Marzieh Nabi

System-2 reasoning refers to the type of thinking that is slow, deliberate, logical, and effortful, as opposed to System-1 reasoning, which is fast, automatic, and intuitive. System-2 reasoning is often associated with tasks that require deep analysis, careful consideration of evidence, and logical deduction, such as solving complex problems, making informed decisions, or understanding nuanced arguments.

System 2 reasoning remains a significant challenge despite recent advancements in AI, including large language models (LLMs). It is a central focus of ongoing research and development at major technology companies like OpenAI and Google. System 2 reasoning is likely one of the biggest barriers to confidently applying AI and LLMs to legal applications, where consistent and trustworthy outcomes are crucial.

As part of this project, we are exploring ways to implement System 2 legal reasoning. One of our initial approaches involves integrating logic programming with LLMs, a technique known as neuro-symbolic AI. The goal is to use LLMs to convert bodies of law into knowledge graphs and logic programs, which can then be applied to specific use cases.

Project Lead: Dr. Marzieh Nabi