HUMBLE: Human-centred Method for Bias-reducing Algorithms with NLP and Qualitative analysis for Improved Health Outcomes of Underserved and Marginalised Populations

There is a wealth of free-text data holding valuable insights that offers a valuable opportunity to address health inequalities by capturing the experiences of underserved groups. However, due to the time-intensive nature of analysing and reporting on free-text data, it is often overlooked and underutilised. Manual analysis falls short in handling large data volumes, while machine learning (ML) and Natural Language Processing (NLP) have potential but are limited by algorithmic biases. These biases, coupled with the increasing use of algorithms in decision-making, amplify disparities in healthcare access and outcomes for underserved and marginalised communities. While Health Psychology and Behaviour Sciences extensively investigate health disparities, there is limited presence of these Sciences in mitigating algorithmic bias. I will develop HUMBLE, a novel AI-Human Collaboration framework combining NLP, qualitative analysis, and tools from Health Psychology and Behaviour Science to address this gap. My approach provides a significant advancement over existing methodologies for analysing large volumes of free-text data, which I will use to report on experiences of healthcare and public health services of those at risk of facing multiple health issues. By minimising algorithmic bias, my innovative tool will lead to greater health democracy for marginalised and underserved populations.