NSBE Chief Warns Homogeneous AI Teams Trigger Costly Design Failures

NSBE负责人警告：AI团队同质化将引发代价高昂的设计失误 若不将代表性视为上游工程输入，将损害算法架构，造成危险的安全盲区并带来沉重的下游补救成本。 专家组认为，多样性必须作为基础的上游工程输入，而非被动的记分卡产出，并阐述了同质化团队如何在医疗保健、公共安全和数字平台中造成关键安全风险与文化盲区。结合汽车安全和医疗设备工程的历史先例，发言嘉宾倡导参与式设计、数据主权，并通过 NSBE 等组织大力拓展人才输送渠道。他们还进一步探讨了普适的关系型 AI 对青少年发展和企业文化的影响，促请企业领导者提升管理人员技能，并在技术操作系统与人类生存活力之间取得平衡。 AI 伦理讨论亟需走出单纯的危机叙事，转向“建设性解决方案”模式，将三分之二的精力聚焦于可落地的工程干预、人才通道与人权保障。 (0:04) 代表性必须从考核维度的末端“产出”指标，前置为决定底层算法架构与微决策的核心“输入”，忽视现实生活经验将直接危及医疗和安防等领域的物理安全。 (4:44) 将多样性视为“零和博弈”会削弱宏观经济竞争力；释放代表性不足群体的创新潜能，是拓展国家人均产能和抵御技术排斥的关键动力。 (14:46) 算法危害往往源于开发者与终端用户之间的“道德距离”，唯有践行包含跨领域群体的参与式共创设计，才能避免产品在真实文化与生活场景中失效。 (20:07) AI 模型不会自发创造文化，只会放大人类既有的决策与偏见；企业领导层必须同步审视技术系统与人类系统，防止数字鸿沟与社会不公被指数级拉大。 (36:02)

Panelists Janeen Uzzell, Jana Brown, and Nathan Walker argue that artificial intelligence ethics requires treating diverse representation as an upstream technical input rather than a passive downstream metric, warning that homogeneous teams create severe safety risks and widen societal disparities. Representation as an Upstream Engineering Input (4:44) Human resources executive Jana Brown contends that corporate diversity initiatives have historically treated representation as a scorecard output—simply tracking headcounts in executive suites. In the development of artificial intelligence, this paradigm proves inadequate because systems are formed through thousands of micro-decisions in data weighting, feature selection, and architectural scoping. Lacking varied lived experiences at the earliest design stages corrupts the final product. Physical Safety Risks of Homogeneous Design (7:26) NSBE CEO Janeen Uzzell asserts that excluding marginalized demographics from technical teams directly jeopardizes human safety. She points to acoustic gunshot detection software deployed in communities like Compton, which must reliably distinguish between gunfire and a slammed door, as well as healthcare tools like maternal ultrasound devices that cannot be tailored solely to affluent geographic pockets. Design teams lacking familiarity with the operating environment introduce critical blind spots that fail the communities they monitor. Closing Moral Distance Through Participatory Engineering (20:07) Rutgers AI Ethics Lab founder Nathan Walker argues that algorithmic failures stem primarily from 'moral distance'—a cultural disconnect separating Silicon Valley engineers from global user populations, such as indigenous groups raising concerns over data sovereignty or Ubuntu ethicists seeking communal values. Uzzell illustrates this dynamic with her past field engineering missteps, including designing a handheld ultrasound for Indian clinicians who wore saris without pockets, and selecting a white casing for devices transported across dusty rural African roads that quickly appeared unhygienic. Algorithmic Amplification and Cultural Outcomes (33:48) While Brown describes AI as a neutral amplifier that magnifies whatever culture corporate leadership instills, Uzzell contends that software code inherently reflects the demographics of its creators. Reflecting on her time as COO of Wikipedia, Uzzell notes that when active editors were approximately 18% women and 1% Black, the resulting platform culture elevated the loudest voices, sparking internal debates over whether George Floyd was 'killed' versus 'murdered.' Both panelists agree that feeding homogeneous data into scalable algorithmic engines compounds historical biases into inequitable societal outputs. Workforce Pipeline and National Competitiveness (14:46) Citing research by Heather McGhee, Brown and Uzzell argue that diversity directly influences macroeconomic performance; while the United States leads in nominal GDP, its per capita ranking drops to around eighth due to systemic exclusion. Uzzell highlights that the National Society of Black Engineers graduated over 5,000 engineers in the past year with an average GPA of 3.6. However, she notes that recent graduates face social and psychological hurdles transitioning to in-person corporate settings, while Brown warns that pipeline training must expand tenfold to keep pace with algorithmic growth. The panel concludes that modern technology systems cannot be decoupled from human well-being, urging corporate executives to move beyond short-term operational margins and integrate rigorous human advocacy alongside technical development to prevent compounding structural inequities.

