Cardiff Study Shows How Social Media & A.I. Can Detect Crime

Twitter Logo AH 1

A new study out of Cardiff University in the UK suggests that various social networks like Twitter and Facebook could be extremely useful in crime detection. The paper focuses primarily “disruptive events” such as riots and similar incidents that can be a public safety or security concern. The aptly titled “Can We Predict a Riot? Disruptive Event Detection Using Twitter” was published in May 2017 and authored by Nasser Alsaedi, Pete Burnap, and Omer Rana. Using computer systems, the team concludes, disruptive or dangerous occurrences can be detected up to an hour faster than they normally would be and can sometimes provide advance warning when such events are about to take place.

The study was carried out using a machine learning algorithm across five steps. That included data collection, pre-processing, classification, online clustering, and summarization. Moreover, machine learning enabled the team to derive contextual information thanks to features created from how Twitter works – including time and timeframes for related tweets, as well as the context of the text itself. They applied to the detection system to tweets posted during the 2011 riots that took place in England. By partnering with the local police force and analyzing data provided by them, the team was able to show that the system performed as well as sources on the ground at the time of those riots and even better in some cases. In fact, it was able to detect problems faster than those sources in all but two of the incidences it analyzed.

According to Burnap, machine learning has already been used in combination with natural language processing and Twitter data to “better understand online deviance” including “antagonistic narratives and cyber hate.” Burnap also points out that, as social media and internet-based communications become more ingrained in daily life, it is possible that systems such as the one created by the Cardiff University team will become an invaluable asset for public service officials in crime detection. Perhaps more importantly, because of the use of social media in times of crisis, the technology may become integral to how police officers and other related personnel respond to disruptive or dangerous events. In some cases, where those communications are used to plan or discuss criminal activity, a level of prevention may also be attainable. To be clear, there are still plenty of obstacles and issues to overcome or solve before such systems become commonplace. Privacy ethics is definitely a major issue that needs to be solved, for example before Twitter can report crimes that have yet to take place to authorities. In the meantime, it may be heartening that there are people conducting studies like this one with hopes to solve problems and improve how we react to problems.