Link Search Menu Expand Document
Publications
  1. Shrinking Big Data

Shrinking Big Data

Researchers exploring the antecedent behavior of public mass killers 47 and intimate partner homicide offenders 48 have identified numerous indicators of online threatening behavior that typically precede these violent crimes including expressing homicidal thoughts, acquiring weapons, articulating specific plans, and engaging in surveillance and cyberstalking of intended victims via social media. It

42 Douglas et al., Crime Classification Manual, 3-7.

43 F. Jeane Gerard, Norair Khachatryan, and Bethany Browning, “Exploration of Crime Scene Characteristics in Cyber-Related Homicides,” Homicide Studies 24, no. 1 (2020): 46-47.

44 Ibid., 47.

45 Ibid., 59.

46 Douglas et al., Crime Classification Manual, 22-23.

47 Adam Lankford, Krista Grace Adkins, and Eric Madfis, “Are the Deadliest Mass Shootings Preventable? An Assessment of Leakage, Information Reported to Law Enforcement, and Firearms Acquisition Prior to Attacks in the United States,” Journal of Contemporary Criminal Justice 35, no. 3 (2019): 316-317, 323-324.

48 Chris Todd, Joanne Bryce, and Virginia N. L. Franqueira, “Technology, Cyberstalking and Domestic Homicide: Informing Prevention and Response Strategies,” Policing and Society 31, no. 1 (2021): 82-83.

appears that law enforcement agencies should be able to prevent or disrupt a substantial number of violent crimes like lone actor killings, stranger-perpetrated homicides, and intimate partner homicides by leveraging online behavioral threat indicators, but the sheer magnitude of data that individuals produce and share on social media makes this seem like a herculean challenge. Fortunately, there are many nascent big data strategies and systems that have tremendous potential as tools for strategic and tactical criminal intelligence, and these tools often have cross-functional capabilities across different types of crisis situations like preventing mass killings and allocating resources during the COVID-19 pandemic.49 Crowdsourcing data, the practice of engaging with a community as valued partners and teaching interested citizens about online behavioral threat indicators that should be shared with law enforcement, can be an effective and flexible strategy for expanding limited resources, maintaining police transparency, and building trusting relationships.50 The value of crowdsourcing is supported by specific online behaviors that have been reported to law enforcement prior to successful mass shooting attacks including website evidence that the 1999 Columbine School shooters made specific advance threats and purchased weapons, and email evidence revealing that the 2009 Fort Hood Army Base shooter was communicating with extremists and might be planning a “heroic” suicide attack.51

While crowdsourcing is certainly a useful resource, it does not allow law enforcement to proactively search for specific behavioral indicators, and a more targeted instrument would be better suited for guiding decisions about surveillance and resource allocation. Social media data mining techniques like link analysis and sentiment analysis can equip law enforcement to actively search for specific types of information and even to analyze and classify the sources of that information.52 Link analysis uses specialized types of search algorithms to explore interconnected links and map related content, while sentiment analysis uses machine-learning methods to identify peoples’ attitudes or opinions about particular topics and to evaluate the relative strength and positive or negative quality of those opinions.53 Social media link analysis can be a particularly powerful tool toward the prevention, disruption, and prosecution of crimes like domestic terror, gang violence, and mass killing because it can be used to map social networks and to chart threatening behavioral indicators like frequenting extremist or violent fetishist websites, and liking or sharing memes that represent fascination with or incitement

49 Konstantinos Domdouzis et al., “A Social Media and Crowdsourcing Data Mining System for Crime Prevention During and Post-Crisis Situations,” Journal of Systems and Information Technology 18, no. 4 (2016): 379.

50 Ibid., 367.

51 Lankford, Adkins, and Madfis, “Deadliest Mass Shootings Preventable,” 324-326.

52 Domdouzis et al., “Data Mining for Crime Prevention,” 368-370.

53 Ibid.

to commit violent crimes. This information can be strengthened by sentiment analysis designed to expose online behavior patterns that reflect threatening attitudes like misogyny, depersonalization of outgroup members, glorification of violence, and strong positive sentiments about lawlessness and chaos.

Table of Contents