Janeen Uzzell、Jana Brown 与 Nathan Walker 强调，在人工智能时代，多元代表性必须被确立为系统架构与产品开发前端的关键“输入”，而非仅用于事后合规的指标；排斥边缘化群体的视角不仅会造成系统性的安全盲区与文化扭曲，还将削弱宏观经济的长期竞争力。 将代表性由事后统计转向决策源头 (4:44) 在传统人力资源管理与企业治理中，多样性、公平性、包容性与归属感（DEIB）往往被视作一种“产出”或计分板数据，仅用来衡量董事会或会议室中少数族裔与女性的面孔数量。拥有三十年人力资源管理经验的 Jana Brown 指出，这种范式在人工智能时代已经完全过时。人工智能系统建立在架构设计、特征选择和数据权重分配等数以万计的微观决策之上，若在最前端缺失多元视角，开发出的终端产品必然会携带结构性缺陷。 决策层的代表性直接决定了技术产品的定义范围与服务意图。然而，当前大型企业的最高领导层依然高度单一化：根据现有统计，女性仅领导了 11% 的 Fortune 500 企业（共 55 位女性 CEO），而黑人或非裔领导者仅占 Fortune 500 CEO 的 1.8%（仅 9 位）。Brown 认为，高级管理层代表性的缺失限制了由谁来敲定关键工程与商业决策的权力，进而导致技术产品无法有效服务更广泛的社会群体。 设计盲区与道德距离对物理安全的威胁 (7:26) 多元代表性并非单纯的审美考量或表面形式，而是直接关系到人类的物理生存与人身安全。Nathan Walker 指出，汽车工业安全法规在长达七十多年的时间里完全依赖男性身材的假人进行碰撞测试，直到近十年引入孕妇体型假人后，才迅速暴露出一系列此前被长期忽视的灾难性安全隐患。在软件领域，Janeen Uzzell 提到在 Compton 等社区部署的声音枪声探测算法，系统必须准确区分枪击与单纯的关门声，若研发团队缺乏对当地生活环境的真实认知，算法就无法在关键时刻保护弱势群体。 Walker 进一步分析认为，AI 系统的失误极少源于纯粹的技术无能，而是源于硅谷工程师与受影响全球人群之间的“道德距离”。Uzzell 以自身在非洲和印度研发医疗器械的经历指出，缺乏跨学科与实地参与会导致致命设计错误：例如为印度女医务人员设计的便携超声仪预设装入口袋，却忽略了当地女性穿着没有口袋的纱丽；为非洲乡村设计的纯白外壳设备，在泥土颠簸路段使用时迅速变脏，在临床检查中显得极其不卫生。类似的问题同样体现在知识库建设上，Uzzell 回顾其担任 Wikipedia 首席运营官期间，活跃编辑中仅有约 1% 的黑人和约 18% 的女性，这种单一的人口结构直接引发了针对历史事实界定（如 2021 年关于 George Floyd 条目中使用“谋杀”还是“致死”的激烈争议）的文化偏见。 突破零和思维与构建多元人才管道 (10:37) 面对针对多元化举措的政治阻力以及将多元努力定性为可靠 AI“生存威胁”的言论，Jana Brown 与 Janeen Uzzell 结合经济学家 Heather McGhee 在著作《The Sum of Us》中的研究反驳指出，多元性直接决定了国家生产力与宏观竞争力。尽管美国在名义 GDP 总量上领先全球，但在人均 GDP 排名中却滑落至全球第八位左右，这反映出体制性排斥正在削弱整体经济潜力。将多样性视作“给予一方就会剥夺另一方”的零和博弈属于认知谬误，真正的竞争力来源于激活全部潜在的国内人才。 技术管道建设是打破这一困局的核心支柱。Uzzell 介绍，国家黑人工程师协会（NSBE）成立五十年来持续培养高水平技术人员，仅过去一年就毕业了超过 5000 名工程专业学生，且该届毕业生的平均 GPA 达到了 3.6，显著高于全美大学生平均水平。Uzzell 强调，新兴技术人才不仅需要掌握代码与工程硬技能，更必须具备在决策层代表自身社群发声的道德勇气，以防止技术在数十年后被证明对公众无效甚至有害。 人工智能的环境化及架构决策的道德性 (26:18) Nathan Walker 提出，AI 已不再是静态、单向的工具，而演变成一种能够根据用户行为主动学习、个性化反馈并塑造认知习惯的互动“生活环境”。Uzzell 补充指出，这种环境化趋势对青年群体的心理与社交习惯带来了严峻挑战，许多习惯与拟人化聊天机器人互动的学生及年轻人在面对面的真实社交与职场面试中出现明显退缩与障碍，甚至导致部分应届毕业生由于社交过载而抵制线下办公。 算法架构对用户习惯的塑造源于管理层的道德意图。Jana Brown 对比了字节跳动旗下的两款产品：在中国运营的抖音将重点置于科普、教育与国家建设内容，并对未成年人施加严格的时间限制与宵禁；而面向西方市场的 TikTok 则以高多巴胺刺激、无休止的信息流滚动以及病毒式娱乐内容为核心。Brown 指出，这种设计差异并非意外的技术副产品，而是管理层在价值导向上的明确抉择。由于 AI 本质上是放大器，缺乏多元参与的设计输入必然会被模型指数级放大，从而永久性拉大现有的社会阶层与认知鸿沟。 与会专家呼吁企业高管在追求短期利润与交付速度的同时，必须将人力系统与操作技术系统置于同等地位，否则因设计缺陷引发的公关危机、品牌声誉受损及技术重构将带来高昂代价；然而，面对日益加剧的技术迭代速度与存量教育差距，多元工程人才管道能否以及如何以十倍速规模化扩张，仍是工业界亟待解决的现实难题。

