The New York police department has acquired a robotic police dog, known as Digidog, and has deployed it on the streets of Brooklyn, Queens and, most recently, the Bronx. At a time that activists in New York, and beyond, are calling for the defunding of police departments – for the sake of funding more vital services that address the root causes of crime and poverty – the NYPD's decision to pour money into a robot dog seems tone-deaf if not an outright provocation. As Congresswoman Alexandria Ocasio-Cortez, who represents parts of Queens and the Bronx, put it on Twitter: "Shout out to everyone who fought against community advocates who demanded these resources go to investments like school counseling instead. Now robotic surveillance ground drones are being deployed for testing on low-income communities of color with underresourced schools." There is more than enough evidence that law enforcement is lethally racially biased, and adding an intimidating non-human layer to it seems cruel.

Islamophobic hate speech on social media inflicts considerable harm on both targeted individuals and wider society, and also risks reputational damage for the host platforms. Accordingly, there is a pressing need for robust tools to detect and classify Islamophobic hate speech at scale. Previous research has largely approached the detection of Islamophobic hate speech on social media as a binary task. However, the varied nature of Islamophobia means that this is often inappropriate for both theoretically-informed social science and effectively monitoring social media. Drawing on in-depth conceptual work we build a multi-class classifier which distinguishes between non-Islamophobic, weak Islamophobic and strong Islamophobic content. Accuracy is 77.6% and balanced accuracy is 83%. We apply the classifier to a dataset of 109,488 tweets produced by far right Twitter accounts during 2017. Whilst most tweets are not Islamophobic, weak Islamophobia is considerably more prevalent (36,963 tweets) than strong (14,895 tweets). Our main input feature is a gloVe word embeddings model trained on a newly collected corpus of 140 million tweets. It outperforms a generic word embeddings model by 5.9 percentage points, demonstrating the importan4ce of context. Unexpectedly, we also find that a one-against-one multi class SVM outperforms a deep learning algorithm.