Automated modeling and extraction of information for decision-makers - AIME-DM Program

  1. Home
  2. Scientific programs
  3. Automated modeling and extraction of information for decision-makers - AIME-DM Program

Automated modeling and information extraction for decision-makers

Scientific program Automated Information Modeling and Extraction for Decision-Makers (AIME-DM)

The Automated Information Modeling and Extraction for Decision-Makers (AIME-DM) scientific program focuses on automating information modeling and extraction to facilitate decision-making in complex environments.

Situational awareness is central to the decision-making process, both for human cognitive mechanisms and for decision-support information systems. It enables us to model and extract the information essential to understanding decision-making issues and their contexts.

Situational awareness is based on two main principles:

  • the matching of information to a clearly defined and assessable business need, through the modeling of complex systems in their contexts;
  • automating the extraction and contextual interpretation of information from massive, heterogeneous data flows.

 

AIME-DM program objectives

The AIME-DM program offers contributions and events focusing on tools for characterizing business needs, capitalizing on expertise, and Artificial Intelligence (AI) methods for automated extraction of exploitable information by induction or deduction. It draws on work carried out in the fields of situational awareness, model and knowledge engineering, and artificial intelligence systems dedicated to the automated analysis and interpretation of multimodal and heterogeneous data. The program aims to share its contributions with the model- and knowledge-based systems (IEEE,) and AI (NeurIPS, ICML, PFIA) communities. Its members are also active in the MaDICS and RADIA GDRs.

 

Industrial applications and use cases

The scientific program also benefits from significant synergy in terms of positioning its contributions in relation to the laboratory's applied research. For example, it could contribute to:

  • FLOWS axis, in the extraction of performance indicators and their evolution in relation to systems of systems (e.g., predictive maintenance, sales forecasting, etc.),
  • DiSCS axis, improving the situational awareness of crisis management actors,
  • TRACE axis, real-time representation of a spatio-temporal system and interactions between different infrastructures,
  • WHOPS axis, automated extraction of information useful for better patient care.

 

Scientific constraints of the program

The program is structured around a four-step process for contextualizing data:

  • Modeling information needs

Meta-modeling of complex systems (such as collective self-consumption loops, virtualized biorefineries, crisis situations) in order to represent the type of business information needed to understand/supervise the system and its environment, particularly in ecosystems subject to uncertainty.

  • Steering information extraction in data flows

Building neuro-symbolic AIs that can be adapted on the fly to contexts and information needs.

  • Extracting concepts and information from data flows

Building frugal and/or self-supervised AIs in contexts where data is scarce.

  • Discovery of decision-making contexts

Build neuro-symbolic AIs that can instantiate and evolve meta-models.