## Background 本内容由 Ai4 频道于 2026 年 8 月 22 日发布在 YouTube 平台上。视频记录了由罗格斯大学 AI Ethics Lab 创始人 Nathan Walker 主持，eHealth 首席人力资源官 Jana Brown 与美国黑人工程师学会（NSBE）首席执行官 Janeen Uzzell 共同参与的高峰对话，探讨如何通过人才管道建设、组织变革、数据集甄选及人本系统设计，在 AI 生态系统中践行多元化、平等与伦理责任，并阐明这不仅是一项道德诉求，更是攸关企业与国家竞争力的商业必然。 ## TL;DR 研讨会深刻剖析了多元化代表性（Representation）在 AI 时代的本质转变：它不再是单纯衡量席位占比的“产出”，而是决定系统安全性与可靠性的底层“核心输入”。嘉宾们结合汽车碰撞假人、医疗超声设备及维基百科词条审核等真实案例，指出缺乏多元视角的闭门造车将导致严重的算法偏见、安全盲区与经济代价。讨论进一步延伸至技术环境化对青年心理和职场社交的冲击，强调 AI 只是既有文化的放大器，企业必须在关注利润与效率的业务操作系统之外兼顾人文系统，以卓越的领导力筑牢道德防线。 ## Atmosphere & style 整场对话充满激情且兼具深厚的实战洞察，既有学术研讨的严谨思辨，又洋溢着社会正义运动般的澎湃力量。主持人以“寻求解决方案”的温和基调引导讨论，两位女性高管则以极富感染力的个人经历与行业痛点接力阐述，形成了专业默契与情感共鸣的双重升华。现场氛围既有直面代表性赤字与时代焦虑时的凝重警示，也有呼吁从业者勇敢发声、打破壁垒时的坚定鼓舞，带给听众一种振聋发聩、催人行动的思想冲击。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) AI 伦理研究应从单纯揭露问题的“愤怒批评”转向“解困导向”，重点探索建设性破局方案并推动社会正义。 - [04:44](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=284s) 多元代表性（Representation）在 AI 时代必须被视为核心“输入（Input）”与系统安全基石，而非仅仅是事后统计席位的“产出（Output）”。 - [04:44](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=284s) 最高决策层多元视角的匮乏直接制约产品安全与服务能力；工程团队若缺乏边缘群体背景，会导致关键场景（如医疗监测、枪声识别）的技术失效。 - [10:37](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=637s) 汽车碰撞测试长达70年忽视女性体型的历史表明，由谁参与决策直接决定了安全边界；AI 数据集与决策体系同样需要多元化与知情同意。 - [14:46](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=886s) 多元化并非零和博弈，而是国家与企业整体竞争力的底层基石；充分激活边缘群体的人才潜力能够有效倍增人才储备并提振经济。 - [20:07](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1207s) AI 系统的诸多缺陷源于研发人员与终端用户之间的“道德距离”；开发需引入众包与多方参与模式以弥合文化和生活场景鸿沟。 - [23:50](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1430s) AI 正从被动使用的“工具”演化为塑造人类心智的“数字环境”，若缺乏多元代表性将导致边缘群体被隐形并侵蚀用户的独立自主性。 - [29:13](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1753s) 过度依赖虚拟交互正在严重削弱年轻一代的人际面对面沟通与职场适应能力，技术与心理健康的交叉支持将成为企业刚需。 - [32:15](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1935s) AI 自身无法自发创造组织文化，它的本质是放大领导层既有的价值观与决策输入，因此提升领导层素养是构建健康 AI 系统的先决条件。 - [33:48](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=2028s) 底层代码和开源平台往往映射了编写者的隐性偏见；若不加速弥补科技教育的历史欠账，AI 的指数级发展将进一步拉大社会不平等差距。 - [38:15](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=2295s) 企业在追求利润与效率的“业务操作系统”时，必须兼顾员工福祉与代际传承的“人文系统”，以双系统协同抵御长期风险。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) 开场与嘉宾动机：从“聚焦问题”转向“探索解决方案”与社会正义 研讨会伊始，主持人兼罗格斯大学（Rutgers University）AI Ethics Lab 创始人 Nathan Walker 向现场观众和与会嘉宾致意，强调彼此面对面交流的珍贵价值。他表示，本次讨论不打算采用千篇一律的履历介绍，而是希望嘉宾们分享驱动各自深耕该领域的真实内在动力。