
RMIT PhD Student Visiting Researcher at University of Bristol
Author Natalie Campbell
Date 18 October 2023
ADM+S PhD student Edward Small from RMIT University has completed a four-month research program with Machine Learning and Computer Vision (MaVi) at the University of Bristol.
As a visiting researcher at the University of Bristol, Edward worked on Explainable AI in Healthcare and Climate Science with the Transparent AI Team (TrAIT), working across two main projects.
One project, working under Professor Santos-Rodriguez alongside Dr. Jeffrey Clark and Christopher McWilliams, looked at using counterfactual explanations to query and reason with AI models that assist in decision-making for intensive care units.
In another project, working with Dr. Jeffrey Clark, Nawid Keshtmand, Michelle Wan, Dr. Elena Fillola Mayoral, Dr. Enrico Werner and Dr. Christopher Bourdeaux, the group investigated a new method that uses counterfactuals as anchor points to assess how a patient is progressing in ICU or how a country/region compares to climate targets.
Edward says, “the work being done at Bristol in AI is deep, broad, and incredibly well supported. If a topic exists in AI, someone at Bristol is working on it.”
Reflecting on his stay, Edward says a key insight from the experience was the effectiveness of problem orientated meetings, opposed to time-boxed ones. This approach meant that the team would discuss a specific aspect of a paper/problem until they felt it was sufficiently solved.
“Sometimes this took 10 minutes, sometimes this took 2 hours, and these discussions were open for anyone to join. These were incredibly helpful in facilitating corkscrew thinking, as well as gaining insight from researchers who we may not usually look to for information,” he said.
“The team I worked with were simply fantastic to be with. Intelligent, supportive, creative, and fun. There wasn’t a single day where I didn’t want to work. The university also had very frequent talks from leading researchers, and the discussions following these talks were always very stimulating.”
During his time with the Transparent AI Team (TrAIT), Edward co-authored two papers, TraCE: Trajectory Counterfactual Explanation Scores, and Counterfactual Explanations via Locally-Guided Sequential Algorithmic Recourse. They also produced two workshop papers; the summary of one can be featured on Montreal AI ethics Institute.
“The work we completed during my time at Bristol has really inspired a new direction for us to look when it comes to XAI tools, especially in high-impact areas like healthcare.”
Upon returning to Australia, Edward’s work with the team at Bristol will continue. As well as gaining interest in new areas of AI, and inspiring new ideas for his own research, Edward says the biggest takeaway from his time at Bristol is the connections made.
“Raul, Jeff and I are meeting once a week to facilitate continued collaboration, with plans to conduct clinical trials with our support tools. We are also exploring new domains to apply our methodologies to, as well as looking to work directly with more ICU clinicians to ensure any work we do for ICU support tools is well aligned with those who will be using it.”
Edward researches fairness, explainability and transparency in automated decision-making, with supervisors Prof Flora Salim, Dr Jeffrey Chan and Dr Kacper Sokol.
His research examines the robustness and stability of current fairness strategies, and looks to resolve the mathematical conflict between group fairness and individual fairness.
The four-month program was supported by ADM+S and Bristol University.