算法透明度:从理论到实践的探索与反思

科学学研究 ›› 2024, Vol. 42 ›› Issue (7): 1354-1360.

算法透明度:从理论到实践的探索与反思

邓克涛¹,张贵红²

1. 中国科学技术大学人文与社会科学学院
2. 中国科学技术大学

收稿日期:2023-04-23 修回日期:2023-08-01 出版日期:2024-07-15 发布日期:2024-07-15
通讯作者: 张贵红
基金资助:
负责任的人工智能及其实践的哲学研究

Algorithmic Transparency: Exploration and Reflection from Theory to Practice

Received:2023-04-23 Revised:2023-08-01 Online:2024-07-15 Published:2024-07-15
Contact: 张 Gui-hongZhang

摘要/Abstract

摘要： 尽管越来越多重要的任务和决策被委托给了算法，但因算法的计算复杂性和不透明性，使得用户难以理解算法的决策过程和结果，导致其难以信任算法，甚至出现了“算法厌恶”的现象。相应地，算法透明度通常被视为可信任人工智能的基础，并在过去几年的学术辩论中获得了相当多的呼吁。然而，在实践层面，算法透明度的实施却存在诸多挑战，甚至还可能引发一定的伦理风险。基于此，本研究通过分析算法透明度实践层面的挑战和风险，指出至少可以从披露、审查以及设计这三个维度来化解当前实践层面的困境

Abstract: Although more and more important tasks and decisions have been entrusted to the algorithm, the computational complexity and opacity of the algorithm make it difficult for users to understand the decision-making process and results of the algorithm, resulting in their difficulty in trusting the algorithm, and even the phenomenon of "algorithmic aversion". Accordingly, algorithmic transparency is often seen as the foundation of trustworthy artificial intelligence and has received considerable attention in academic debates over the past few years. However, at the practical level, there are many challenges in implementing algorithmic transparency, which may even trigger certain ethical risks. Based on this, this study analyzes the challenges and risks at the practical level of algorithmic transparency, and points out that at least three dimensions of disclosure, review, and design can be used to solve the current practical difficulties.

邓克涛张贵红. 算法透明度:从理论到实践的探索与反思[J]. 科学学研究, 2024, 42(7): 1354-1360.