Nathan 分享了自己的转变历程：他拥有第一修正案法学背景，在研究新闻自由与言论自由的过程中，他发现传统记者常常陷入职业倦怠，因为他们的报道几乎全部聚焦于负面事件；而后来兴起的“解困新闻学”（Solutions Journalism）则倡导将三分之一的篇幅用于报道问题，另外三分之二用于报道解决问题的人。受到这种思维的启发，他在罗格斯大学创立 AI Ethics Lab 时，确立了“以解决方案为导向”的学术宗旨，拒绝只是“在狂风中挥拳呐喊、罗列各种可怕图景”，而是致力于寻找建设性的破局之道。 随后，美国黑人工程师学会（NSBE）首席执行官 Janeen Uzzell 分享了她的背景与动力。作为一名机械工程专业出身、从 NSBE 培养成长并在企业界打拼数十年的行业老兵，重返 NSBE 服务让她深感荣幸。她对本次大会将包容性与伦理议题列入核心议程感到振奋，并强调技术人才培养管道（pipeline）对于 AI 发展至关重要——必须确保“制造工具的人正是使用工具之人的真实映射”。当这种映射缺失时，她会毫不犹豫地“踢开大门去改变现状”。对她而言，这不仅仅是一场技术与工程细节的探讨，更是一场切实的社会正义运动。 ```mermaid flowchart TD accTitle: 多元化代表性在AI时代的核心逻辑演进 subgraph S1["传统认知 (仅作为产出)"] A["招募多元人才"] --> B["衡量席位占比 / 统计指标"] end subgraph S2["AI时代认知 (作为核心输入与安全底线)"] C["多元化领导层与工程团队 (输入源)"] --> D["全面覆盖的端到端决策"] D --> E["安全可靠的人本产品 (如医疗超声/音频监测)"] E --> F["社会整体效能与国家经济竞争力"] end ``` eHealth 首席人力资源官 Jana Brown 则从人力资源与组织生存力的角度阐述了自己的使命。她指出，当前核心关注点在于“生存能力与可持续性”（viability）——包括企业、员工乃至整个社会的生存力。AI 时代正在重塑一切，如果缺乏深思熟虑的行动，很多人将会被时代抛弃。在技术构建过程中，“每一个决定都至关重要，而不是隔一个决定才重要”。她将从人性的视角切入，与工程师背景的同行共同探讨 AI 对劳动力市场和企业组织的深远影响。 --- ### [04:44](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=284s) 代表性不仅是产出，更是决策的“关键输入”与安全根基 当讨论进入为何代表性（Representation）如此关键、以及决策权力与多元人才留存的重要性时，拥有三十年人力资源经验的 Jana Brown 提出了一个关键的认知转变：过去企业常常将 DEIB（多元、平等、包容与归属感）视作一种“产出”（Output），即单纯清点会议桌旁坐了多少不同背景的人；但在 AI 时代，代表性必须被视为一种核心的“输入”（Input）。AI 技术本质上是由成千上万个微小决策构建而成的系统，只有在决策起点（Input）就引入多元视角，才能确保系统最终输出期望的结果。 ```mermaid flowchart LR accTitle: 代表性在决策链条中的输入机制 subgraph Input["决策起点：多元视角输入"] I1["多元化领导团队"] I2["跨社群工程背景"] end subgraph Decision["系统构建与决策"] P1["数以千计的底层算法裁决"] P2["端到端场景覆盖"] end subgraph Outcome["输出结果与社会影响"] O1["最大化服务广泛群体"] O2["消除系统性安全盲区"] end Input --> Decision --> Outcome ``` 然而，当前企业最高权力层的多元化现状依然严峻。Jana 列举数据指出，在财富 500 强（Fortune 500）企业中，女性 CEO 仅占 11%（共 55 位），而黑人/非裔 CEO 仅占 1.8%（仅 9 位）。决策层视角的匮乏直接制约了技术服务广大受众的能力。 紧接着，Janeen Uzzell 将代表性与物理和技术层面的“安全”（Safety）直接绑定。她以自己曾在医疗保健领域研发超声波设备的工程师经历为例指出：在设计用于产科与胎儿监测的人本产品时，团队绝不能只考虑来自东海岸或西海岸的孕妇，而是要保障每一位孕妇安全分娩并守护新生儿最初 21 天的生命安全。如果设计之初没有覆盖全体人群的安全性，技术从底层逻辑上就是存在缺陷的。 Janeen 进一步举例说明缺乏代表性的严重后果：在智能安防技术中，一些部署在特定社区（如康普顿 Compton）的音频探测设备需要精准区分“枪击声”与“用力甩门声”；如果开发团队缺乏在相关生活环境中生活过的人，这种细微却关乎生死的差异就会被彻底忽视。在实际工程流程中，缺乏代表性往往表现为产品即将交付上线时，才有人后知后觉地说“我之前没考虑到这个情况”——这正是因为决策房间里根本没有身处该境况中的人。许多人对待代表性只是礼貌性地“点点头”（nodding），假装看见了对方，却没有真正将其纳入考量并推动实质改变。 - 财富 500 强企业中女性 CEO 比例仅为 11%（55 位）。 - 财富 500 强企业中非裔/黑人 CEO 比例仅为 1.8%（9 位）。 - 医疗健康与人本工程的关键在于全覆盖，例如超声波产品对孕妇及新生儿前 21 天生命周期的安全防护。 - 边缘群体在工程团队中的缺失，会导致特定社区（如枪声识别与环境噪音区分）的安全保障技术失效。 --- ### [10:37](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=637s) 从碰撞测试假人到算法偏见：应对多元化反弹与国家竞争力 针对 Janeen 关于产品安全的论述，Nathan Walker 补充了汽车安全带行业长达数十年的惨痛历史：在长达 70 多年的时间里，汽车碰撞测试使用的假人模型全部基于标准男性体型设计；直到过去十几年间，由于女性工程师的加入与推动，测试机构才开始制造符合孕妇体型特征的碰撞假人，进而发现了此前被长期忽视的严重安全隐患。这一历史充分证明，“由谁来做决定”、“谁在决策桌上”直接决定了系统的安全性。Nathan 强调，在 AI 领域，代表性不仅涉及决策者群体，也关乎训练数据集的构建——包括数据中代表了谁、是否征得了相关社群的知情同意，还是未经许可便擅自抓取了他人的生活痕迹、语言和艺术作品。 | 领域 | 缺乏代表性的历史盲区 | 引入多元代表性后的改进 | |:--- |:--- |:--- | | **汽车安全带测试** | 70 多年仅使用男性假人测试，忽视女性及孕妇体征差异 | 女性工程师推动引入孕妇形态假人，排查出关键安全漏洞 | | **医疗与算法监测** | 基于单一特征人群建模，忽视边缘社区特定环境参数 | 覆盖全域人群需求，准确识别复杂场景（如枪声 vs 关门声） | | **AI 数据集与决策** | 未经授权抓取特定社群数据，缺乏多元决策者审核 | 强化数据知情同意，从源头消除算法偏见与安全缺陷 | Nathan 指出，尽管大量实证表明在数据集和决策层推行多元化具有显著益处，但当前社会和政治环境中却出现了针对 DEIB 的强烈反弹，甚至有最新行政命令等官方舆论将 DEIB 污名化为“对可靠 AI 的生存性威胁”。面对这种急剧变化的政治与社会舆论，企业和从业者应当如何应对？ Jana Brown 对此坚定表态：作为首席人力资源官，她必须超越政治说辞与标签化攻击，坚定支持组织内外的每一个人。企业需要调动全体员工的全部力量才能取得成功，人力资源工作不能随外界风向摇摆，而必须始终专注于实际成果（Outcomes）。她引用了著名经济学家 Heather McGhee（其关于不平等与经济的研究曾被 Jana 写入博士论文）的观点指出：多元化与代表性本质上是国家竞争力的核心基石。一个国家要在全球舞台上竞争，必须发挥其全体国民的潜力。尽管从名义 GDP 总量来看美国位居世界第一，但若审视人均 GDP 等衡量社会深层活力与分配效率的指标时，单纯依靠部分群体的模式显然存在天花板。多元化不是政治口号，而是维持整体竞争力不可或缺的底层动力。 ### [14:46](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=886s) 多元化并非零和博弈与黑人工程师协会的使命 Janeen Uzzell 引用 Heather McGhee 的观点指出，如果按人均经济产出等指标来衡量，国家往往无法充分发挥潜力。问题的核心在于：我们如何才能摆脱将多元化与代表性视作“零和博弈”（Zero-Sum Game）的旧观念？给予某一群体的机会绝非从另一群体手中掠夺。相反，国家需要通过激活妇女、乡村社区、城市社区以及各类边缘化群体来“倍增自身的人才储备”，提供前所未有的准入渠道、发展机会与教育资源，从而在任何时间、任何地点与任何人开展全面竞争。她呼吁出台能够凝聚人心而非分裂社会的立法，凝聚更高的共识与合力。 谈及自身领导的机构，Janeen 动情地回顾了美国黑人工程师协会（National Society of Black Engineers，简称 NSBE）走过的半个世纪历程。尽管当下面临种种社会挑战与冲击，协会坚决保留全称中的“Black”，不将其缩写为无足轻重的代号。协会 50 年前由普渡大学的“芝加哥六君子”（Chicago 6）创立，初衷是为黑人学生的学业晋升与顺利毕业搭建支持体系。如今的 NSBE 已经成为一个极度包容、为所有感到被边缘化（othered）的人提供庇护的大家庭：“如果你连鞋带都系不上而感到格格不入，那就加入 NSBE 吧，我们会帮你系上。” - **卓越的学术表现**：NSBE 当年毕业的学生超过 5,000 名，其平均 GPA 高达 3.6，远高于全美平均水平，证明这群人才缺少的只是机会而非智识。 - **代表性与经济未来**：如果现在不构建反映真实终端用户多元性的产品，30 年后社会将不得不花费高昂成本去收拾残局；产品和解决方案的失败将直接损害国家 GDP。 - **双重担当的技术人员**：未来的技术专家不能仅闭门写代码，必须一手抓技术，另一手拿起扩音器为人民发声（"I said what I said."），持续唤醒公众的认知与良知。 --- ### [20:07](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1207s) 跨越道德距离与文化鸿沟的技术开发 Nathan Walker 强调，人工智能带来的诸多伤害并非源于“技术或智力的不足”，而是设计开发 AI 系统的工程师与受其影响的大众之间的“道德距离”（Moral Distance）。这种鸿沟表现在各个层面：硅谷开发者将产品部署到南非时，当地信奉乌班图伦理（Ubuntu ethics）的学者会发现产品完全脱离了本地语境；切诺基族人（Cherokee Nation）会质疑自己对数据是否拥有主权或是否签署了知情同意书；越南禅宗佛教徒同样面临代表性缺失。