Picture  of Joseph Kwarteng

Dr Joseph Kwarteng

Lecturer In Computing & Communications, Ai

Knowledge Media Institute

joseph.kwarteng@open.ac.uk

LinkedIn icon

Biography

Joseph Kwarteng is a Lecturer in Computing and Communications (Artificial Intelligence) at The Open University's School of Computing and Communications. His work sits at the intersection of responsible AI, computational social science, and technology for social justice, bringing together technical rigour and critical social inquiry to address some of the most pressing questions facing AI today: how to make it fair, trustworthy, and genuinely beneficial to all.

Research and Scholarship

Joseph's research spans responsible artificial intelligence, AI ethics, online safety, and educational technology. A central thread of his scholarship concerns intersectional hate speech online, in particular misogynoir, the specific and compounding forms of anti-Black misogyny directed at Black women through the combined operation of racism and sexism. His current research examines how generative AI (GenAI) and large language models (LLMs) can be deployed responsibly in education, with particular attention to Retrieval-Augmented Generation (RAG) systems, algorithmic fairness, and the mitigation of bias. He is a core contributor to the SAGE-RAI project (Smart Assessment and Guided Education with Responsible AI) and the AI Digital Assistant (AIDA) initiative, where he has contributed to the design and evaluation of RAG architectures for personalised learning platforms.

More broadly, he is committed to understanding how AI shapes and is shaped by structures of power, race, and gender. He contributes to interdisciplinary work exploring these dynamics across diverse social contexts, and actively develops ethical guidelines and best practices for the responsible adoption of generative AI in higher education.

Teaching Interests

Joseph is Co-Chair of TM342: Intelligence and Ethics in AI (working title; first presentation in February 2029), a module that forms part of the new Computer Science with Artificial Intelligence qualification. He is also a Module Team Member for AI at Work, a Continuing Professional Development short course for workplace professionals that explores the opportunities and risks of AI in communication, collaboration, and organisational life, with a critical focus on bias, ethics, and responsible use.

He's contributed to the development of AI Language Technology in the Workplace, an Open University short course co-created with students from diverse backgrounds in partnership with the Faculties of Wellbeing, Education and Language Studies, and Languages and Applied Linguistics. He has also taught database design, data management, and programming at the University of Nottingham and the University of Education, Winneba, Ghana.

Projects

SAGE-RAI: Smart Assessment and Guided Education with Responsible AI

AIEd researchers have for decades been motivated by Bloom’s famous 1984 paper which found that students taught 1-to-1 performed 2 standard deviations better than students taught in a standard classroom setting, with the average 1-to-1 student performing better than 98% of the students taught in the control group. AI offers the potential to unlock low-cost personalised teaching to massively assist both students and teachers and dramatically improve learning outcomes. The advent of Generative AI has brought this dream closer and has captured the attention of many educators. For example, a recent study of more than 1,000 K-12 teachers and 1,000 students found that 51% of teachers reported using ChatGPT - 40% weekly. Interest and take-up is similar in the Higher Education Arena. For example, an analysis of eight Russell Group universities found that there were over 1 million visits to the ChatGPT site during December 2022 and January 2023. All of the above use is via a platform provided by a US company which has not been transparent on how the underlying Large Language Model was trained and for which there are a number of issues including: •Misinformation - responses may not be truthful. •Copyright - text may be generated which is similar to existing copyrighted content. •Bias - bias in training data may mean that responses replicate existing biases and prejudices. In SAGE-RAI we will build upon a number of existing proof of concept and pilot education-oriented Generative AI tools, created by the partners, to target Bloom’s 2 Sigma problem through Responsible AI whilst also addressing the issues raised above. In particular we will build a platform that can support assessment and provide student guidance through Responsible Generative AI.