SocialGoodGPT Project

In this SocialGoodGPT project @ PIKE, we delve into the potential applications of large language models (LLMs), such as ChatGPT and Gemini, to contribute positively to humanities and societies through innovative approaches. Our focus is on the development of computational tools and methods to explore and implement these ideas, while increasing our understanding on the "human-in-the-loop" issues in these problems and solutions. Currently, our research encompasses the following areas:

1. Enhancing K12 Education: In collaboration with researchers at BUET, we are investigating the use of LLMs, particularly leveraging their multi-modal Q&A capabilities, to enhance English and Mathematics learning for K12 students. This initiative aims to provide novel and effective LLM-based educational tools.

2. Addressing LLM Challenges: Acknowledging the concerns about LLMs contributing to certain issues, we are actively studying how to mitigate these challenges. One notable issue is the generation of diverse hallucinations, which can contribute to the creation and dissemination of disinformation at scale. In response, in the spirit of "fighting fire with fire," we are exploring the potential of LLMs to detect both human-written and (hallucination-based) disinformation.