Semantic Web technologies and bias in artificial intelligence: A systematic literature review

Abstract

Bias in Artificial Intelligence (AI) is a critical and timely issue due to its sociological, economic and legal impact, as decisions made by biased algorithms could lead to unfair treatment of specific individuals or groups. Multiple surveys have emerged to provide a multidisciplinary view of bias or to review bias in specific areas such as social sciences, business research, criminal justice, or data mining. Given the ability of Semantic Web (SW) technologies to support multiple AI systems, we review the extent to which semantics can be a “tool” to address bias in different algorithmic scenarios. We provide an in-depth categorisation and analysis of bias assessment, representation, and mitigation approaches that use SW technologies. We discuss their potential in dealing with issues such as representing disparities of specific demographics or reducing data drifts, sparsity, and missing values. We find research works on AI bias that apply semantics mainly in information retrieval, recommendation and natural language processing applications and argue through multiple use cases that semantics can help deal with technical, sociological, and psychological challenges.

Assessment of the bias of artificial intelligence generated images and large language models on their depiction of a surgeon

“…If this data underrepresents a certain demographic, the output will likely be skewed. [20][21][22][23][24] Despite using unbiased algorithms, research shows gender biases persist. 25,26 Another concern is 'Programmer Bias', where non-representative developer demographics might introduce biases into software.…”

“…AI GANs and LLMs are trained on pre‐existing online data. If this data underrepresents a certain demographic, the output will likely be skewed 20–24 . Despite using unbiased algorithms, research shows gender biases persist 25,26 .…”

Policy advice and best practices on bias and fairness in AI

“…Primarily, models trained on raw data fail to capture the nuances found in the less-represented segments of the data distribution (Mallen et al, 2023), which often correspond to underprivileged communities. While using external knowledge sources to compensate these inequalities holds promise (Lobo et al, 2023), this objective is not central to current knowledge-informed approaches. Typically, external sources support the so-called knowledge-intensive tasks, which are those tasks requiring a significant amount of real-world knowledge (e.g., fact verification) (Petroni et al, 2021).…”

Knowledge-Grounded Target Group Language Recognition in Hate Speech

“…Semantic knowledge integration has helped to address bias related to the data annotation [30], to generalise better to unseen data [31], and to explain model predictions [32]. These examples imply, on the one hand, that it should be plausible to better overcome the discussed training data specificities given an adequate source of knowledge representation of the language from these communities.…”