Predicting the Spread and Damage of Hate Speech for Effective Prevention and Intervention of Cyberhate

The project will develop novel Natural Language Processing (NLP) methods and metric for the automated prediction of the spread and damage of online hate speech. Despite significant research into studying online hate speech in the last decade, a critical limitation of existing work is that the focus has been to address 'where hate is', rather than 'where hate will be'. Predicting where hate speech will emerge, how it will spread, and to what extent it may damage online communities is crucial for the effective use of limited resources for prevention and intervention, where social media companies and NGOs invest hundreds of millions every year and yet still struggle to mitigate cyberhate. Therefore, this project will make an invaluable step towards addressing this gap in research, and develop more effective approaches and tools to address the problem.

The project will be based around the combination of computational and social sciences research methods. Although both areas have studied the problem, there has been little crossover despite the benefits a combined approach would bring. This project will: 1) create new datasets addressing online hate speech prediction tasks; 2) develop metrics to evaluate the damage of online hate speech and its damage to online communities; and 3) develop new NLP methods to identify and predict the diffusion of hate speech.