2084. An MCDM process integrating human judgments and AI-facilitated objective decisions
Invited abstract in session TC-10: Pairwise comparisons and preference relations 1, stream Multiple Criteria Decision Aiding.
Tuesday, 12:30-14:00Room: Clarendon SR 1.06
Authors (first author is the speaker)
| 1. | Andrej Bregar
|
| Informatika |
Abstract
AI and big data have become heavily utilized in business and production processes, services, and information systems, especially in Industry 4.0 and 5.0. Although autonomous decisions can substitute human judgments and may directly trigger actions, many issues arise related to ethics, bias, transparency, lack of expertise, ability to express subjective preferences correctly, etc. We introduce an MCDM process that aggregates holistic AI-facilitated objective decisions with subjective judgments and knowledge of human decision-makers to guarantee unbiased actions. The process is aligned with the principles of Industry 5.0 and includes phases of data processing and prediction, synthesis of decision-making knowledge, specification of objective and subjective preferences, and consolidation. It incorporates personal constraints on value functions, preference mappings, and a convergence mechanism. We demonstrate the process with a use case from the energy domain. We conduct a simulation study to show the efficiency of human-centric decision-making that complements and enriches autonomous decisions. We examine how subjective judgments enhance the objectively inferred recommendations through several factors, including the consistency of judgments, richness of discriminating information, and extremeness of preferences. We generate pairwise comparison matrices to approximate human reasoning with slightly inconsistent preferences and compare the inferred decisions with consistent ones.
Keywords
- Decision Analysis
- Decision Support Systems
- Analytics and Data Science
Status: accepted
Back to the list of papers