在技术哲学的视野下，人工技术物不再是单纯的客体，而是逐渐展现出道德能动性的特征，尤其是伴随人工道德主体概念的提出，智能机器的道德地位与道德主体性正在不断的被重新审视，自动驾驶汽车作为现阶段大规模投入使用的人工智能体之一，其道德主体性问题不容忽视。作为对人类驾驶的替代方案，自动驾驶汽车必然面临对诸多道德场景的处理，也因此被期望具有自主的道德决策能力，以实现更为高效、安全的人工智能驾驶。

自动驾驶汽车如何能够准确地识别出道德场景？如何能够自主地做出道德决策？这些关键性的技术问题成为了对自动驾驶道德决策伦理问题分析的出发点。作为自动驾驶系统运行的核心——感知系统与视觉算法在其中扮演了至关重要的角色，它不仅决定了自动驾驶系统的工作能力，更是整个自动驾驶驾驶过程安全性的重要保障。视觉算法作为自动驾驶技术的核心组成部分，在车辆感知、环境识别和决策制定等方面发挥着关键作用，更是自动驾驶得以完成道德情景识别、决策、操作执行的基础性技术。也正是由于自动驾驶所展现出的道德决策与行动的自主性，自动驾驶技术不仅改变了传统交通模式，更是引发了人车关系与交通实践中的重要变革。一方面，自动驾驶系统的使用令传统人类驾驶员的角色与义务发生了根本性转变，另一方面，由于自动驾驶汽车所追求的高自主性，自动驾驶汽车从传统的交通工具转变为具有积极行动能力的交通参与者。

但基于视觉算法的道德决策逻辑本质上是单一的，算法所要求的对道德事实的数据化与结构化需求使得道德场景趋于扁平化，由此产生了更多的伦理风险。更为关键的是，自动驾驶汽车的道德地位也因此产生了重要变动，车辆从传统的技术工具转向了具有道德能动性的主体。在此背景下，如何理解自动驾驶道德决策的性质，如何界定自动驾驶汽车的道德主体性，以及如何回应可能产生的道德归责疑难就成为了至关重要的问题。

From the perspective of technology philosophy, artificial technological objects are no longer purely objects but gradually exhibiting the characteristics of moral agency, especially with the introduction of the concept of artificial moral agents. The moral status and moral subjectivity of intelligent machines are constantly being re-examined, and the issue of moral subjectivity of autonomous vehicles, as one of the AI entities being used extensively at this stage, cannot be ignored. As an alternative to human driving, autonomous vehicles inevitably face the handling of many moral scenarios and are therefore expected to have independent moral decision-making capabilities to achieve more efficient and safe artificial intelligent driving.

How can autonomous vehicles accurately identify moral scenarios? How can they make moral decisions independently? These crucial technical issues serve as the starting point for analyzing the ethical issues of moral decision-making in autonomous driving. As the core of the autonomous driving system, the perception system and visual algorithms play a crucial role. They not only determine the working capabilities of the autonomous driving system but also provide important safeguards for the safety of the entire autonomous driving process. As a core component of autonomous driving technology, visual algorithms play a crucial role in vehicle perception, environmental recognition, and decision-making, and are the fundamental technology for autonomous driving to complete moral scenario recognition, decision-making, and operation execution. It is precisely because of the autonomy of moral decision-making and actions exhibited by autonomous driving that autonomous driving technology has not only changed traditional transportation modes but also triggered significant changes in the relationship between humans and vehicles as well as in transportation practices. On the one hand, the use of autonomous driving systems has fundamentally transformed the roles and obligations of traditional human drivers. On the other hand, due to the high autonomy pursued by autonomous vehicles, they have transformed from traditional transportation tools into active participants in traffic with the ability to take action.

However, the moral decision-making logic based on visual algorithms is essentially unitary, and the data-driven and structured demands of the algorithm on moral facts tend to flatten moral scenarios, thereby generating more ethical risks. More crucially, the moral status of autonomous vehicles has undergone fundamental changes, transforming from traditional technological tools to subjects with moral agency. Against this backdrop, how to understand the nature of moral decision-making in autonomous driving, how to define the moral subjectivity of autonomous vehicles, and how to respond to possible moral attribution difficulties have become crucial issues.