数字金融场景中人工智能模型可解释性风险的特征与治理研究
Interpretability Risksof AI Models in Digital Finance Scenarios: Features and Governance
在数字金融快速演进的背景下,人工智能(AI)模型深度嵌入风险评估、资产定价与反欺诈等关键业务环节,其可解释性不足逐渐演化为制约金融安全与信任的重要风险源。本文旨在探索AI模型可解释性风险的成因、危害、识别与治理。研究发现,AI模型的可解释性风险主要源于算法结构的高复杂度、数据样本的隐性偏倚、建模目标与可解释监管目标的不一致以及模型迭代导致的解释失效。基于此,本文从金融稳定、社会包容、法律监管与技术安全4个维度系统揭示了AI模型可解释性风险的多层次危害,同时基于透明性量化、偏见识别、合规验证与安全检测为核心的识别思路,构建了AI模型可解释性风险的识别框架。最后,提出了一种涵盖模型工程优化、数据治理与特征管理、多方审计与监管协同、标准体系建设与责任界定的综合治理体系,以期在数字金融与AI的协同发展中实现技术效率、监管可控与社会信任的动态平衡。
In the context of the rapid evolution of digital finance, artificial intelligence (AI) models are deeply integrated into critical business processes such as risk assessment, asset pricing, and anti-fraud. The resultant lack of model interpretability has progressively become a significant source of risk, constraining financial stability and public trust. This study aims to comprehensively explore the causes, harms, identification, and governance of AI model interpretability risks. It finds that the interpretability risks of AI models primarily stem from the high complexity of algorithmic structures, implicit biases within data samples, inconsistency between modeling objectives and interpretability regulatory goals, and failure of explanations due to continuous model iteration. Building upon this, the study systematically reveals the multi-layered harms of AI model interpretability risks across four key dimensions: financial stability, social inclusion, legal compliance, and technical security. Concurrently, an identification framework for AI model interpretability risks is constructed, centered on the core methodology of transparency quantification, bias identification, compliance validation, and security detection. Finally, we propose a comprehensive governance system encompassing model engineering optimization, data governance and feature management, multi-party auditing and regulatory coordination, and construction of standards systems and responsibility delineation. This framework seeks to achieve a dynamic balance among technological efficiency, regulatory controllability, and social trust in the collaborative development of digital finance and AI.
| [1] |
Li X H, Xiong H Y, Li X J, et al. Interpretable deep learning: Interpretation, interpretability, trustworthiness, and beyond [J]. Knowledge and Information Systems, 2022, 64(12): 3197‒3234. |
| [2] |
Mhlanga D. The role of big data in financial technology toward financial inclusion [J]. Frontiers in Big Data, 2024, 7: 1184444. |
| [3] |
Thekdi S, Aven T. Understanding explainability and interpretability for risk science applications [J]. Safety Science, 2024, 176: 106566. |
| [4] |
Ding W P, Abdel-Basset M, Hawash H, et al. Explainability of artificial intelligence methods, applications and challenges: A comprehensive survey [J]. Information Sciences, 2022, 615: 238‒292. |
| [5] |
Arrieta A B, Díaz-Rodríguez N, Ser J D, et al. Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI [J]. Information Fusion, 2020, 58: 82‒115. |
| [6] |
Leben D. Explainable AI as evidence of fair decisions [J]. Frontiers in Psychology, 2023, 14: 1069426. |
| [7] |
Eshete B. Making machine learning trustworthy [J]. Science, 2021, 373(6556): 743‒744. |
| [8] |
Lundberg S M, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees [J]. Nature Machine Intelligence, 2020, 2(1): 56‒67. |
| [9] |
Wachter S, Mittelstadt B, Russell C. Counterfactual explanations without opening the black box: Automated decisions and the GDPR [EB/OL]. (2017-11-01)[2025-07-10]. https://arxiv.org/abs/1711.00399. |
| [10] |
Rudin C, Chen C F, Chen Z, et al. Interpretable machine learning: Fundamental principles and 10 grand challenges [EB/OL]. (2021-03-20)[2025-07-10]. https://arxiv.org/abs/2103.11251. |
| [11] |
Ribeiro M T, Singh S, Guestrin C. "Why should I trust you?": Explaining the predictions of any classifier [R]. San Francisco: The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016. |
| [12] |
