Applied and Explainable AI (AXAI) Group
We are a research group in the School of Engineering and Computer Science at Te Herenga Waka—Victoria University of Wellington. We study how to build AI systems that people can understand and trust, and we apply them to problems in Aotearoa New Zealand.
Most of our projects are collaborations with researchers in other fields. We work with ecologists and geographers, legal and policy scholars, cognitive scientists, health researchers and software engineers.
Meet the group ·
Group meetings ·
Publications
Research
Each project below has a longer write-up, with the team and selected publications, on
Andrew's personal website.
Recognising taonga with AI
We are building AI that recognises individual kākā by sight and sound, so that ecologists can track birds across Wellington without catching and banding them. The project is co-led by
Dr Rachael Shaw (School of Biological Sciences) and funded by a Ministry of Business, Innovation and Employment (MBIE) Smart Ideas grant.
Students:
Nicolas Samelson,
Ryan Jaggers,
Abigail Clennell and
Fraser Campbell.
Read more about the kākā project
Explainable AI
We aim to develop methods that can better explain how AI models reach their outputs. Much of our work uses genetic programming, which evolves models as formulas that a person can read. We also find counterfactual explanations, which show the smallest change to an input that would change a model's prediction. And we study mechanistic interpretability, which explains a large language model by examining its internal parts.
Students:
Hayden Andersen and
Tan Nguyen.
Read more about explainable AI with evolutionary computation
AI, law and policy in Aotearoa
We study how New Zealand's laws and government agencies can oversee the use of AI. We ask whether current law can deal with AI-enabled harms such as sexualised deepfakes, how the government can best enable trustworthy AI, and how New Zealand should prepare for advanced AI more generally. We work with
Prof Ali Knott and
Dr Cassandra Mudgway (University of Canterbury).
Students:
Ben Cravens and
Johniel Bocacao.
Read more about AI, law and policy
Machine learning for ecosystems and climate
We use machine learning to study how New Zealand's ecosystems are changing, with data that ecologists and earth scientists have collected over decades. We mapped which tree species are becoming more or less common, and where, using forest plots surveyed between 1969 and 2019. We also mapped floods across New Zealand every week from 2016 to 2026 using satellite images, and we are training a model on those maps to forecast floods.
Dr Emily O'Riordan uses physics-informed neural networks for scientific and weather modelling. These networks are trained to follow known physical laws as well as to fit the data.
Students:
Nimasha Herath and
Ilya Shabanov.
Read more about ecosystems and climate
Other collaborations
- Health: through the Explainable AI for Health Equity fellowship, funded by Royal Society Te Apārangi, we work with researchers in the Netherlands and with New Zealand hospitals to explain how cancer treatment choices relate to patient outcomes.
- Psychology: with Prof Bart Ellenbroek, we study how stress shapes the way people use large language models. Marie Katsanos studied the impact of generative AI on creativity.
- Software security: we work on Task 4 of Tūwhana, an MBIE research programme on finding and reducing threats in software supply chains, led by A/Prof Jens Dietrich. Alix Schultze uses machine learning to detect tampering in compiled software.
- Sensing and signal processing: Mayuravaani Mathuranathan works on voice processing for hearing aids, and Sofie Claridge works on temperature sensing with optical fibres.
Working with us
Andrew's thesis supervision is currently full, so he is not taking on new PhD or MSc students except in exceptional circumstances.