인공지능 공정성의 이론과 실제
Theory and Practice of AI Fairness
0
조회
0
다운로드
3
인용
0
참고문헌
키워드
초록
챗GPT의 등장 이후 인공지능에 대한 사회적 관심이 다시 급격히 고조되고 있다. 많은 논자들이 챗GPT가 가져올 수 있는 다양한 위험을 우려하고 있지만, 공정성 이슈에 대한 주목도는 오히려 이전에 비해 낮아진 것으로 보인다. 이에 대해 일각에서는 인공지능 공정성 연구와 이를 기술적으로 구현한 도구들이 양적, 질적으로 성장하면서 이러한 성취를 거두었다고 긍정적으로 바라보고 있다. 이에 대한 올바른 규범적 평가를 위해서는 인공지능 공정성 분야에 대한 치밀한 이해가 선행될 필요가 있다. 이러한 문제의식에서 본고는 인공지능 공정성 담론을 이론적, 실제적 관점에서 탐구하는 작업을 진행하였다. 먼저 이론적 관점에서 인공지능 공정성 분야가 신뢰할 수 있는 인공지능 거버넌스라는 국내외 논의 지형도에서 어떤 역할을 담당하고 있는지를 살펴보았다. 이후 이러한 논의를 포괄하는 용어로 경쟁적 개념인 평등, 정의 대신 공정성이 활용되고 있는 이유에 대한 탐구를 진행하였다. 이어서 인공지능 공정성 개념을 배분적 공정성과 표상적 공정성의 하위범주로 세분화하고, 유사 개념인 편향성과 불공정성이 인공지능 맥락에서 어떤 관계에 있는지를 논의하였다. 이어서 본고는 인공지능 공정성의 이론적 내용을 실무적으로 구현하기 위한 기술적 방법론에 대해 탐구하였다. 여기에는 크게 입력 중심 접근과 출력 중심 접근이 있는데, 오늘날 실무에서 전자보다는 후자에 방점을 두는 이유가 무엇인지를 살펴보았다. 이후 두 가지 접근의 구체적 내용을 탐구하면서, 특히 출력 중심 접근의 공학적 구현물인 인공지능 공정성 지표에 대한 이해를 본격적으로 심화하였다. 여러 가지 인공지능 공정성 지표가 있을 수 있고, 이들 사이의 상충관계가 있을 수 있다는 것은 기술규제가 만능이 아니고, 이것이 인간 사회의 공정성 논의와 많은 부분 닮아 있다는 것을 보여주는 대목으로 해석될 수 있다.
Abstract
The emergence of ChatGPT has led to a significant resurgence of social interest in artificial intelligence (AI). While many scholars express concerns about the potential risks associated with ChatGPT, there seems to be a decreased focus on fairness issues compared to the past. However, some perceive this as a positive development, given the increasing quantity and quality of research on AI fairness and the corresponding technical tools. Ultimately, a comprehensive understanding of the field of AI fairness is essential to provide an accurate normative evaluation of this situation. With this concern in mind, this paper aims to explore the discourse on AI fairness from both theoretical and practical perspectives. First, from a theoretical standpoint, we examine the role of AI fairness within the broader discussions on trustworthy AI governance. Additionally, we investigate why fairness is favored over competing concepts like equality and justice. Subsequently, we delve into the subcategories of AI fairness, namely distributive fairness and representational fairness, and thoroughly explore the relationship between bias and unfairness, which share conceptual similarities within the AI context. Next, this paper delves into the technical methodologies for the practical implementation of AI fairness theory. These methodologies can generally be classified into input-centered and output-centered approaches. We analyze the current emphasis on the latter over the former. Furthermore, while examining the specific details of both approaches, we focus on gaining a deeper understanding of the AI fairness metrics, which serve as an engineering implementation of the output-centered approach. It is worth noting that the existence of various fairness metrics for AI and the potential conflicts between them highlight the fact that technical regulation is not omnipotent. Furthermore, this parallel can be seen as reminiscent of ongoing discussions on fairness within human society.
