Conspiracy to Commit: Information Pollution, Artificial Intelligence, and Real-World Hate Crimes

• In Event: Online Extremism and Hate Speech

Abstract

Reimagining existing approaches to studying online information pollution, our paper shifts the focus from supply (articles, blogs, and posts) to demand (internet searches). Utilizing the power of state-of-the-art deep learning algorithms and capturing the online demand for information pollution via Google Trends, we study the relationship between forty racially- and politically-charged conspiracy theories’ consumption and hate crimes in the state of Michigan. This work introduces to criminology the nascent use of machine learning as a method of identifying correlations, assigning the magnitude and presence to a given variable’s capacity to improve a model’s prediction’s accuracy and its robustness against a permutation test. We find no evidence of a connection between most conspiracy theories and registered hate crimes; however, increases in online research regarding the Rothschild family, QAnon, and the Great Replacement better inform the prediction of our deep learning model to when hate crimes are likely to occur. Our work, done in parallel with Horizon Europe’s FERMI project, brings to the table new evidence on the link between online and offline actions, particularly regarding crime, as well as, contributing to understanding the benefits and limitations of using AI-based methods in the social sciences.

Authors

• Michael Victor Lo Giudice, Università Cattolica del Sacro Cuore - Transcrime

• Alberto Aziani, Università Cattolica del Sacro Cuore - Transcrime

• Ali Shadman Yazdi, Università Cattolica del Sacro Cuore - Transcrime