Informatics and Applications

2025, Volume 19, Issue 2, pp 63-68

MACHINE LEARNING AND TRUST IN CLASSIFICATION RESULTS

A. A. Grusho
N. A. Grusho
M. I. Zabezhailo
V. O. Piskovski
E. E. Timonina

Abstract

It is generally accepted that trust in the artificial intelligence system is determined by the confidence of the consumer and regulatory organizations that this system is capable ofperforming the tasks assigned to it with the required quality. In the scientific literature, we are talking only about increasing trust but not about guaranteeing trust in the results ofartificial intelligence. In the interpretation ofincreasing trust, it is natural to believe that there is no trust in the results ofthe work ofartificial intelligence. In this article, a mathematical model is built, within the framework of which it is proved that in the class of artificial intelligence systems built on machine learning, there can be no guarantees of trust. The concept of "the trust in classifier" is defined if it correctly classifies new data with probability 1. The result was obtained under the conditions of the classical data space RL and a set of uniform distributions. The model can be complicated by leaving the space metric and the distributions continuous. In this case, trust does not depend on the capabilities of the classifier and on the generalization property.

[+] References (20)

GOST R 59276-2020. 2021. Sistemy iskusstvennogo intellekta. Sposoby obespecheniya doveriya. Obshchie polozheniya [Artificial intelligence systems. Methods for ensuring trust. General]. Moscow: Standardinform Publs. 16 p.
Garbuk, S. V. 2024. Funktsional'nost' i bezopasnost' sistem iskusstvennogo intellekta: kachestvo dannykh [Functionality and security of artificial intelligence systems: Data quality]. Otkrytye sistemy. SUBD [Open Systems. DBMS] 1:18-22. doi: 10.51793/OS.2024.95.90.004. EDN: IKAMSA.
Grusho, A. A., N.A. Grusho, M. I. Zabezhailo, V. O. Piskovski, E. E. Timonina, and S. Ya. Shorgin. 2024. Logika obmana v mashinnom obuchenii [Logic of deception in machine learning]. Informatika i ee Primeneniya - Inform. Appl. 18(1):78-83. doi: 10.14357/ 19922264240111. EDN: SCDSHX.
Artificial intelligence risk management framework (AI RMF 1.0). 2023. NIST AI100-1. NIST. 42 p. doi: 10.6028/ NIST.AL.100-1.
The NIST cybersecurity framework (CSF) 2.0. 2024. NIST. 27 p. doi: 10.6028/NIST.CSWP.29.
Issledovatel'skiy tsentr doverennogo iskusstvennogo intellekta (ISP RAN) [Research Center of the Trusted Artificial Intelligence ISP RAS]. Available at: https:// www.ispras.ru/ai-center (accessed May 26, 2025).
Rowe, J. 2019. Informatics association seeks tweaks to FDA's AI regulatory framework. AI Powered Healthcare. Available at: https://www.healthcareitnews.com/ai- powered-healthcare/informatics-association-seeks- tweaks-fda%E2%80%99s-ai-regulatory-framework (accessed May 26, 2025).
AI and the potential liability issues arising from use in a clinical setting. 2019. Digital Health. Available at: https://www.digitalhealth.net/2019/08/ai-and-the- potential-liability-issues-arising-from-use-in-a-clinical- setting (accessed May 26, 2025).
Grusho, A. A., M. I. Zabezhailo, and E. E. Timonina. 2020. O kauzal'noy reprezentativnosti obuchayushchikh vyborok pretsedentov v zadachakh diagnosticheskogo tipa [On causal representativeness of training samples of precedents in diagnostic type tasks]. Informatika i ee Primeneniya - Inform. Appl. 14(1):80-86. doi: 10.14357/ 19922264200111. EDN: LAQCMA.
Gusev, A. V., E. M. Astapenko, I.V Ivanov, et al. 2022. Printsipy formirovaniya doveriya k sistemam iskusstvennogo intellekta dlya sfery zdravookhraneniya [Principles for building confidence in artificial intelligence systems for healthcare]. Vestnik Roszdravnadzora [Bulletin of Ros- zdravnadzor] 2:25-33. EDN: BSMOFV.
Grusho, A., N. Grusho, and E. Timonina. 2020. Method of several information spaces for identification of anomalies. Intelligent distributed computing XIII. Eds. I. Kotenko, C. Badica, V. Desnitsky, D. El Baz, and M. Ivanovic. Studies in computational intelligence ser. Springer. 868:515- 520. doi: 10.1007/978-3-030-32258-8_60.
Grusho, A., N. Grusho, M. Zabezhailo, and E. Timonina. 2022. Evaluation of trust in computer-computed results. Distributed computer and communication networks. Eds. V M. Vishnevskiy, K. E. Samouylov, and D.V. Kozyrev. Communications in computer and information science ser. Cham, Springer. 1552:420-432. doi: 10.1007/978-3- 030-97110-6_33.
Grusho, A.A., N.A. Grusho, M.I. Zabezhailo, V.V. Kulchenkov, E. E. Timonina, and S.Ya. Shorgin. 2023. Prichinno-sledstvennye svyazi v zadachakh klassifikatsii [Causal relationships in classification problems]. Informatika i ee Primeneniya - Inform. Appl. 17(1):43-49. doi: 10.14357/19922264230106. EDN: DTQZPK.
Finn, V.K. 2023. On empirical regularities in the JSM method of automated research support. Automatic Documentation Mathematical Linguistics 57(6):362–381. doi:10.3103/S0005105523060055
Finn, V. K. 2024. On rank r empirical regularities in the JSM method of automated research support. Automatic Documentation Mathematical Linguistics 58(1):10-31. doi: 10.3103/S0005105524010035.
Finn, V. K. 2024. DSM-metod avtomatizirovannoy podderzhki issledovaniy i doverie k ego rezul'tatam [JSM method for automated research support and trust in its results]. Iskusstvennye obshchestva [Artificial Societies] 19(1):9. 36 p. doi: 10.18254/S207751800030180-7. EDN: PLMOHH.
Finn, V. K. 2024. DSM-metod avtomatizirovannoy podderzhki issledovaniy i apparat ponyatiy dlya iskusstvennogo intellekta [The JSM-method of automated research support and the conceptual apparatus for artificial intelligence]. Iskusstvennye obshchestva [Artificial Societies] 19(2):9. 75 p. doi: 10.18254/S207751800030907-6. EDN: FVVMNB.
Grusho, A. A., N. A. Grusho, M. I. Zabezhailo, E. E. Timonina, and S. Ya. Shorgin. 2023. Slozhnye prichinno- sledstvennye svyazi [Complex cause-and-effect relationships]. Informatika i ee Primeneniya - Inform. Appl. 17(2):84-89.doi: 10.14357/19922264230212. EDN: TGXQIW.
Grusho, A. A., N.A. Grusho, M.I. Zabezhailo, D. V. Smirnov, and E. E. Timonina. 2023. Klassifikatsiya s pomoshch'yu prichinno-sledstvennykh svyazey [Classification by cause-and-effect relationships]. Informati- ka i ee Primeneniya - Inform. Appl. 17(3):71-75. doi: 10.14357/19922264230310. EDN: AKWBZD.
8 shagov dlya ukrepleniya doveriya k II [8 steps for building trust in AI]. 2025. Geeks for geeks. Available at: https/www.geeksforgeeks.org/steps-for-building- trust-in-ai (accessed May 26, 2025).

[+] About this article