Multi-party responsibility contribution and ethical decision-making for autonomous driving under AI moral dilemmas

doi:10.3969/j.issn.1674-8484.2026.04.003

Abstract

Abstract:

A multi-party responsibility contribution analysis and ethical decision-making method was proposed to address the unclear criteria for identifying artificial intelligence (AI) moral dilemmas in autonomous driving and the insufficient quantification of ethical decision-making basis and responsibility contributions. Based on typical accident scenarios, variables related to vehicle states, road environments, traffic participants, human-machine interactions, and data traceability were extracted to establish the association between responsibility factors and responsible agents. A comprehensive ethical cost function was constructed, and joint simulations using SCANeR and MATLAB were conducted. The results show that, under the baseline condition, the emergency braking strategy achieves the lowest comprehensive ethical cost with a value of 0.276, and the relative responsibility contributions of external traffic participants and the autonomous driving system are 0.50 and 0.26, respectively. Furthermore, the ranking of candidate strategies remains stable under weight perturbations. The proposed method provides a reference for ethical decision-making and accident process analysis in autonomous driving systems.

Key words: autonomous driving, artificial intelligence (AI) moral dilemma, multi-party responsibility contribution, ethical decision-making, comprehensive ethical cost function

CLC Number:

U467.1

XIA Xue, XU Shucai, LI Haoran, QIAN Chuang. Multi-party responsibility contribution and ethical decision-making for autonomous driving under AI moral dilemmas[J]. Journal of Automotive Safety and Energy, 2026, 17(4): 450-458.

Figures/Tables 13

References 16

[1]	郑志峰. 自动驾驶汽车“电车难题”的伦理困境与立法因应[J]. 法学, 2024(11): 107-123.
	ZHENG Zhifeng. Ethical dilemma and legislative response to the “trolley problem” of autonomous vehicles[J]. Law Sci, 2024(11): 107-123. (in Chinese)
[2]	冯洁语. 人工智能技术与责任法的变迁:以自动驾驶技术为考察[J]. 比较法研究, 2018(2): 143-155.
	FENG Jieyu. Artificial intelligence technology and the transformation of liability law: Taking autonomous driving technology as an example[J]. J Comparat Law, 2018(2): 143-155. (in Chinese)
[3]	孔祥稳. 面向人工智能风险的行政规制革新:以自动驾驶汽车的行政规制为中心而展开[J]. 行政法学研究, 2020(4): 18-33.
	KONG Xiangwen. Innovation of administrative regulation for artificial intelligence risks: Centered on the administrative regulation of autonomous vehicles[J]. Admin Law Rev, 2020(4): 18-33. (in Chinese)
[4]	Bonnefon J F, Shariff A, Rahwan I. The social dilemma of autonomous vehicles[J]. Science, 2016, 352(6293): 1573-1576.
[5]	李克强, 戴一凡, 李升波, 等. 智能网联汽车(ICV)技术的发展现状及趋势[J]. 汽车安全与节能学报, 2017, 8(1): 1-14.
	LI Keqiang, DAI Yifan, LI Shengbo, et al. Development status and trend of intelligent connected vehicle technology[J]. J Autom Safe Energ, 2017, 8(1): 1-14. (in Chinese)
[6]	Goodall N J. Ethical decision making during automated vehicle crashes[J]. Transport Res Record: J Transport Res Board, 2014, 2424(1): 58-65.
[7]	彭湃, 耿可可, 王子威, 等. 智能汽车环境感知方法综述[J]. 机械工程学报, 2023, 59(20): 281-303.
	PENG Pai, GENG Keke, WANG Ziwei, et al. Review of environment perception methods for intelligent vehicles[J]. J Mech Engi, 2023, 59(20): 281-303. (in Chinese)
[8]	Awad E, Dsouza S, Kim R, et al. The Moral Machine experiment[J]. Nature, 2018, 563(7729): 59-64.
[9]	朱冰, 贾士政, 赵健, 等. 自动驾驶车辆决策与规划研究综述[J]. 中国公路学报, 2024, 37(1): 215-240.
	ZHU Bing, JIA Shizheng, ZHAO Jian, et al. Review of decision-making and planning for autonomous vehicles[J]. Chin J Highw Transport, 2024, 37(1): 215-240. (in Chinese)
[10]	Nyholm S, Smids J. The ethics of accident-algorithms for self-driving cars: An applied trolley problem?[J]. Ethi Theor Mora Pract, 2016, 19(5): 1275-1289.
[11]	Thornton S M, Pan S, Erlien S M, Gerdes J C. Incorpora-ting ethical considerations into automated vehicle control[J]. IEEE Trans Intel Transport Syst, 2017, 18(6): 1429-1439.
[12]	胡云峰, 曲婷, 刘俊, 等. 智能汽车人机协同控制的研究现状与展望[J]. 自动化学报, 2019, 45(7): 1261-1280.
	HU Yunfeng, QU Ting, LIU Jun, et al. Research status and prospect of human-machine cooperative control for intelligent vehicles[J]. Acta Auto Sini, 2019, 45(7): 1261-1280. (in Chinese)
[13]	Elli M S, Weast J. Towards a formal model for safe and scalable automated vehicle decision-making: A brief survey on responsibility-sensitive safety[J]. SAE Int’l J Connect Autom Vehi, 2021, 4(1): 9-22.
[14]	黄鹏程, 裴晓飞, 周洪龙, 等. 基于多指标耦合的自动驾驶汽车实时轨迹规划算法[J]. 汽车安全与节能学报, 2023, 14(4): 480-487.
	HUANG Pengcheng, PEI Xiaofei, ZHOU Honglong, et al. Real-time trajectory planning algorithm for autonomous vehicles based on multi-index coupling[J]. J Autom Safe Energ, 2023, 14(4): 480-487. (in Chinese)
[15]	赵亮. 自动驾驶汽车监管制度研究[D]. 长春: 吉林大学, 2019.
	ZHAO Liang. Research on the regulatory system of auto-nomous vehicles[D]. Changchun: Jilin University, 2019. (in Chinese)
[16]	张虹. 自动驾驶汽车侵权责任法律问题研究[D]. 长春: 吉林大学, 2021.
	ZHANG Hong. Research on legal issues of tort liability for autonomous vehicles[D]. Changchun: Jilin University, 2021. (in Chinese)

