Copyright: ThalesSource : Thales


AI x Decision-Making

This challenge is over, applications are closed

Challenge Overview

Description du challenge

Decision-making involves a risk of error whose consequences can be critical, typically with consequences on human life. This risk being possibly increased by an alteration of the operator's capacities (distraction, fatigue, time pressure, stress...). The development of a cognitive twin would provide assistance to the operator under pressure by automatic learning of decision strategies and automation of tasks and/or alert in case of overload/drift.The objectives of the challenge are to:

  • Perform automatic learning of the decision-making strategies of an operator in a critical decision-making situation and subject to a disturbance of his attention capacity: overload, pressure, stress, fatigue but also potentially underload
  • Allow automatic identification of deviations from the usual or nominal decision strategy
  • Obtain an alert in case of unusual or deviant decisions
  • Explain to the operator the reasons for the alert (explanation of the automatic process) to avoid cognitive biases in the use of the system, typically too much confidence in the system
  • Allow partial automation of the decision making task to relieve the operator
Copyright: Thales
Copyright: Thales
Copyright: Sergey Nivens
Copyright: Sergey Nivens


Expert in:

  • Machine learning
  • Frugal Learning
  • Complex systems modeling
  • Risk prediction and forecasting

Experience in:

  • Identification
  • Explainability
  • Automation
  • Optimization

Projects you could be working on

Use case on medical diagnostic assistance:

  • Physiological parameters (temperature, pulse...) are acquired by the system via dedicated sensors and the medical staff informs the observations that cannot be digitized (symptoms)
  • The medical staff informs the system of its diagnosis: suspected pathology
  • The system, which has learned over a period of time the diagnostic strategies of the medical staff, alerts the medical staff when the diagnosis given is not consistent with the usual decision of this particular staff in a context of similar parameters
  • The system gives back to the medical staff the explanation of its automatic diagnosis

Use case on assistance in the targets' categorization:

  • The characteristics of the monitored targets are acquired by the system thanks to the sensors and the monitoring operator informs the system of the non-digitizable observations
  • The surveillance operator informs the system of its category decision
  • The system, which has learned for a certain time the categorizations of the monitoring operator, alerts the latter when the informed classification is not in accordance with the usual decision of this particular operator
  • The system returns to the surveillance operator the explanation of its automatic categorization decision