Unsupervised Detection of Contextualized Embedding Bias with Application to Ideology
This work addresses bias detection in embeddings for applications like ideology analysis, but it is incremental as it builds on existing techniques for unsupervised learning and bias analysis.
The authors tackled the problem of detecting bias in contextualized embeddings by proposing an unsupervised method that uses social network data and combines orthogonality regularization, structured sparsity, and graph neural networks to identify an ideological subspace. Their experiments showed that this subspace encodes abstract evaluative semantics and reflects political shifts during Donald Trump's presidency.
We propose a fully unsupervised method to detect bias in contextualized embeddings. The method leverages the assortative information latently encoded by social networks and combines orthogonality regularization, structured sparsity learning, and graph neural networks to find the embedding subspace capturing this information. As a concrete example, we focus on the phenomenon of ideological bias: we introduce the concept of an ideological subspace, show how it can be found by applying our method to online discussion forums, and present techniques to probe it. Our experiments suggest that the ideological subspace encodes abstract evaluative semantics and reflects changes in the political left-right spectrum during the presidency of Donald Trump.