참고문헌 (67)
- 김도균, "한국 사회에서 정의란 무엇인가 - 우리 헌법에 담긴 정의와 공정의 문법" (2020), 아카넷
- 김범수, "한국 사회에서 공정이란 무엇인가 : 공정한 나를 지켜줄 7가지 정의론" (2022), 아카넷
- 대니얼 카너먼, "노이즈" (2022), 김영사
- 마이클 샌델, "공정하다는 착각" (2020), 와이즈베리
- 마이클 키언스, "알고리즘 윤리" (2021), 에이콘
- 존 롤즈, "정의론" (2002), 이학사
- 케이트 크로퍼드, "AI 지도책" (2022), 소소의책
- 고학수, "인공지능과 차별", 저스티스, 171, pp.199~277 (2019), 한국법학원[DOI]
- 김병필, "인공지능 윤리와 거버넌스" (2021), 박영사
- 김주현, "디지털 시대의 평등권", 법과사회, (71), pp.255~282 (2022), 법과사회이론학회
- 박도현, "인공지능과 해악 ― 창발적 해악론을 중심으로", 박사, pp.2021
- 박도현, "인간 편향성과 인공지능의 교차", 서울대학교 법학, 63(1), pp.139~175 (2022), 법학연구소[DOI]
- 박도현, "신뢰할 수 있는 인공지능의 이론적 고찰", 법철학연구, 25(2), pp.321~364 (2022), 한국법철학회
- 심헌섭, "정의원리의 구도", 법철학연구, 20(3), pp.291~302 (2017), 한국법철학회[DOI]
- 관계부처 합동, "사람이 중심이 되는 인공지능을 위한 신뢰할 수 있는 인공지능 실현전략(안)", ~ (2021)
- Aristotle, "Nicomachean ethics" (2014), Cambridge University Press
- Barocas, Solon, "Fairness and machine learning:Limitation and opportunities", ~ (2022)
- Fricker, Miranda, "Epistemic injustice : Power and the ethics of knowing" (2007), Oxford University Press
- Agarwal, Alekh, "Fair regression : Quantitative definitions and reduction-based algorithms", International Conference on Machine Learning (2019), PMLR
- Barocas, Solon, "Big data’s disparate impact", California Law Review, 104(3) (2016)
- Bender, Emily M., "On the dangers of stochastic parrots : Can language models be too big?", Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (2021)
- Berk, Richard, "A convex framework for fair regression", ~ (2017)
- Binns, Reuben, "On the apparent conflict between individual and group fairness", Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (2020)
- Buolamwini, Joy, "Gender shades: Intersectional accuracy disparities in commercial gender classification", Conference on Fairness, Accountability and Transparency (2018), PMLR
- Chander, Anupam, "The racist algorithm?", Michigan Law Review, 115(6) (2017)
- Corbett-Davies, Sam, "Algorithmic decision making and the cost of fairness", Proceedings of the 23rd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining (2017)
- Crawford, Kate, "Excavating AI : The politics of images in machine learning training sets", AI & Society, 36(4) (2021)
- Danks, David, "Algorithmic bias in autonomous systems", Proceedings of the 26th International Joint Conference on Artificial Intelligence (2017)
- Datta, Anupam, "Proxy non-discrimination in data-driven systems", ~ (2017)
- Deng, Jia, "ImageNet : A large-scale hierarchical image database", 2009 IEEE Conference on Computer Vision and Pattern Recognition (2009), IEEE
- Dwork, Cynthia, "Fairness through awareness", Proceedings of the 3rd Innovations in Theoretical Computer Science Conference (2012)
- Fazelpour, Sina, "Algorithmic bias : Senses, sources, solutions", Philosophy Compass, 16(8) (2021)
- Friedman, Batya, "Bias in computer systems", ACM Transactions on Information Systems, 14(3) (1996)
- Geman, Stuart, "Neural networks and the bias", variance dilemma, 1992, pp.4(1), ~
- Gillis, Talia B., "Big data and discrimination", The University of Chicago Law Review, 86(2) (2019)
- Goodman, Bryce W., "A step towards accountable algorithms? : Algorithmic discrimination and the European Union general data protection", 29th Conference on Neural Information Processing Systems (2016)
- Hardt, Moritz, "Equality of opportunity in supervised learning", Advances in Neural Information Processing Systems, 29 (2016)
