A new study from the University of Birmingham proposes using sociolinguistic insights to improve large language models, addressing critical issues like social bias and misinformation. Researchers highlight the urgent need for diverse language data to create fairer and more ethical AI systems.
New research from the University of Birmingham is shedding light on a critical pathway to improving large language models (LLMs) by integrating sociolinguistic principles. This innovative approach could significantly enhance the fairness and reliability of artificial intelligence systems like ChatGPT, addressing prevalent issues of misinformation and societal biases.
The study, published in Frontiers in Artificial Intelligence, emphasizes that popular AI systems often falter due to inadequacies in the language databases used for their training. These databases fail to accurately represent the diverse dialects, registers and temporal changes intrinsic to any language, leading to AI outputs that can perpetuate harmful stereotypes and inaccuracies.
“When prompted, generative AIs such as ChatGPT may be more likely to produce negative portrayals about certain ethnicities and genders, but our research offers solutions for how LLMs can be trained in a more principled manner to mitigate social biases,” lead author Jack Grieve, a professor in the Department of Linguistics and Communication at the University of Birmingham, said in a news release.
The study suggests that if LLMs are fine-tuned on datasets reflecting the full spectrum of language diversity, the societal value of these AI systems can be greatly improved. This approach can balance the representation of different social groups and contexts, ensuring that AI systems are not only more accurate but also more ethical.
“We propose that increasing the sociolinguistic diversity of training data is far more important than merely expanding its scale,” added Grieve. “For all these reasons, we therefore believe there is a clear and urgent need for sociolinguistic insight in LLM design and evaluation.”
The implications of this research are far-reaching. By embedding a deeper understanding of societal structures and their impact on language use into the design of LLMs, the study paves the way for AI systems that better serve humanity. As AI continues to embed itself in various sectors of society, from customer service to personalized recommendations, ensuring these systems operate without bias is crucial.