The QUT Centre for Data Science is celebrating its graduates. Meet Mythreri Velmurugan who just celebrated her graduation with a PhD!
My PhD focused on understanding how reliable explainable artificial intelligence (XAI) techniques are. Modern machine learning models can be quite complex and their decision-making become incomprehensible to human users. This lack of transparency means we need special techniques to interpret their decisions safely and responsibly. My research specifically looked at these interpretation methods, known as post-hoc algorithms, and examined how the nature of different datasets affects the quality of explanations. While tabular data is common in data science, its impact on the explainability of machine learning models hasn’t been thoroughly explored. My thesis delved into what makes an explanation effective, how we can evaluate these explanations, and how well some popular explainable AI (XAI) techniques work with tabular data.
How did you come to do a PhD on this? Did things change along the way?
I started my PhD with a focus on applying explainable AI (XAI) techniques to business process management, which is a field that deals with the dynamic activities within organizations. The temporal aspect in business process management makes the application of AI and XAI quite complex. To tackle this, I first explored the fundamental aspects of XAI, particularly with tabular data, to understand the basics and see how these methods might scale to more complex scenarios. As I delved deeper into the field, I realized that XAI is rapidly evolving, so I needed to adapt my approach to stay current with the latest developments. This meant shifting my focus slightly to lay the groundwork for future research that could address these more complex applications effectively.
How was your PhD journey? Any surprises?
My PhD journey was both challenging and rewarding. One of the biggest surprises was how quickly the field of XAI evolved during my research. I started with a clear focus and scope, but as the technology advanced and XAI techniques became more diverse, I had to frequently adjust my approach to keep up with new developments. Another surprise was the complexity of working with different types of data, especially tabular data. I expected it to be straightforward, but it turned out to be much more nuanced and impactful on the quality of explanations than I had initially anticipated. Overall, the journey was a dynamic learning experience, full of unexpected turns and insights that enriched my research and deepened my understanding of both AI and its practical applications.
Did you do anything else at QUT besides your PhD?
Alongside my PhD, I was actively involved in several activities at QUT. I worked as a tutor and guest lecturer, where I developed and delivered teaching materials and assessments. I even taught international students at an overseas university through an exchange program. I also participated in the Three Minute Thesis competition in 2022, where I won within the Faculty of Science and went on to compete in the university finals. Additionally, I completed a research internship with Lockheed Martin Australia, applying my skills in analyzing complex, time-based data to a real-world project. These experiences were incredibly valuable—they helped me see how my research fits into a broader context, understand its practical implications, and adapt my work for different purposes. They also enhanced my ability to communicate my research to various audiences effectively.
How was your experience with the Data Science Centre?
My experience with the CDS was fantastic. It allowed me to connect with experts and peers from various disciplines, beyond just data science and information systems. This exposure broadened my understanding of how data science is applied across different fields and provided a wider perspective on my own work. Data science is a rapidly evolving field with ongoing debates and uncertainties, especially regarding ethics and responsible practices. Engaging with experts and interacting with academics and students from diverse areas helped me navigate these complex challenges and explore potential solutions. Overall, it was a valuable opportunity for professional growth and provided insights that have been incredibly useful in my development as a data scientist.
What’s next – or what are you doing now?
Now that I’ve graduated, I’m excited to apply my expertise to new challenges in data science and develop innovative solutions. I’m continuing to be involved in teaching, aiming to support and educate the next generation of data scientists and IT professionals. I’m also actively seeking job opportunities where I can leverage my skills to make a meaningful impact. One key takeaway from my experiences, especially through my interactions with the Data Science Centre, is that data science has the potential to influence any field.