SFI MediaFutures invites to a seminar on bias in hate speech detection hold by Debora Nozza, on 12 of December in our office in Bergen.

Abstract:

Online hate speech is on the rise, often leading to offline harm. While NLP methods, including LLMs, offer tools for detection, fine-tuning these models can introduce severe unintended bias—performing inconsistently across comments involving specific identity terms (e.g., gay, Muslim, woman). This talk explores the risks of deploying off-the-shelf classifiers, methods for measuring and mitigating bias, and the ongoing challenge of balancing effective hate speech detection with fairness in online spaces.

Bio:

Debora Nozza is an Assistant Professor in Computing Sciences at Bocconi University. Her research interests mainly focus on Natural Language Processing, specifically on the detection and counter-acting of hate speech and algorithmic bias on Social Media data in multilingual context. She was recently awarded a €1.5m ERC Starting Grant project 2023 for research on personalized and subjective approaches to Natural Language Processing.

The event is free of charge.