参考文献

[1] Jauernig J, Uhl M, Walkowitz G. People prefer moral discretion to algorithms: Algorithm aversion beyond intransparency[J]. Philosophy and Technology, 2022, 35(1):1-25. [2] Burrell J. How the machine ‘thinks’: Understanding opacity in machine learning algorithms[J]. Big data and society, 2016, 3(1):1-12. [3] Council of the European Union. European Council AI Act[EB/OL].https://artificialintell igenceact.eu/the-act/, 2021-04-21/2023-04-15. [4] ACM Technology Policy Office.Statement on Principles for Responsible Algorithmic System[EB/OL].https://www.acm.org/binaries/content/assets/public-policy/final-joint-ai-statement-update.pdf, 2022-10-26/2023-04-15. [5] Hood C, Heald D. Transparency: The key to better governance[M]. New York: Oxford University Press, 2006:7-9. [6] Larsson S , Heintz F . Transparency in artificial intelligence[J]. Internet Policy Review, 2020, 9(2):1-16. [7] Bovens M, Goodin R E, Thomas S. The Oxford Handbook of Public Accountability[M].New York: Oxford University Press, 2014:511. [8] Diakopoulos N, Koliska M. Algorithmic transparency in the news media[J]. Digital journalism, 2017, 5(7):809-828. [9] Bitzer T, Wiener M, Cram W A. Algorithmic Transparency: Concepts, Antecedents, and Consequences–A Review and Research Framework[J]. Communications of the Association for Information Systems, 2023(52):293-331. [10] Rader E, Cotter K, Cho J. Explanations as mechanisms for supporting algorithmic transparency[C]//Mandryk R. Proceedings of the 2018 CHI conference on human factors in computing systems. New York, NY, USA:Association for Computing Machinery, 2018:1-13. [11] Hayes P, Van De Poel I, Steen M. Algorithms and values in justice and security[J]. AI & Society, 2020, 35(3):533-555. [12] Grimmelikhuijsen S. Explaining Why the Computer Says No: Algorithmic Transparency Affects the Perceived Trustworthiness of Automated Decision-Making[J]. Public Administration Review, 2023, 83(2):241-262. [13] Bovens M. Analysing and assessing accountability: A conceptual framework[J]. European law journal, 2007, 13(4):447-468. [14] Matthias A. The responsibility gap: Ascribing responsibility for the actions of learning automata[J]. Ethics and information technology, 2004, 6(3) :175-183. [15] Martin K. Ethical implications and accountability of algorithms[J]. Journal of business ethics, 2019, 160(4):835-850. [16] Mallin C. The Relationship between Corporate Governance, Transparency and Financial Disclosure[J]. Corporate Governance: An International Review, 2010, 10(4):253-255. [17] Text-H.R.6580-117th Congress. Algorithmic Accountability Act of 2022[EB/OL]. www. congress.gov/bill/117th-congress/house-bill/ 6580/text, 2022-02-03/2023-04-15. [18] 张贵红, 邓克涛. 社会化研究框架下算法公平性的实现策略研究[J/OL]. 科学学研究, 2023-03-24/2023-04-15. [19] Shin D, Lim J S, Ahmad N, et al. Understanding user sensemaking in fairness and transparency in algorithms: algorithmic sensemaking in over-the-top platform[J/OL]. AI & Society, 2022-07-03/2023-04-15. [20] Scantamburlo T. Non-empirical problems in fair machine learning[J]. Ethics and Information Technology, 2021, 23(4):703-712. [21] Johnson, D G .Algorithmic accountability in the making[J]. Social Philosophy and Policy, 2021, 38(2):111-127. [22] Rubel A, Castro C, Pham A. Algorithms and autonomy: the ethics of automated decision systems[M]. Cambridge: Cambridge University Press, 2021:21. [23] Tsamados A, Aggarwal N, Cowls J, et al. The ethics of algorithms: key problems and solutions[J]. AI and Society, 2022, 37(1):215-230. [24] Vaassen B. AI, Opacity, and Personal Autonomy[J]. Philosophy and Technology, 2022, 35(4):1-20. [25] Markus D D, Pasquale F, Das S. The Oxford handbook of ethics of AI[M].New York: Oxford University Press, 2020:206-210. [26] De Laat P B. Algorithmic decision-making based on machine learning from big data: can transparency restore accountability[J]. Philosophy and Technology, 2018, 31(4):525-541. [27] 陈昌凤, 张梦. 智能时代的媒介伦理：算法透明度的可行性及其路径分析[J]. 新闻与写作, 2020(8): 75-83. Chen C F，Zhang M. Media Ethics in the Intelligent Era: Feasibility and Path Analysis of Algorithmicc Transparency[J]. News and Writing, 2020(8): 75-83. [28] O'neill O. A question of trust: The BBC Reith Lectures 2002[M].Cambridge: Cambridge University Press, 2002:73. [29] N guyen C T. Transparency is surveillance[J]. Philosophy and Phenomenological Research, 2022, 105(2):331-361. [30] Wang H. Transparency as manipulation? Uncovering the disciplinary power of algorithmic transparency[J]. Philosophy and Technology, 2022, 35(3):1-25. [31] Henin C, Le Métayer D. Beyond explainability: justifiability and contestability of algorithmic decision systems[J]. AI & Society, 2022, 37 (4):1397-1410. [32] Turilli M, Floridi L. The ethics of information transparency[J]. Ethics and Information Technology, 2009, 11(2):105-112. [33] Hildebrandt M. Privacy as protection of the incomputable self: From agnostic to agonistic machine learning[J]. Theoretical Inquiries in Law, 2019, 20(1):83-121. [34] 褚建勋, 刘新生. 试论人工智能产品可靠性与企业伦理责任[J]. 自然辩证法研究, 2022, 38(10):71-77. Chu J X, Liu X S. On Artificial Intelligence Product Reliability and Corporate Ethical Responsibility[J]. Studies in Dialectics of Nature, 2022, 38(10):71-77. [35] 闫坤如. 人工智能“合乎伦理设计”的实现路径探析[J]. 大连理工大学学报(社会科学版), 2019, 40(06):123-128. Yan K R. Exploration on Ethically Aligned Design of Artificial Intelligence [J]. Journal of Dalian University of Technology (Social Sciences), 2019, 40(06):123-128. [36] 杜严勇. 论人工智能系统的透明性[J]. 科学学研究, 2022, 40(9):1537-1543. Du Y Y. On the transparency in artificial intelligence system[J]. Studies in Science of Science, 2022, 40(09): 1537-1543.