Bio
Anna Gausen is researcher who is passionate about using her skillset to address socio-cultural impacts of AI. She has both academic and industry experience of implementing AI. However, her goal as a computer scientist to ensure that AI is a positive force for society.
She is pursuing a PhD in Safe and Trustworthy AI at Imperial College London. Her research addresses how to technically improve the transparency of recommendation systems on social media. Alongside her studies, she founded Accessible AI. This is a collaboration with a creative practitioner that tells stories from the intersection of technology, society and culture. The call to action is: AI is everywhere yet people don’t feel like they understand it. For AI to work for everyone, everyone must feel like they can engage with it.
Research interests
The internet and social media have been a huge force for change in our society, transforming the way we communicate. Due to the scale of information on these platforms, newsfeed recommendation algorithms have been developed to curate what users see. However, these algorithms are opaque and it is difficult to understand their impact on human communication flows. They have been criticised for heightening online polarization and the proliferation of misinformation.
Anna’s research is focused on developing a novel approach to improve the transparency of these algorithms using agent-based modelling (ABM). ABM, a type of simulation, offers the opportunity to model the complex interactions between users, algorithms and flows of information. Unlike existing methods used to audit recommendation algorithms, this enables long-term and dynamic analysis with large numbers of agents. To date, this research has resulted in two publications: evaluating the impact of misinformation countermeasures and of different recommendation objectives.
Recommendation algorithms are a flourishing application of artificial intelligence. They curate posts on social media, recommend songs on music platforms and suggest clothes through advertising. However, the underlying algorithms are usually “black box” making it difficult to understand how they are making their decisions. ABM could be a promising new avenue of research to better understand these opaque algorithms. Future work in this field could enable the use of ABM in the design and auditing of recommender algorithms, ensuring they serve society.