AI has a language gap. That could be a problem for some voters.
AI models underperform in non-English languages, which national organizations and experts say unintentionally disenfranchises communities using AI for election information.
AI chatbots are unintentionally providing inaccurate information to non-native English speakers about the upcoming midterms, creating a barrier for some voters, according to community activists and researchers. This language gap arises because AI models are predominantly trained in English, failing to capture the nuances of other languages.
This unintentional disenfranchisement disproportionately affects non-English speaking voters, who make up more than 20% of Americans. National voter organizations are working to address this issue, but Spanish-speaking voters may still be at risk of receiving incorrect information, potentially excluding them from reliable answers.
Researchers from the Institute of Strategic Dialogue found that accuracy of AI chatbots declined by 16% when prompted in Spanish about voter information. AI companies, including OpenAI, Anthropic, and Google, have stated that they are working to improve their models for non-English languages, but their APIs may not fully represent how consumers interact with chatbot interfaces.
Some organizations, like Asian Texans for Justice, have decided not to use AI for their in-language outreach due to these inaccuracies.
Written by urgent.news from POLITICO's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
