Hannah Claus, who is in her third year of studying Artificial Intelligence & Robotics, will present at the 36^th Association for the Advancement of Artificial Intelligence (AAAI) event. Hannah’s paper –‘The Importance of Hyperparameter Optimisation for Facial Recognition Applications’ – looks into how current facial recognition technologies frequently exhibit racial, gender and a number of other biases, making them less effective at recognising the faces of already marginalised groups, and how the technology can be improved.

The paper addresses a very current issue; over half of the world’s countries use facial recognition technology for surveillance technology, however most of them exhibit these underlying group biases. Hannah believes that the topical nature of her paper contributed to why it was selected for presentation at the AAAI-22 event. She said:

“The fact that so many countries use this technology, but that most show a bias against certain groups of people, means it is really important that we find universal ways to improve the quality of this type of AI so that it can be implemented everywhere without the risk of excluding already marginalised people. Facial recognition technology is everywhere. It’s in our phones, in law enforcement, at public places, some even have it within their own homes. But how quickly this technology has progressed has also resulted in problematic applications.

“I think this is why my paper, which I originally submitted as a three-minute pitch, has been chosen. It’s really exciting. I will get to connect with other researchers in the field and the audience will be genuinely interested in the topic, so I will do my best to impress them.”

Hannah’s interest in the topic comes from a personal experience of how the failings of facial recognition technology can make life harder for certain people based on their race or gender. Her paper delves into the technology, data and application of facial recognition and provides possible solutions for improving its accuracy. She continued:

“Out of 100 darker-skinned women, 35 will be misclassified compared to 1 out of 100 lighter-skinned men. I, myself, am biracial and I know that the low quality of this technology will definitely affect my family. This is why it is so important to me; it is essential that we finally talk about racial and gender bias within the new technologies which we use every day. I really want to keep doing research in this area and fight for equality in our algorithms.”

Hannah is currently collaborating as a student researcher for the German Aerospace Center (DLR), helping optimise the humanoid robot Rollin’ Justin for space exploration.

Professor of Robotics and course coordinator for Hannah’s degree, Dayou Li, commented:

“One of the biggest dreams for anybody involved in Computer Science is to receive recognition from a top class society and, from the AAAI event, Hannah has done just that. She is so enthusiastic about AI and her work on the assignment that led to this paper was outstanding. Hannah is the first AI and Robotics student from the University to present work at a top-level conference like this, so it is a brilliant result.”

Dr Vitaly Schetinin, Senior Lecturer in Computing and Information Systems and supervisor of Hannah’s project, added: “The AAAI conference is the main international conference of its type, meaning it is very impressive that Hannah has been chosen to present her paper. It is important to us that research carried out at the University is shared around the world so it can influence practice on a major scale.”

Hannah’s paper has been published as part of a chapter in Big Data Intelligence for Smart Applications. The AAAI-22 conference runs online from 22 February – 1 March 2022. For more information visit: https://aaai.org/Conferences/AAAI-22/