Dr Debasis Ganguly he/him
Explainable and Fair Agentic AI
Lecturer of Computing Science
School of Computing Science
Research interests:
Retrieval-Augmented Generation, Information Retrieval, Performance Prediction, Agentic AI
Research fields:
Fair Generative AI to build a better (less opinionated) society, Mitigating Exposure Biases from Generative AI, Explainable Agentic Workflows
Strategic Research Areas
Why do you want to join the DiveIn community?
Get involved with broader topics beyond the realms of computer science and AI focussing mainly on the societal application side.
Personal profile:
In our group, we actively work on Information Retrieval and Retrieval-Augmented Generation. Obvioulsy, this opens up opportunities of developing agentic RAG systems for a range of different application areas, including the healthcare domain where hallucinations need to be reduced and it’s important to cite reliable evidences, which act as explanations for the model output.
Now this is not an easy problem to fix and there’re many open problems and interesting research questions that need thorough investigation.
Another aspect of information related to the EDI is that these generative models can often be biased to specific demographics or they may exhibit particular types of topical biases, e.g., presenting only a partial perspective. In the past I’ve worked on fair AI (AAAI’20 paper) and one of my PhD students is working on fair generative AI mitigating the bias towards demographic groups. As a commitment to EDI, I want to explore a PhD topic build around the idea of mitigating information exposure biases from the generative models.
