One technology used to identify dangerous information, such hate speech, is automated classifiers. People’s experiences of damage occurring online can be greatly decreased by implementing these safety precautions. These techniques are also used by researchers to determine the effects of platform modifications (such as altering rules or banning certain users or content) on the prevalence of hate speech. But as per a recent Ofcom investigation, it is crucial for researchers to disclose the classifiers they have employed and their respective performance metrics. This is because to the wide range of performance that classifiers can exhibit. For instance, popular classifiers might not work well with certain datasets.
https://www.ofcom.org.uk/news-centre/2024/how-accurate-are-online-hate-speech-detection-tools