Algebraic and logical methods in computer science and artificial intelligence

The task-based approach: a new paradigm for building trustworthy artificial intelligence

A. V. Nechesov^a, E. E. Vityaev^ab, S. S. Goncharov^ab, D. I. Sviridenko<sup>a

a</sup> The Artificial Intelligence Research Center of Novosibirsk State University, Novosibirsk, Russian Federation
^b Sobolev institute of mathematics SB RAS, Novosibirsk, Russian Federation

Abstract: While AI systems excel at reasoning within formal frameworks, their tendency to hallucinate remains a critical challenge. This paper proposes a task-based approach to enhance reliability. By focusing on the specific task and its resolution criteria, we ensure AI solutions are informed by a deep understanding of the problem's inherent limitations, including its defining axioms and theorems. This comprehension of the problem's structure and constraints is key to mitigating hallucination and building trustworthy AI.

Keywords: artificial intelligence, machine learning, agent-based approach, task approach, computability, semantic modeling.

UDC: 519.7

MSC: 68W30, 68U35

Received: 11.06.2025
Revised: 21.07.2025
Accepted: 22.07.2025

Language: English

DOI: 10.26516/1997-7670.2025.54.96