British MP’s AI avatar struggles with Yorkshire accents

Leeds MP Mark Sewards has launched what’s billed as the first AI avatar of a British MP, allowing constituents to interact with a digital version that responds in his voice. However, testing by The Guardian revealed significant limitations when the AI encountered Yorkshire accents and local dialect, highlighting ongoing challenges with voice recognition technology for regional speech patterns.

How it works: The “Sewardsbot” features a Pixar-style cartoon representation of the Labour MP for Leeds South West and Morley that can field questions from constituents.
• The AI responds in Sewards’ actual voice, offering advice, support, or promising to pass messages to his team.
• All conversations are recorded with the aim of identifying key topics constituents are discussing.
• The website includes a disclaimer warning that responses are AI-generated and shouldn’t be taken as fact.

The accent problem: Testing with authentic Yorkshire dialect exposed serious transcription issues that could limit the tool’s effectiveness for local constituents.
• Common Yorkshire greetings like “now then” (meaning hello) weren’t recognized as greetings.
• Regional phrases were often transcribed as “unreadable gobbledegook.”
• The AI failed to understand that glottal stops before words indicate “the” in local speech patterns.
• A complex sentence about a neighbor blocking an alley with furniture was largely misunderstood.

Mixed results on policy questions: While the AI struggled with accent recognition, it showed some ability to identify broad topic areas.
• When asked about Gaza using colloquial language (“Are you gonna do summat about Gaza?”), the bot recognized the topic but provided no explanation of government position.
• For a local issue about fly-tipping, the AI incorrectly suggested contacting police about an “abandoned vehicle” rather than the appropriate council services.

Why this matters: The experiment reveals a fundamental tension between AI innovation in public services and accessibility for diverse communities, particularly those with strong regional accents who may already face barriers in accessing government services.