To address prejudice, researchers assert that we need to explore the foundational structures within extensive language models – and how our viewpoints shape results.