The University of Glasgow is to lead a consortium with support from the universities of Edinburgh, Sheffield, Stirling, Strathclyde, York and King’s College London to develop new methods “for maximising the potential benefits of predictive and generative AI while minimising their potential for harm.”
The “potential for harm” can arise from bias and “hallucinations” where AI tools present false or invented information as fact.
“The project will pioneer participatory AI auditing, where non-experts including regulators, end-users and people likely to be affected by decisions made by AI systems will play a role in ensuring that those systems provide fair and reliable outputs,” said the University of Glasgow.
“The project will develop new tools to support the auditing process in partnership with relevant stakeholders, focusing on four key use cases for predictive and generative AI, and create new training resources to help encourage widespread adoption of the tools.
“The predictive AI use cases in the research will focus on health and media content, analysing data sets for predicting hospital readmissions and assessing child attachment for potential bias, and examining fairness in search engines and hate speech detection on social media.
“In the generative AI use cases, the project will look at cultural heritage and collaborative content generation.
“It will explore the potential of AI to deepen understanding of historical materials without misrepresentation or bias, and how AI could be used to write accurate Wikipedia articles in under-represented languages without contributing to the spread of misinformation.”
Dr Simone Stumpf of the University of Glasgow’s School of Computing Science is the project’s principal investigator.
She said: “AI is a fast-moving field, with developments often at risk of outpacing the ability of decisionmakers to ensure that the technology is used in ways that minimise the risk of harms.
“Regulators around the world are working to ensure a balance between harnessing AI’s potentially transformative benefits for society and the most effective level of oversight on its outputs.
“Auditing the outputs of AI can be a powerful tool to help develop more robust and reliable systems, but until now auditing has been unevenly applied and left mainly in the hands of experts.
“The PHAWM project will put auditing power in the hands of people who best understand the potential impact in the four fields these AI systems are operating in. That will help produce fairer and more robust outcomes for end-users and help ensure that AI technologies meet their regulatory obligations.
“By the project’s conclusion, we will have developed a robust training programme and a route towards certification of AI solutions, and a fully-featured workbench of tools to enable people without a background in artificial intelligence to participate in audits, make informed decisions, and shape the next generation of AI.”
The University of Glasgow said its researchers will play leading roles in projects supported by £12 million in new funding from Responsible AI UK (RAi UK).
It said Glasgow computing scientists are involved in two of the three new initiatives announced by RAi UK during the CogX conference in Los Angeles.
The projects will look to tackle emerging concerns of generative and other forms of AI currently being built and deployed across society.
Stumpf will lead the £3.5 million Participatory Harm Auditing Workbenches and Methodologies (PHAWM) project. Meanwhile, Professor Dame Muffy Calder and Dr Michele Sevegnani will play key roles in PROBabLE Futures – Probabilistic AI Systems in Law Enforcement Futures, a £3.5 million project led by the University of Northumbria.
RAi UK is led from the University of Southampton and backed by UK Research and Innovation (UKRI), through the UKRI Technology Missions Fund and EPSRC.
UKRI has also committed an additional £4 million of funding to further support these initiatives.
The £3.4 million PROBabLE Futures project, led by Northumbria University’s Professor Marion Oswald MBE, brings together researchers from the Universities of Glasgow, Northampton, Leicester, Cambridge and Aberdeen universities with a number of law enforcement, commercial technology, third-sector and academic partners.
The project will focus on the uncertainties of using AI for law enforcement. Professor Oswald said that AI can help police and the courts to tackle digital data overload, unknown risks, and increase operational efficiencies.
Professor Muffy Calder and Dr Michele Sevegnani, of the Glasgow’s School of Computing Science, will lead the university’s contribution to PROBabLE Futures. Last year, Professor Calder co-authored a report from The Alan Turing Institute that examined how to balance the needs of national security with individual human rights in the use of AI.
Professor Calder, also head of the University’s College of Science & Engineering, said: “I’m pleased to be part of PROBabLE Futures. This project is well-placed to help ensure that AI can be integrated effectively and ethically into law enforcement infrastructures to help keep us all safer.
“Last year, the University established our Centre for Data Science and AI to bring together our broad, multidisciplinary research base across the theory and application of artificial intelligence and machine learning.
“These two important research projects from RAi UK will help strengthen the University’s links with other leading institutions, as well as further establishing the UK as a leader in ethical AI.”
Funding has been awarded by Responsible AI UK (RAi UK) and form the pillars of its £31 million programme that will run for four years. RAi UK is backed by UK Research and Innovation (UKRI), through the UKRI Technology Missions Fund and EPSRC.