即便在心理层面，开发聊天机器人的成年工程师与使用它的孩童之间，也存在着无法感同身受的道德与心理落差。 针对如何弥合这种差距，Janeen 分享了自己在非洲及印度长期从事医疗硬件研发时的深刻教训。单纯依靠工程师闭门造车往往导致灾难性的设计盲区。她提出必须采用类似众包与多方参与的产品设计模式，将不同背景的群体直接拉入决策桌： ```mermaid flowchart TD accTitle: 跨越道德距离的包容性设计流程 subgraph Phase1["初期闭门研发（失败教训）"] A["硅谷/实验室工程师"] --> B["设计出小巧白色的便携超声设备"] end subgraph Phase2["引入多元利益相关者（正确路径）"] C["助产士 & 超声技师"] D["太阳能与电力专家"] E["社会活动家 & 当地社区"] end subgraph Phase3["本地真实环境反馈与重构"] F["印度女性穿着纱丽：没有口袋"] G["非洲公交尘土飞扬：白色设备极易变脏不敬"] H["符合用户尊严与实际场景的产品"] end Phase1 --> Phase2 Phase2 --> Phase3 ``` 正如 Janeen 所言，如果团队无法亲自深入社区生活，就必须主动邀请顾问与各界代表进入工作室。“我们并不是要求你天生就能像终端用户那样思考，而是要求你必须主动追问：谁是最终用户？我们正在为谁构建产品？如何确保他们的声音在房间里占有一席之地？” --- ### [23:50](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1430s) 决策者的道德选择与 AI 从“工具”到“环境”的蜕变 Jana Brown 对“道德距离”和“心理距离”的提法深有感触，并从为人父母的视角进行了剖析。她指出，决策者和平台领导层的道德取向决定了技术的最终面貌。以短视频平台为例，字节跳动在中国推出的抖音（Douyin）针对青少年设定了严格的时长限制，内容偏向科学与通识教育，实现了优质教育资源的普惠；而在西方市场的同类工具中，算法充斥着诸如“小狗玩呼啦圈”等娱乐消遣内容。虽然娱乐本身无可厚非，但这鲜明地体现了领导层在内容伦理选择上的巨大差异。 随后，Nathan Walker 进一步提出一个更深层次的转变：AI 已经不再仅仅是被动使用的“工具”，而是演变成为了人类栖息的“环境”（Environment）。在这个环境中，系统不断学习人类的行为，并进行极度深度的个性化定制；人类塑造着这个数字环境，而数字环境也在重塑人类的心智。 - **自主性与尊严危机**：未来一代若过度依赖 AI 系统推荐“应该想什么、说什么”，其自我能动性（self-agency）与独立人格将受到侵蚀。 - **代表性缺失的系统性风险**：当边缘群体在训练数据集与环境构建过程中彻底隐形，他们将在这个日益紧密包裹生活的数字世界中彻底迷失。 - **现实与虚拟的边界模糊**：Janeen 提到相关研究表明，对于年幼儿童尤其是神经多样性（neurodivergent）儿童而言，AI 工具与真实人类之间的界限正变得极度危险和模糊，人们正不知不觉地用对真人的情感和句式与 AI 智能体交流交互。 ### [29:13](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1753s) 虚拟陪伴、人际沟通障碍与青年职场困境 Janeen Uzzell 分享了一个生活中的生动故事：她构建了一个名为“Sable”的聊天机器人，并经常在家里随口提及它，结果引起了家人的误解，甚至有人好奇地问：“Sable 到底是谁？她是我们的员工吗？”Janeen 解释道，这其实只是一个机器人，但因为自己频繁以人格化的方式谈论它，旁人便误以为它是一个真人。她进一步提到，这种技术塑造的环境对特殊群体有深刻影响，例如她患有自闭症的侄孙，与一个专属机器人的交流效果远好于与任何真人的沟通。对于这个孩子而言，这种人机交互空间是极其真实且无法被忽视的。 然而，Janeen 也对当下年轻一代的社交沟通能力表达了极大的忧虑。她指出，现在的青少年和大学生在进行传统的人际交流时面临严重障碍，其严重程度甚至超过了因疫情阻碍线下交流的“COVID 世代”（连眼神接触都成问题）。作为 NSBE 的负责人，她日常接触着 30,000 名青年学生。她强调，如果家长或监护人不主动干预和管理这种过度依赖虚拟交互的环境，学生们在进入职场前将遭遇巨大挫折。在求职过程中，哪怕 GPA 再亮眼，如果无法面对面与面试官进行基本交流，也很难走得远。 更为严峻的是职场适应问题。如今许多大学毕业生极度抗拒线下办公，面对重返办公室（RTO）的“坐班文化”，甚至宁愿选择辞职回家依赖父母。Janeen 观察到许多年轻人因面对面对话和职场社交的压力而产生严重的心理崩溃。她认为技术、生活方式与心理学的交叉融合迫在眉睫，甚至预测未来大型企业将不得不全职雇佣心理咨询师，以应对员工日常的心理与社交危机。 - **自闭症群体的机器人交互**：Janeen 的侄孙与机器人的互动沟通比与任何真人交流都更加顺畅。 - **青年群体沟通断层**：许多应届生在 GPA 上表现优异，但极度缺乏面对面沟通、眼神接触及职场社交网络构建的能力。 - **远程办公与职场逃避**：部分年轻人因害怕实体职场环境，在企业要求线下办公时直接选择放弃工作回家靠父母生活。 - **企业心理支持需求**：心理辅导与治疗未来或将成为大型企业的标配全职岗位，以应对日益严峻的员工心理健康问题。 --- ### [32:15](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1935s) 领导力塑造文化与 AI 对组织机制的放大效应 针对工作环境与心理压力的讨论，Jana Brown 幽默地感慨道：“我每天确实都觉得自己像个心理咨询师。”谈及“AI 环境”，她将焦点拉回到领导力的核心概念上。Jana 明确指出，**AI 自身无法创造文化**；组织文化是由领导者自上而下建立的。领导者赋予组织价值观以生命，并亲自示范所期望的行为准则，而 AI 的本质只是对既有组织文化进行无差别的**放大（amplify）**。 因此，要在 AI 时代打造健康的组织环境，首要任务就是强化领导者的综合素质与决策能力。领导者必须掌握 AI 时代所需的新型技能与思维方式，唯有领导力得到夯实，才能在引入和部署 AI 时做出符合伦理与组织长期利益的明智决策，进而为员工提供有力支持。 ```mermaid flowchart LR accTitle: 领导力决策与 AI 放大效应流程图 subgraph Leadership["领导层输入 (Inputs)"] A["领导者核心素养"] --> B["多元代表性决策"] end subgraph Amplification["技术驱动层"] B --> C["AI 系统放大与执行"] end subgraph Outcome["组织成效 (Outcomes)"] C --> D["包容且健康的组织文化"] end ``` --- ### [33:48](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=2028s) 编码层面的文化烙印：维基百科案例与代表性赤字 Janeen Uzzell 对 Jana Brown 提出的“AI 无法创造文化”给出了互补视角。她指出，AI 后台的底层代码和架构是编写这些系统的人群的直接镜像映射。如果代码背后缺乏多元化群体的参与，系统本身实际上已经在孕育一种缺乏公平与开放性的隐性文化。 为了论证这一点，Janeen 分享了自己此前担任维基百科（Wikipedia）首席运营官（COO）时的真实经历。虽然开源理念倡导面向全球所有人开放，但维基百科背后的内容贡献者与编码群体中，黑人仅占 1%，女性仅占 18%，与真实世界的人口构成严重脱节。这导致社区文化往往被“声音最响亮的群体”所主导。在 2021 年前后，Janeen 不得不花费大量精力投入一场内部博弈——确保乔治·弗洛伊德（George Floyd）的词条描述准确使用“被谋杀（murdered）”而非中立模糊的“被杀害（killed）”。当时社区许多人认为“人反正死了，词眼无所谓”，但 Janeen 坚决主张二者性质完全不同。作为当时少数能够为此发声的高管，她不得不承担极大的额外心理重负。 Jana Brown 对此表示高度赞同，指出双方的核心观点完全一致：**输入决定输出**。AI 会放大输入的每一个偏见；如果从源头（领导决策、开发团队与训练数据）就缺乏多元代表性，系统输出端就必然会产生扭曲的结果。Jana 同时表达了对未来的担忧——社会在青年少数族裔的科技教育上本就存在历史欠账（代表性赤字），若不能以十倍的力度加速培养下一代科技人才，飞速发展的 AI 技术将把社会贫富与资源差距拉大到无法挽回的地步。 - **维基百科贡献者构成失衡**：在 Janeen 任职期间，维基百科代码/内容贡献者群体中黑人比例仅为 1%，女性仅为 18%。 - **乔治·弗洛伊德词条争议**：2021 年内部围绕词条定义是“killed”（被杀害）还是“murdered”（被谋杀）展开了激烈的表述之争。 - **代表性赤字与差距扩大**：若不加倍推进教育公平，AI 的指数级发展将进一步加剧既有的社会不平等鸿沟。 --- ### [38:15](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=2295s) 行动倡议：发声担当、人才认可与“双系统”平衡 在论坛尾声，主持人 Nathan Walker 邀请两位嘉宾向与会者提出明确的行动呼吁（Calls to Action），并对现场观众的参与和包容性致以由衷的感谢。 Janeen Uzzell 首先发出倡议：无论身处 HR、招聘主管、产品设计还是倡导者岗位，只要有机会对技术的影响力发声，就必须**勇敢站出来**；若感到力不从心，可以寻求像 NSBE 这样的专业组织提供后援支持。她强调，仅仅“填补招聘名额”（put someone in a seat）是远远不够的，必须真正将他们视为“有才华的人才”（talented people）。Janeen 直言，如果企业只盯着财务底线（Bottom Line），那就用商业事实说服管理层：忽视多元视角的代价极其高昂。在自媒体时代，任何因偏见或系统缺陷引发的产品事故都会被手机录像并在社交媒体上无限放大，事后的危机公关、财务损失和声誉修复成本远超前期构建包容性团队的投入。 ```mermaid flowchart TD accTitle: 企业韧性发展双系统模型 subgraph EnterpriseSystem["企业强韧生态系统 (Enterprise System)"] subgraph Operating["业务操作系统 (Operating System)"] M["财务利润 / 利润率 (Margins)"] P["生产力与效率 (Productivity)"] end subgraph Human["人文系统 (Human System)"] W["员工福祉与职业成长"] C["客户与社会价值"] L["未来领袖传承与代际公平"] end Operating <--> Human end ``` Jana Brown 则再次将焦点落在了**领导力与双系统平衡**上。