类型	代表案例	主要争议焦点	处理特点
感知失效型	2016年京港澳高速特斯拉Model S事故；2018年美国亚利桑那州Uber行人致死事故	环境识别是否有效，技术边界提示是否充分	追责制造商、测试主体
规则冲突型	2016年美国加州Mountain View谷歌无人驾驶汽车与公交车相撞案；2023年德国高速公路奔驰自动驾驶事故	复杂场景下目标冲突与风险分配	技术与规则并重
人机协作失效型	2022年杭金衢高速绍兴辖区小鹏P7事故；2025年铜陵小米NOA辅助驾驶事故	提示时机、接管条件、控制权归属是否明确	驾驶人、车企责任并存
高阶运营型	2023年美国加利福尼亚州Cruise Robotaxi事故	系统处置逻辑与运营安全管理是否充分	责任转向系统与运营主体

类型	代表案例	主要争议焦点	处理特点
感知失效型	2016年京港澳高速特斯拉Model S事故；2018年美国亚利桑那州Uber行人致死事故	环境识别是否有效，技术边界提示是否充分	追责制造商、测试主体
规则冲突型	2016年美国加州Mountain View谷歌无人驾驶汽车与公交车相撞案；2023年德国高速公路奔驰自动驾驶事故	复杂场景下目标冲突与风险分配	技术与规则并重
人机协作失效型	2022年杭金衢高速绍兴辖区小鹏P7事故；2025年铜陵小米NOA辅助驾驶事故	提示时机、接管条件、控制权归属是否明确	驾驶人、车企责任并存
高阶运营型	2023年美国加利福尼亚州Cruise Robotaxi事故	系统处置逻辑与运营安全管理是否充分	责任转向系统与运营主体

变量类别	主要内容	分析作用
车辆状态，V	车速、加速度、制动转向角、碰撞时间等	判断避险能力、碰撞风险和稳定性
道路环境，E	道路类型、附着系数、信号标识等	判断场景约束条件和环境修正因素
交通参与者，P	行人与车辆位置、通行权、违法行为、受影响数量	判断风险来源、规则状态与损害后果
人机交互，H	自动驾驶等级、接管请求与响应、提示方式	分析控制权状态和接管响应
数据追溯，T	系统日志、接管记录、事故数据	支撑责任贡献分析与事故过程复核

变量类别	主要内容	分析作用
车辆状态，V	车速、加速度、制动转向角、碰撞时间等	判断避险能力、碰撞风险和稳定性
道路环境，E	道路类型、附着系数、信号标识等	判断场景约束条件和环境修正因素
交通参与者，P	行人与车辆位置、通行权、违法行为、受影响数量	判断风险来源、规则状态与损害后果
人机交互，H	自动驾驶等级、接管请求与响应、提示方式	分析控制权状态和接管响应
数据追溯，T	系统日志、接管记录、事故数据	支撑责任贡献分析与事故过程复核

主车初始速度， v₀ / (km·h^-1)	40~60	路面附着系数，µ	0.7~0.8
行人横穿速度， v_p m / s	1.2~1.5	系统响应时间，t_r / s	0.3~0.8
行人与主车初始纵向距离，d_p / m	15~25	a_x_{, max}/ (m·s^-2)	8
后车与主车初始距离，d_r / m	10~20	a_y_{, max}/ (m·s^-2)	3
侧向车辆与主车纵向距离，d_s / m	8~16	-	-