Goethe Institut case study hero image

Goethe Institut

AI Translation Bias

Back to all case studies

The Problem / Context

Machine translation systems inherit human bias from the data they are trained on. Gender and racial stereotypes get embedded into AI outputs, and are hardest to detect in complex literary texts where context and nuance matter most. The Goethe Institut brought together translators, activists, and developers from around the world to hack at this problem over 52 hours.


My Role

Sole designer and researcher in a multidisciplinary winning team. I led the user-centred aspects of the project — investigating translator needs, considering the broader impact on authors and readers, defining the problem space, and shaping the direction of the tool. The project continued beyond the hackathon, with follow-up interviews with professional translators and recommendations delivered to the Goethe Institut on improving fairness in AI translation and building the corresponding tools.


Methodology

  • Problem framing and opportunity identification during the hackathon
  • Qualitative interviews with professional translators
  • Synthesis of findings
  • Recommendations for tool development

Key Findings

  • Complex literary content is where bias is most likely to occur and hardest to catch, making human review essential but difficult to target efficiently.
  • Translators needed a tool that surfaced susceptible passages rather than flagging everything, preserving their expertise while reducing cognitive load.
  • Racial and gender bias often co-occur in the same passages, requiring an intersectional approach to detection.

What I Delivered

A translation bias reduction tool concept that identifies and highlights sentences susceptible to racial and gender bias, helping translators focus their review where it matters most. Post-hackathon: translator interviews, synthesis report, and recommendations to the Goethe Institut for improving fairness in AI-assisted literary translation.

Read more about the Artificially Correct Hackathon


Outcomes / Impact

First place winner at the Goethe Institut's Artificially Correct Hackathon. Continued mentoring and institutional support to develop the solution further.