English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Text-to-Image (T2I) generation is enabling new applications that supportcreators, designers, and general end users of productivity software bygenerating illustrative content with high photorealism starting from a givendescriptive text as a prompt. Such models are however trained on massiveamounts of web data, which surfaces the peril of potential harmful biases thatmay leak in the generation process itself. In this paper, we take amulti-dimensional approach to studying and quantifying common social biases asreflected in the generated images, by focusing on how occupations, personalitytraits, and everyday situations are depicted across representations of(perceived) gender, age, race, and geographical location. Through an extensiveset of both automated and human evaluation experiments we present findings fortwo popular T2I models: DALLE-v2 and Stable Diffusion. Our results reveal thatthere exist severe occupational biases of neutral prompts majorly excludinggroups of people from results for both models. Such biases can get mitigated byincreasing the amount of specification in the prompt itself, although theprompting mitigation will not address discrepancies in image quality or otherusages of the model or its representations in other scenarios. Further, weobserve personality traits being associated with only a limited set of peopleat the intersection of race, gender, and age. Finally, an analysis ofgeographical location representations on everyday situations (e.g., park, food,weddings) shows that for most situations, images generated through defaultlocation-neutral prompts are closer and more similar to images generated forlocations of United States and Germany.

Social Biases through the Text-to-Image Generation Lens