인공지능 알고리즘의 편향성에 대한 판단 - 기계학습을 중심으로 -
A Study on the Bias of Artificial Intelligence Algorithms - On Machine Learning
- 발행
- 2020 등록KCI API에는 발행일 항목이 없어 원문 등록일을 쓰고 있습니다. 실측으로 발행보다 최대 6년 늦습니다 — 재수집하면 발행연월로 바뀝니다.
- 소속·발행
- 목포대학교
- 출처
- 국내 KCI
- 원문
- 원문 보기 ↗
개념
키워드
Artificial Intelligence, Algorithm, Bias, Machine Learning
초록
Decisions made by artificial intelligence algorithms have biases or potential errors. This is because machine learning takes in various biased elements located on the basis of an individual or society without ethical judgment. Current laws regulate the structure of algorithms, but this focuses only on privacy. Forced disclosure of artificial intelligence algorithms can result in the designers' freedom of economic activity and disregarding the property value of the algorithms. Thus, the software designer's self-regulatory, or ethical design, should precede. It should also be given the right to know how much artificial intelligence is used in human decision-making and refuse decisions based on artificial intelligence alone. In addition, if the use of algorithms results in significant damage to social values, the law needs to be amended to recognize the designer's fault for the illegal bias of algorithms.