Lundberg S M, Lee S I. A unified approach to interpreting model predictions [R]. Long Beach: The 31st International Conference on Neural Information Processing Systems, 2017. |
| [13] |
Visani G, Bagli E, Chesani F, et al. Statistical stability indices for LIME: Obtaining reliable explanations for machine learning models [J]. Journal of the Operational Research Society, 2022, 73(1): 91‒101. |
| [14] |
Zaeem R N, Barber K S. The effect of the GDPR on privacy policies: Recent progress and future promise [J]. ACM Transactions on Management Information Systems, 2020, 12(1): 1‒20. |
| [15] |
Martens D, Hinns J, Dams C, et al. Tell me a story! narrative-driven XAI with large language models [J]. Decision Support Systems, 2025, 191: 114402. |
| [16] |
Feng M K, Gu J J, Qiu J, et al. From news to forecast: Integrating event analysis in LLM-based time series forecasting with reflection [R]. Vancouver: Advances in Neural Information Processing Systems 37, 2024. |
| [17] |
Cao Y P, Chen Z, Cui Z Y, et al. FinCon: A synthesized LLM multi-agent system with conceptual verbal reinforcement for enhanced financial decision making [R]. Vancouver: Advances in Neural Information Processing Systems 37, 2024. |
| [18] |
Wang H P, Yang Z J. A multi-agent approach to investor profiling using large language models [R]. Barcelona: 2025 International Conference on Control, Automation and Diagnosis (ICCAD), 2025. |
| [19] |
Bank of England. Machine learning explainability in finance: an application to default risk analysis [EB/OL]. (2019-08)[2025-07-10]. https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2019/machine-learning-explainability-in-finance-an-application-to-default-risk-analysis.pdf. |
| [20] |
招商银行. 智能投顾资产配置解释系统 [EB/OL]. (2025-08-27)[2025-08-28]. https://www.finebi.com/blog/article/68ae76dc28946ecca8878efc. |
| [21] |
China Merchants Bank. Intelligent investment advisory asset allocation explanation system [EB/OL]. (2025-08-27)[2025-08-28]. https://www.finebi.com/blog/article/68ae76dc28946ecca8878efc. |
| [22] |
中国建设银行. 智能客服系统解释框架[EB/OL]. (2025-08-27)[2025-08-28]. https://www.finebi.com/blog/article/68ae76dc28946ecca8878efc. |
| [23] |
China Construction Bank. Interpretation framework for intelligent customer service system [EB/OL]. (2025-08-27)[2025-08-28]. https://www.finebi.com/blog/article/68ae76dc28946ecca8878efc. |
| [24] |
微众银行. 联邦反洗钱可解释模型 [EB/OL]. (2024-08-14)[2025-08-28]. https://cloud.baidu.com/article/3324961. |
| [25] |
WeBank. Federated anti-money laundering explainable model [EB/OL]. (2024-08-14)[2025-08-28]. https://cloud.baidu.com/article/3324961. |
| [26] |
Talaat F M, Aljadani A, Badawy M, et al. Toward interpretable credit scoring: Integrating explainable artificial intelligence with deep learning for credit card default prediction [J]. Neural Computing and Applications, 2024, 36(9): 4847‒4865. |
| [27] |
刘璇, 张蕾, 刘钟, 基于数据挖掘技术的企业信用风险评估 [J]. 山西财经大学学报, 2023, 45(S2): 89‒91. |
| [28] |
Liu X, Zhang L, Liu Z, et al. Enterprise credit risk assessment based on data mining technology [J]. Journal of Shanxi University of Finance and Economics, 2023, 45(S2): 89‒91. |
| [29] |
Hakkoum H, Idri A, Abnane I. Global and local interpretability techniques of supervised machine learning black box models for numerical medical data [J]. Engineering Applications of Artificial Intelligence, 2024, 131: 107829. |
| [30] |
刘朝. 算法歧视的表现、成因与治理策略 [J]. 人民论坛, 2022 (2): 64‒68. |
| [31] |
Liu C. Manifestations, causes and governance strategies of algorithm discrimination [J]. People's Tribune, 2022 (2): 64‒68. |
| [32] |
Ford J. Fairness in focus: quantitative insights into bias within machine learning risk evaluations and established credit models [J]. Management System Engineering, 2025, 4(1): 8. |
| [33] |
罗熠琛, 周新衡. 算法自动化决策中就业性别歧视的类型与规范路径 [J]. 中国人力资源开发, 2025, 42(1): 92‒103. |
| [34] |
Luo Y C, Zhou X H. Types and normative pathways of employment gender discrimination in algorithmic automated decision-making [J]. Human Resources Development of China, 2025, 42(1): 92‒103. |
