FAIRCROWD II: Decentralised and Fair Data Evaluation for Machine Learning

Position: Postdoctoral Research Associate

Machine learning models are trained on large datasets. But how do different data points help the model? This question has been a recent important research topic spanning explainablity, fairness, privacy and other aspects of machine learning. It is a subtle question that requires the investigation of the detailed behaviour of models and training algorithms. Both the definition of value and value assignemnt algorithms require in-depth theoretical and exaperimental study (see for example this and this paper written in an earlier project and their references).

Data valuation is a topic of great long term importance in the development and adoption of AI. The biggest businesses today run on models trained on personal user data. But how much is a particular person's data worth? How useful is it? Unfortunately, we the users get little information about the value of our data or how it is used. So a question all of us would like answered is: What is my data worth? How much benefit is the ML model getting from using my data? We can then make better decisions, preserve privacy, and develop better data oriented policies and regulations.

The FAIRCROWD II project aims to create efficient, private and decentralised valuations. Imagine getting quick information about the value of our data, making informed deicsions and getting fair compensations. The research will include theoretical and experimental work.

The candidate should have experience in machine learning, either in ML theory and algorithms or implementation of ML training and models. Experience in algorithm development and decentralised algorithm can also be useful. Optionally, some experience in decentralised ML, differential privacy, economics etc will be an advantage.

Areas of Interest: Those in the areas of Machine learning (theory and experimental), Algorithms, decentralised learning, differential privacy, fairness etc would find this topic interesting.

The project is flexible, and the postdoc will have opportunity to develop their own research ideas.

Position Details


Please put "FAIRCROWD II position" in the title of your email.

The FAIRCROWD project is part of the REPHRAIN National Research Centre on Privacy, Harm Reduction and Adversarial Influence Online