Zaupanje in uspešnost umetne inteligence v medicini

Miklavčič, Jonas (author)

Evropska komisija je aprila 2021 izdala predlog pravilnika, ki bo omogočil regulacijo umetne inteligence (UI). Mnogi kompleksni algoritemski sistemi, ki v medicini ponujajo največ upanja za drastičen napredek, strogih kriterijev, ki jih v svojem predlogu navaja Evropska komisija, pogosto žal ne dosegajo – npr. ne delujejo transparentno. Kot enega izmed kriterijev za presojanje etičnosti uporabe UI tako predlagam kriterij uspešnega delovanja, saj bi morda lahko sistemom, ki konsistentno delujejo uspešno, zaupali tudi, če kriterija transparentnosti ne bi dosegali. Tu naletimo na problem odnosa med zaupanjem in uspešnostjo. Ne le, da je naše zaupanje v odnosu do UI odvisno od uspešnega delovanja sistemov, pač pa je tudi uspešnost njihovega delovanja odvisna od našega zaupanja – saj njihova odličnost temelji tudi na uporabi podatkov, ki jim jih zaupamo v učne namene. Morda je edini način za rešitev tega krožnega problema ta, da sistemom zaupamo, tudi ko ti zaupanja še niso povsem vredni.

• Language: Slovenian
• Keywords: umetna inteligenca, medicina, transparentnost, zaupanje, uspešno delovanje
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: TEOF - Theological Faculty
• Publication status: Published
• Publication version: Version of Record
• Publication date: 01.01.2021
• Year: 2021
• Number of pages: Str. 935-946
• Numbering: Letn. 81, [št.] 4
• PID: 20.500.12556/RUL-164050
• UDC: 004.89:61
• ISSN on article: 0006-5722
• DOI: 10.34291/BV2021/04/Miklavcic
• COBISS.SI-ID: 91442179
• Publication date in RUL: 16.10.2024

Record is a part of a journal

• Title: Bogoslovni vestnik : glasilo Teološke fakultete v Ljubljani
• Shortened title: Bogosl. vestn.
• Publisher: Teološka fakulteta
• ISSN: 0006-5722
• COBISS.SI-ID: 8745472

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.

Secondary language

• Language: English
• Title: Trust and success of artificial intelligence in medicine

Abstract

The European Commission has proposed legislation to enable the regulation of artificial intelligence (AI). However, many complex algorithmic systems that offer the greatest hope for dramatic advances in medicine often fail to meet the strict criteria of the European Commission – e.g. they often do not operate transparently. As one of the criteria for assessing the ethical use of AI, I propose the criterion of successful performance since perhaps systems that work consistently well could be trusted even if they do not meet the criterion of transparency. Here we encounter the problem of the relationship between trust and performance. Not only does our trust in AI depend on the successful operation of the systems, but the success of their operation depends on our trust since their performance also relies on the use of data we provide for algorithm’s learning. Perhaps we need to trust the systems before they can be trustworthy.

• Keywords: artificial intelligence, medicine, transparency, trust, successful performance