Intelligent conversational agents (chatbots) and gender stereotypes in addressing Gender-Based Violence: A taxonomy of potential threats from a threat modelling approach

Abstract

The use of artificial intelligence (AI) in the fight against gender-based violence (GBV) has gained increasing relevance, particularly through conversational agents or chatbots designed to support victims. However, the implementation of these tools is not without risks. This article analyzes the threats associated with the use of intelligent conversational agents or chatbots in addressing GBV and proposes a taxonomy of these threats based on the threat modeling approach. Based on an analysis of specialized literature and existing experiences, technological vulnerabilities, algorithmic biases, and ethical limitations are identified that may compromise the security and effectiveness of these tools. Furthermore, the discussion explores how Feminist AI can contribute to the creation of more inclusive systems that are sensitive to the needs of victims. The proposed taxonomy provides a reference framework for designing chatbots with an ethical approach, prioritizing user privacy, autonomy, and security