Research Fellow, Computing Science Department, University of Aberdeen
I’m a computer scientist with a background in Artificial Intelligence (AI). I completed my PhD on provenance and the Semantic Web at the University of Southampton. From here, I took up the post of Research Fellow on the Realising Accountable Intelligent Systems (RAInS) project at the University of Aberdeen. The RAInS project is an EPSRC-funded project that started in January 2019 and will be running till June 2021. It seeks to find ways to make intelligent systems accountable by taking both technical and socio-legal perspectives. RAInS emphasises that accountability is both proactive and reactive, and information needs to be recorded at each step of building and operating an intelligent system. We call for recording meaningful information proactively about how the system is designed and built, how it operates, and how it is used and maintained, so that if and when errors occur, we can look at that logged information and see what went wrong and why, and who is accountable.
Our approach not only takes algorithmic accountability into consideration, it looks at the whole intelligent system including software and hardware it is integrated with. So, while errors may result from biased datasets or from low performance metrics of the machine learning models, they may also result from faulty sensors or user errors. Because we are taking an interdisciplinary approach, we are working with lawyers and legislators to see what information they might need to deduce liability. We then analyse their requirements to identify what information is useful for them and how it can be recorded and audited efficiently.
One of the challenges that we face is that there is no consensus among the different computer science communities on a definition of accountability within the context of intelligent systems. In RAInS, we argue that such accountability forms an umbrella which benefits from other themes being studied, including transparency, explainability and interpretability, traceability, and auditability. The issue is further compounded when we take the perspectives of lawyers, lawmakers, and regulators into consideration, because laws and regulations are esoteric and not easily understood by the technical communities who sometimes find them technically unfeasible. Addressing these issues allows us to pave the way for trustworthy AI systems.
Working in computer science excites me as it allows me to be part of one of the most rapidly advancing fields. Think about all the major advancements in technology that have happened in the last three decades. I lived through the explosion of these and the rapid advancement and adoption of AI and smart phones and devices, and despite taking it for granted, I sometimes still find myself awed at how our daily lives are so different and much easier. Computer Science and AI do not only affect our daily lives but are now entrenched in every other field. The advances in technology aid advancements in all other fields, from medicine and healthcare, to allowing classes to carry on during pandemic lockdowns, to landing on Mars! I’m excited to see what the next advancements will be.