Kai Xu

Data Visualisation and Human-AI training

Data Visualisation presents data visually so users can utilise their cognitive power to discover new insights and patterns that are difficult to detect with computation approach (such as statistics) alone. The simplest forms of data visualisation are bar chart and scatter plot created in Excel. Another common form of data visualisation is the dashboards in data science applications.

I am currently working on applying data visualisation to machine learning models to provide better human-AI collaboration. This involves making the model more transparent and explainable (AI to user), and on the other side (user to AI) providing an intuitive way for domain experts to integrate their knowledge and experience with the models through human-AI interaction.

I am always interested in the Computer Graphics; probably played too many games during the school years. Then I discovered data visualisation and its impact in data science applications, particularly making data science accessible to a much larger audience who are struggling with their data problems.

Besides the usual duties such as teaching, marking, paper/grant writing, I really enjoy talking to users, which include general public, business, and academics from other schools and faculties, to understand their data needs and come up with possible solutions.

Also, I like to turn the research outcomes into something more tangible (than papers) that can be used by its intended users.

The thing I enjoyed most is to see the work/software I produced making an impact on user’s professional or daily life.

Be realistic and make the most out of the resources that are available. Also collaborate widely, as most challenges are too much for any individual to tackle.

I have been trying to increase the awareness of ‘provenance’ in the context of research and data science. It is critical to reproducibility of research results and integral to the most challenging data science problems that can only be solved by AI and domain experts working together.

I hope I can turn some of the more promising research results into products so they can research more users that can benefit from them.

There is a great demand for data visualisation in the context of data science education. I try to ensure that the teaching contents include the latest in both research knowledge and practical skills.

Put your research in the context of a bigger problem (such as climate change). This will help you think about your work at a higher level, likely make the work more satisfying (by contributing to the solution to a bigger problem), and may provide guide for future work (what are the gaps in the big problem) and collaborations (whom you need to work with to solve them).