Can learned word relationships contain social bias?
Yes. Training data can encode stereotypes and unequal associations, which may appear in embeddings. Evaluate their effects in the intended application instead of assuming numerical representations are neutral.
Explore This Question
Search for additional explanations using these tools:
More FAQs in How does AI determine that words such as “king” and “queen” are related?
Usually there is no guaranteed human-readable coordinate for a concept such as royalty. Relevant information can be distributed across many dimensions and transformed across layers.
No. The outcome varies with the embedding model, vocabulary, normalization, and comparison method. It is an illustrative analogy, not a mathematical law of language.
Still have questions?
Browse all our FAQs or reach out to our support team