Context of Natural and Artificial Intelligence Misjudgment in Nuclear Technology

Petkov, Gueorgui; Petkov, Ivan

Permalink

https://hdl.handle.net/2142/121807 Copy

Description

Title

Context of Natural and Artificial Intelligence Misjudgment in Nuclear Technology

Author(s)

• Petkov, Gueorgui
• Petkov, Ivan

Issue Date

2023

Keyword(s)

• Human Risk Assessment (HRA)
• Probalistic Risk Assessment (PRA)
• Symptom
• Context
• Judgment
• Misjudgment
• Linda problem
• Conjunction fallacy
• EOC

Abstract

The PRA methodology serves as a basis for making safety decisions, understanding and communicating the full risk profile for prioritization, effectiveness and correctness of judgments. These decisions must be based on scientific knowledge and sound statistics and heuristics, using interacting deterministic and/or probabilistic models. In order to make balanced decisions, we need to consider all short- and long-term risks, conditions, circumstances, i.e., contexts for situations that the decision-maker (non-AI or AI) experiences. Only by tracking and understanding the complementarity of the behavior of subject-object situational (SOS) systems in their dynamic context, it would be possible to provide reliable and safe measures and decisions. General automation technologies should include artificial intelligence/machine learning (AI/ML) tools to support decision-makers in operations, maintenance and repair, as well as specific needs for understandable and reliable solutions in the field of nuclear technology. This effort must follow human-centered AI principles, Human Factors Engineering (HFE) Design and Human Reliability Assessment (HRA) to understand the context model of an iterative cognitive process and explain AI/ML interface. The paper presents opportunities of the context quantification procedure of the Performance Evaluation of Teamwork method for modeling dynamic contexts of the cognition, communication and decision-making as an iterative mental process applicable for explainable AI/ML principles and interface. The lack of appropriate interaction between random/probabilistic and deterministic models to describe thought processes is a general weakness of PRA theory and practice. HRA limitations are related to the lack of the methods of mathematical psychology to give a holistic assessment of human factors (HF) affecting thought processes. The simplified replacement of context with step-like levels of cognitive behavior, an artificially interacting group of cognitive functions and mechanisms (mostly through tree-like structures - FT & ET) or decomposition and converting human action into a code of binary commands disregarding the intellect's need for justification before doing what it attempts, prescribes or decides and representing the cognitive context by manipulating and multiplying HFs by guessing their relative influence through subjective expert judgment. In this way, the real context of operation of the SOS system is not taken into account, which is actually the basis for the diversity and adaptability of natural and artificial intelligence. The paper demonstrates a procedure of the Performance Evaluation of Teamwork (PET) method for a clear and rational interpretation of the decision-making process by assuming and using overlapping types of cognitive process to quantify it in ambiguous and comparative contexts. The main idea for overcoming judgement uncertainties is to use the proposition of the symptom-wave dual nature of cognition and decision-making process to justify that at any given time the reasons for the judgment are based on tracing all possible trajectories of the actual recognized context. The purpose of this paper is to present heuristic modeling and solving of the risk and natural and artificial intelligence misjudgment issues in an ambiguous and comparative context (object & subject in situation) in nuclear technology.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121807

Owning Collections