| [35] |
Weber P, Carl K V, Hinz O. Applications of explainable artificial intelligence in finance—A systematic review of finance, information systems, and computer science literature [J]. Management Review Quarterly, 2024, 74(2): 867‒907. |
| [36] |
Abbasi W, Mori P, Saracino A. Trading-off privacy, utility, and explainability in deep learning-based image data analysis [J]. IEEE Transactions on Dependable and Secure Computing, 2025, 22(1): 388‒405. |
| [37] |
陈贞竹, 李力, 余昌华. 我国货币政策传导效率及信号效应研究——基于金融市场高频识别的视角 [J]. 经济学(季刊), 2023, 23(1): 56‒73. |
| [38] |
Chen Z Z, Li L, Yu C H. Transmission of monetary policy shocks and signaling channels in China—Based on high frequency data in financial markets [J]. China Economic Quarterly, 2023, 23(1): 56‒73. |
| [39] |
许荣, 田文涛, 胡学峰. 量化交易的风险影响与监管体系构建研究 [J]. 金融监管研究, 2025 (2): 74‒92. |
| [40] |
Xu R, Tian W T, Hu X F. Research on the impact of quantitative trading risks and the construction of the regulatory system [J]. Financial Regulation Research, 2025 (2): 74‒92. |
| [41] |
Zhang R X. Toward interpretable machine learning: Evaluating models of heterogeneous predictions [J]. Annals of Operations Research, 2025, 347(2): 867‒887. |
| [42] |
郑挺国, 葛厚逸. 中国股市羊群效应的区制转移时变性研究 [J]. 金融研究, 2021 (3): 170‒187. |
| [43] |
Zheng T G, Ge H Y. A study of the time-varying characteristics of herding effects in China's stock market based on a regime-switching model [J]. Journal of Financial Research, 2021 (3): 170‒187. |
| [44] |
荆思寒, 王振山, 隋聪, 股票间的风险传染——基于对股价崩盘风险的预测 [J]. 系统工程理论与实践, 2022, 42(11): 3090‒3104. |
| [45] |
Jing S H, Wang Z S, Sui C, et al. The risk contagion among stocks—Based on the prediction of crash risk on stock price [J]. Systems Engineering—Theory & Practice, 2022, 42(11): 3090‒3104. |
| [46] |
Garcia A C B, Garcia M G P, Rigobon R. Algorithmic discrimination in the credit domain: What do we know about it? [J]. AI & Society, 2024, 39(4): 2059‒2098. |
| [47] |
Cui P, Athey S. Stable learning establishes some common ground between causal inference and machine learning [J]. Nature Machine Intelligence, 2022, 4(2): 110‒115. |
| [48] |
刘艳红. 人工智能的可解释性与AI的法律责任问题研究 [J]. 法制与社会发展, 2022, 28(1): 78‒91. |
| [49] |
Liu Y H. Research on the interpretability of artificial intelligence and the legal responsibility of AI [J]. Law and Social Development, 2022, 28(1): 78‒91. |
| [50] |
李青青, 张凯, 李晋国, 基于集成学习的入侵检测系统对抗攻击检测 [J]. 计算机工程与设计, 2025, 46(3): 850‒856. |
| [51] |
Li Q Q, Zhang K, Li J G, et al. Adversarial attacks detection in intrusion detection systems with ensemble learning [J]. Computer Engineering and Design, 2025, 46(3): 850‒856. |
| [52] |
Mo G L, Zhang G L, Tan C Z, et al. Reassessment of corporate credit risk identification: Novel discoveries from integrated machine learning models [J]. Computational Economics, 2025, 66(4): 2791‒2841. |
| [53] |
Toy T. Transparency in AI [J]. AI & Society, 2024, 39(6): 2841‒2851. |
| [54] |
储节旺, 夏莉. 嵌入生命周期理论的科学数据管理体系构建研究——牛津大学为例 [J]. 现代情报, 2020, 40(10): 34‒42. |
| [55] |
Chu J W, Xia L. Research on scientific dta management construction based on life cycle theory [J]. Journal of Modern Information, 2020, 40(10): 34‒42. |
| [56] |
Al-Okaily M, Al-Okaily A. Financial data modeling: An analysis of factors influencing big data analytics-driven financial decision quality [J]. Journal of Modelling in Management, 2025, 20(2): 301‒321. |
| [57] |
邓胜利, 丁威威, 汪璠, “工具 ‒ 结构”视角下国内外生成式人工智能监管政策比较研究 [J]. 信息资源管理学报, 2025, 15(1): 54‒68. |
| [58] |
Deng S L, Ding W W, Wang F, et al. A multinational comparative study of regulatory policies for generative AI from the perspective of "tools‒structure" [J]. Journal of Information Resources Management, 2025, 15(1): 54‒68. |
| [59] |
Gramegna A, Giudici P. SHAP and LIME: An evaluation of discriminative power in credit risk [J]. Frontiers in Artificial Intelligence, 2021, 4: 752558. |
| [60] |
Dastani M, Yazdanpanah V. Responsibility of AI systems [J]. AI & Society, 2023, 38(2): 843‒852. |
国家自然科学基金项目(72401144)
/
| 〈 |
|
〉 |