The Truth About Bias and Halluciation in AI

AI is already finding its way into social housing, but how much can we trust what it tells us? In this Social Housing Round Table session, Matt Goodwin-Baird was joined by George McFerran, Director at Wyser, to explore the truth about bias and hallucination in AI and what these risks mean when the technology is used in a tenant-first environment. The discussion looked at how automatic speech recognition and large language models work, why AI can inherit human biases and produce inaccurate information, and the importance of getting the foundations right before using AI at scale. From regional accents and speech recognition to data protection, consent, governance and increasingly demanding regulatory requirements, the conversation explored both the opportunities and the risks. A recurring theme was that AI should enable people rather than replace them, combining technology's ability to process information at scale with the judgement, empathy and experience that only people can provide. A practical and thought-provoking conversation for anyone working with data, technology, customer services, governance or digital transformation in social housing. Big thank you to REACT by CMSG for sponsoring the Data and Technology Stream and to Alertacall Ltd for sponsoring The Social Housing Round Table, without them, none of this would be possible.