Decoding bias: Data-driven AI and transgender representation

Abstract

Artificial intelligence (AI) has taken over nearly all areas of industry. Data-driven AI systems primarily rely on large data sets, transforming identity recognition and decision-making across various industries. Although large data sets are utilised, it is often observed that societal bias against transgender individuals is evident. The study explores the influence of AI on bias across various domains. The paper focuses on three main domains: facial recognition, natural language processing (NLP), and automated hiring systems, utilising foundational research and industry examples as sources for the secondary data. The research aims to investigate how biases develop in training datasets. The lack of representation of transgender individuals in facial recognition could impact NLP datasets, potentially leading to considerable exclusion. The research proposes efficient remedies, such as gathering transgender-inclusive data collections, utilising unbiased algorithms, and promoting transgender contributions in AI systems. Tackling the biases present in AI can significantly promote transgender inclusion and fairness. This study emphasises the importance of recognising both the potential for harm and the positive impact within AI systems.

Ähnliche Arbeiten