- Jobin, Anna, "The global landscape of AI ethics guidelines", Nature Machine Intelligence, 1(9) (2019)
- Kamiran, Faisal, "Classifying without discriminating", 2009 2nd International Conference on Computer, Control and Communication (2009)
- Kilbertus, Niki, "Avoiding discrimination through causal reasoning", Advances in Neural Information Processing Systems, 30 (2017)
- Kleinberg, Jon, "Inherent trade-offs in the fair determination of risk scores", ~ (2016)
- Krizhevsky, Alex, "ImageNet classification with deep convolutional neural networks", Proceedings of the 25th International Conference on Neural Information Processing Systems (2012)
- Kusner, Matt J., "Counterfactual fairness", Advances in Neural Information Processing Systems, 30 (2017)
- Lakkaraju, Himabindu, "The selective labels problem : Evaluating algorithmic predictions in the presence of unobservables", Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (2017)
- Lerman, Jonas, "Big data and its exclusions", Stanford Law Review Online, 66 (2013)
- Mitchell, Shira, "Algorithmic fairness : Choices, assumptions, and definitions", Annual Review of Statistics and Its Application, 8 (2021)
- Offe, Claus, "Democracy and trust" (1999), Cambridge University Press
- Park, Sungjoon, "KLUE : Korean Language Understanding Evaluation", ~ (2021)
- Pedreshi, Dino, "Discrimination-aware data mining", Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (2008)
- Pessach, Dana, "A review on fairness in machine learning", ACM Computing Surveys (CSUR), 55(3) (2022)
- Sap, Maarten, "The risk of racial bias in hate speech detection", Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (2019)
- Sen, Amartya, "The Tanner Lecture on Human Values" (1979)
- Verma, Sahil, "Fairness definitions explained", Proceedings of the International Workshop on Software Fairness (2018)
- Weidinger, Laura, "Ethical and social risks of harm from language models", ~ (2021)
- Yang, Kaiyu, "Towards fairer datasets : Filtering and balancing the distribution of the people subtree in the ImageNet hierarchy", Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (2020)
- Angwin, Julia, "Machine bias", ProPublica, ~ (2016)
- Barr, Alistair, "Google mistakenly tags black people as ‘gorillas,’ showing limits of algorithms", The Wall Street Journal, ~ (2015)
- Crawford, Kate, "The hidden biases in big data", Harvard Business Review, ~ (2013)
- Dieterich, William, "COMPAS risk scales : Demonstrating accuracy equity and predictive parity", Northpointe Incorporation, ~ (2016)
- European Commission, "Ethics guidelines for trustworthy AI", ~ (2019)
- Hardt, Moritz, "How big data is unfair", Medium, ~ (2014)
- Heaven, Will Douglas, "Geoffrey Hinton tells us why he’s now scared of the tech he helped build", MIT Technology Review, ~ (2023)
- ISO, "TR 24028:2020 Information technology ― Artificial intelligence ―Overview of trustworthiness in artificial intelligence", ~ (2020)
- Larson, Jeff, "How we analyzed the COMPAS recidivism algorithm", ProPublica, ~ (2016)
- McCurry, Justin, "South Korean AI chatbot pulled from Facebook after hate speech towards minorities", The Guardian, ~ (2021)
- OpenAI, "GPT-4 technical report", ~ (2023)
- Vincent, James, "Twitter taught Microsoft’s AI chatbot to be a racist asshole in less than a Day", The Verge, ~ (2016)