Neuro-Symbolic Methods for Formalizing Regulations

Context
This internship takes place in the context of the ANR/FAPESP international collaboration project Normative Artificial Intelligence for regulating MANufacturing (NAIMAN).

The digital transformation of manufacturing industries provides a nurturing environment for the adoption of AI technologies that can quickly and flexibly respond to endogenous and exogenous changes, while being transparent and complying with sustainable regulations.

The adoption of AI technologies in these industries imposes significant challenges on companies. The integration of AI technologies in their systems and processes raises particular concerns, especially related to their compliance with the complex regulatory landscape and requirements, e.g., the EU AI Act, NIS2, the Cyber Resilience Act, GDPR, and various environmental standards. The governance of these solutions to assure their compliance with regulations and standards becomes essential.

We focus on automating the governance of systems and processes by explicitly representing regulations and standards, and using autonomous agents to assure their compliance with these regulations and standards.

For machine handling, however, regulations and standards, usually represented in natural language, have to be coded into a formalized normative language.

Recent advances in Large Language Models (LLMs) show its potential in transforming unstructured text into structured content. However, their general applicability hinders them in formalizing quality and accurate norms representing regulations and standards.

Objective
The objective of this internship is to develop neuro-symbolic methods that enhances the LLM capabilities to formalize regulations and standards into a specific normative language, the Normative Programming Language (NPL).

Responsibilities

• Design and implement a neuro-symbolic method that enhances the extraction and formalization of norms from regulations and standards documents
• Evaluate the accuracy of the designed method, possibly comparing it with competing methods
• Develop a Multiagent System using JaCaMo that uses the generated norms
• Write reports about the evolution and final outcomes of the developments

Application Procedure
If you are interested in this internship, apply to this position by sending a Letter of Motivation highlighting your main strengths for this position, a CV, the Master transcripts, and any additional information you deem relevant to Victor CHARPENAY (victor.charpenay@emse.fr) and Luis Gustavo NARDIN (gnardin@emse.fr).