Home » Database

AVID-2025-R0002

Description

Geopolitical bias in sentiment analysis for neutral phrases

Details

As demonstrated in a tweet on March 19, 2022, which has since been incorporated into the model card as a known bias, the model provides biased analysis of sentiment for a neutral phrase “This film was filmed in [country]”. The tweet mentions “India” and “Iraq” explicitly.

References

AVID Taxonomy Categorization

  • Risk domains: Ethics
  • SEP subcategories: E0100: Bias/Discrimination, E0101: Group Fairness
  • Lifecycle stages: L06: Deployment

Affected or Relevant Artifacts

  • Developer: HuggingFace
  • Deployer: HuggingFace
  • Artifact Details:
    TypeName
    Modeldistilbert-base-uncased-finetuned-sst-2-english

Other information

  • Report Type: Advisory
  • Credits: Aurélien Géron
  • Date Reported: 2025-01-17
  • Version: 0.2
  • AVID Entry