Counteracting potential bias created by AI algorithms

An OU researcher are researching the bias that can arise through traditional Artificial Intelligence (AI)-based decision-making processes and will develop fairness-aware algorithms to counteract it.

AI-based systems are widely employed nowadays to make decisions that have far-reaching impacts on individuals and society, such as being denied a credit, a job, a medical treatment, or specific news.

Prof Harith Alani and  Dr Miriam Fernandez, in the OU’s Knowledge Media institute, are partners on the NoBias project, which received £233,000 from the EU’s Horizon 2020 Innovative Training Networks programme.

The project, which involves both academic and industry partners, will begin on 1 January and run for four years, during which time it will research where bias can arise when data is collected, when algorithms turn data into decision making capacity, or when results of decision making are used in applications

NoBIAS will develop novel methods for AI-based decision making by considering ethical and legal considerations in the design of technical solutions and to understand legal, social and technical challenges of bias in AI-decision making.

Professor Alani said:  "NoBias is a very significant opportunity to bring together leading expertise in machine learning, sociology, and law, to jointly train the next generation of AI researchers, who will be equipped with essential and interdisciplinary skills to produce ethical and lawful AI.”

Dr Fernandez; the OU leading scientist on NoBias added: “Bias can crop into AI at multiple levels, from data collection, to algorithms, to decision making. NoBias is designed to target the whole pipeline, to understand the different sources of bias, detect them, and mitigate their effect in various applications and industries.”

The full list of partners is: University of Hannover (Coordinator), GESIS, Schufa Holding AG, Centre for Research and Technology Hellas, The Open University UK, University of Southampton, University of Leuven, in addition to 10 companies across Europe, including Astra Zeneca, Expert System Iberia, Orange Labs Research, and Vodafone.