Urgent Times Call for Urgent Measures

Joanna Maniti | October 7, 2020

USC ISI scientists research urgency detection in short crisis situations through social media

Photo/Pixabay

Photo/Pixabay

During the midst of the COVID-19 pandemic, millennials continue to utilize social media as real-time news digest, using their platforms to draw attention to informative outlets. From calling out racial injustice to enforcing voting rights, users continue to do their part in advocating for attainable solutions.

Mayank Kejriwal, a research lead at the USC Information Sciences Institute (ISI), and his team have been working on a way to harness this aspect of social media. The team is developing an urgency detection algorithm system to mobilize resources in the events of natural disasters that can be used by first responders, law enforcement, and non-governmental organizations.

Using information retrieval and minimal supervision, the research aims to detect and rank relevant crisis messages so experts can respond to urgent matters during time-sensitive events. For instance, if a COVID-19 case were to be detected in a certain city, short message streams can inform news outlets in order to ensure that that area is avoided.

Keeping Updates Accurate

A direct application of this urgency detection algorithm is to help first responders and aid agencies assess needs in crisis-stricken areas and mobilize resources effectively. Kejriwal defines an urgent message in the crisis context as “one that expresses an actionable need that needs to be resolved in a short time frame.”

Social media is a powerful tool for this. Along with the posting of tongue-in-cheek memes, sharing memories, and providing spaces for important dialogue, social media also serves as a news hub for many people. Kejriwal and his team’s urgency detection system can accurately separate relevant and irrelevant content, so it’s important that users don’t promote news that can potentially cause harm, such as conspiracy theories, hate speech, and misinformation.

“Millennial users should continue to put out accurate information on social media, while calling out those who spread misinformation and conspiracy theories,” commented Kejriwal, also a research assistant professor in the department of Industrial and Systems Engineering at USC Viterbi. “We can’t help if we can’t detect, so it’s important for the users to continue using social media platforms actively and responsibly.”

As the spread of COVID-19 cases continue to disrupt everyday life, it’s important to know where hotspots are located and emerging before they’ve completely exhausted the local health care system. Kejriwal suggests how this research discovery can treat already existing related needs for first responders.

“Urgency detection can help health workers and policymakers predict and respond quickly to emerging cases (e.g., the algorithm could look for severity of symptoms or potential COVID-19-related markers as loss of taste or smell), potentially recommending actions, such as imposing a local lockdown, before the situation spirals out of control as it recently did in Florida,” he said. “However, an even bigger value proposition that our algorithm offers is that it can be trained and deployed quickly, using very small quantities of training data, unlike other data-intensive systems, like those based on deep neural networks.”

The algorithm uses multiple methods to rank and detect relevant messages and documents. It also transfers knowledge it learned from previous disasters easily. For example, a system trained on data from New York could be applied to other cities, ensuring that users country-wide can benefit from annotation, however small it may be.

Disasters can happen anytime, anywhere, whether it be a global pandemic or terror incident. With the ability to detect urgency automatically from a live social media stream, a system could lead first responders and law enforcement to a volatile situation even before an emergency call is made. “We’re not saying that law enforcement should follow these systems blindly, but they can serve as an early warning signal that, once corroborated through a secondary source (such as a patrol car that might be in the neighborhood at the time), could lead to lives being saved due to intervening just a few minutes early,” Kejriwal explained.

In countries where government assistance may not be as readily accessible or expected, the system can be used by NGOs and nonprofits to provide improved basic care. “Today, everyone has a phone, and even cheap smartphones are becoming rapidly available,” he continued. “Cellular connections are often available in forests, deserts and even warzones (such as in pockets of war-torn Syria). As more of the world comes online and the technology continues to improve, there’ll be even greater need for AI applications like urgency detection.”

Misuse of Urgency Detection

Although some false alarms remain inevitable, one plausible solution is to use algorithms that can automatically flag such problematic elements, while encouraging users to flag and report such information themselves. Keywords involving slang that may be mislabeled as urgent can be filtered out by using these types of algorithms.

“The only true way to prevent abuse is for social media platforms themselves to have a clear and enforceable set of policies in place,” Kejriwal said. “The interference with elections shows that this is a problem that can occur at the highest levels of technological sophistication involving the likes of state-sponsored hackers. By carefully and proactively monitoring bots and other nefarious actors, we can (as a community) bring down the noise and abuse on social media to enable social-good algorithms like what we have presented to work accurately.”

Kejriwal’s team developed the urgency detection system to fit into a “larger whole,” which requires more research in this field. The system can make informed decisions on when something on social media should be flagged as urgent and when something should be flagged as misinformation that can cause its own damage.

In the wake of the COVID-19 crisis, a number of projects in network science have addressed crises. In developing countries, as well as ‘overburdened’ counties here in the US and other wealthy nations, resource allocation is a significant issue, and Kejriwal’s system can help provide an alternative means of detection and communication to help combat the pandemic.

“In poorer communities, COVID-19 may exist and may even manifest through symptoms, but people may be afraid of going to the doctor or hospital,” he said. “Social media may help in detecting these cases and preventing the virus from spreading further by directing volunteers or health workers to such households. With targeted messaging, it may be possible to put their fears to ease as well. All of these interventions can together lead to a noticeable dent in halting the virus.”

Share This Story