, unusual river crests, damages of infrastructures such as mudslides making streets near hospitals impassable, eroded zones containing electric power lines, cries for help, or reports on mutual aid initiatives soon complement the crisis situational picture. This additional crowdsourced information is available long before electric energy suppliers, local infrastructure authorities, and communities would report to crisis management in a structured manner. Thus, crowdsourcing could facilitate the estimation of evolvement and impacts of the flood, to (pro-)actively enact proper counter measures and to inform the affected citizens appropriately, increasing their preparedness.