Core Viewpoint - The recent move by Huolala to disclose its AI safety and accountability algorithms represents a significant step towards transparency in the platform economy, addressing the long-standing "algorithm dilemma" faced by new employment form workers, particularly truck drivers [1][4]. Group 1: AI and Employment - The digital economy now accounts for over 40% of GDP, with more than 80 million new employment form workers, making the use of AI a matter of "who it serves" rather than "whether to use it" [4]. - For platforms, AI serves as a tool for efficiency and risk management, while for drivers, it should act as a partner for safety and fairness [4][5]. Group 2: Safety and Accountability - Safety and accountability are critical areas for AI optimization, as fatigue driving and hazardous material transport pose significant risks for truck drivers [5]. - A survey by the China Logistics and Purchasing Federation revealed that 44.34% of drivers desire more humane fatigue driving management, and 28.13% often face demands for overloading [5]. Group 3: Huolala's AI Practices - Huolala's AI practices address these pain points by implementing fatigue reminders after four hours of continuous driving and alerting drivers about hazardous materials in real-time [5][7]. - The platform has shifted from a one-sided algorithm notification system to a collaborative approach, involving drivers in the development of AI accountability measures [7]. Group 4: Technological Evolution - The transformation from a "control logic" to a "coexistence logic" in platform economics is crucial for maturity, as it integrates drivers' voices into the technological framework [7][9]. - The goal is to turn drivers' needs for safety and fair accountability into identifiable features within AI systems and collaborative mechanisms [9][10].

据货拉拉介绍，目前，平台的AI安全防控已覆盖从用户下单至司机运输等各个环节，可实时监测订单 运输情况。用户下单时，平台算法会检测用户是否存在违规行为，如识别到存在高危场景，则阻止用户 下单，并做出相应安全提醒。在司机接单运输及完单过程中，平台也会通过图片、语音等算法综合分析 平台运营数据，识别本次运输货物是否合规，是否存在违规载人等，并基于违规类型和风险严重程度分 别采取"强制取消订单""给司机发送弹窗或语音提醒，引导司机无责取消订单""给司机/用户发送安全中 心提醒"等不同干预动作。 同时，算法也会实时监测司机驾驶时长，司机连续驾驶4小时后，每小时给司机发送一次疲劳驾驶弹窗 提醒，直至司机休息。货拉拉平台数据显示，AI全面应用后，危险品运输和违规载人的日均风险单量 下降了30%，日均强制取消违规订单超千单，超限识别平均准确率超80%。在司机安全教育上，近一 年，平台累计推送安全课程43节，超1700万人次完成学习，同时通过新就业形态劳动者职业伤害保险为 司机提供兜底保障，试点以来累计投入1.75亿元，保障8.9亿单。 "AI可以在第一时间快速、准确地处理海量数据，通过算法模型精准识别出危险品运输等违规行为， ...

人机协同判责，公正效率"两手抓" 9月，货拉拉公开行为分取消判责原则时提到，平台是以"判责必须充分取证，证据不足一律无责"为原则，通过人机结合的方式科学判责。那么机器是如何 判责的？机器判责是否真的能保证公平？本次货拉拉的算法公开围绕AI判责进行了专门的解读。 据货拉拉运营副总监靳威介绍，订单取消后，平台AI判责算法会自动获取订单基本信息、订单行驶轨迹等各类数据，并提取其中的货物状态描述、订单备 注等关键信息，结合平台规则进行分析，初步判定责任归属，在这个过程当中，算法可高效、全面收集相关数据，客观公正地做出分析。 11月4日，继9月公开行为分取消判责原则后，货拉拉再次公开平台AI安全防控及判责算法，聚焦于提升货运司机的安全保障与判责体验。据悉，AI安全防 控系统由货拉拉技术团队历时四年自主研发，通过感知、检测和自动化处置的能力，可实时监测订单是否存在危险品运输、违规载人、疲劳驾驶和货物超限 等情况，并做出强制取消订单、司机端弹窗提醒等干预处理。 订单取消后，平台AI判责系统算法则会自动采集各类数据，并从海量数据（603138）中提取其中的关键信息，结合平台规则进行分析，初步判定责任归 属，如果是违规订单取消 ...