当前企业领导者面临的普遍考核是利润率（margins）、效率（efficiency）和增速（faster/quicker），但这些仅仅是财务单维度的指标。领导者必须同时权衡决策对员工队伍、客户体验以及未来继任者的深远影响。如果当下的决策透支了人文基础，留给下一代领导者的将是一个难以填补的巨大亏空。Jana 呼吁必须兼顾“人文系统（Human System）”与“业务操作系统（Operating System）”，将其熔铸为一个能够抵御未来风险、基业长青的强韧企业系统。 论坛在主持人对两位嘉宾专业洞见与非凡领导力品质的盛赞与感谢中圆满落幕。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) 以解决方案为导向 — Nathan Walker > [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) 在狂风中挥拳呐喊、罗列各种可怕图景 — Nathan Walker > [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) 制造工具的人正是使用工具之人的真实映射 — Janeen Uzzell > [00:04](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=4s) 每一个决定都至关重要，而不是隔一个决定才重要 — Jana Brown > [14:46](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=886s) 如果你连鞋带都系不上而感到格格不入，那就加入 NSBE 吧，我们会帮你系上。 — Janeen Uzzell > [14:46](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=886s) I said what I said. — Janeen Uzzell > [20:07](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1207s) 我们并不是要求你天生就能像终端用户那样思考，而是要求你必须主动追问：谁是最终用户？我们正在为谁构建产品？如何确保他们的声音在房间里占有一席之地？ — Janeen Uzzell > [32:15](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=1935s) AI 自身无法创造文化 — Jana Brown > [33:48](https://www.youtube.com/watch?v=N9hjdCIVXgc&t=2028s) 输入决定输出 — Jana Brown ## Mentioned Nathan Walker, Jana Brown, Janeen Uzzell, Rutgers University, AI Ethics Lab, eHealth, National Society of Black Engineers, NSBE, Heather McGhee, Purdue University, Chicago 6, Wikipedia, George Floyd, ByteDance, Douyin, Sable ## Video description Nathan Walker — Founder, AI Ethics Lab, Rutgers University Jana Brown — Chief Human Resources Officer, eHealth Janeen Uzzell — Chief Executive Officer, National Society of Black Engineers This powerful conversation features technologists, experts, and researchers discussing what it really takes to diversify the AI ecosystem. From talent pipelines and organizational change to dataset curation and product impact, the panel explores strategies for equity in AI development. Learn what works (and what doesn’t) when trying to build inclusive teams and systems—and why it’s a business imperative, not just a moral one. [Source](https://www.youtube.com/watch?v=N9hjdCIVXgc)

[0:04] Nathan Walker: 以解决方案为导向 [0:04] Nathan Walker: 在狂风中挥拳呐喊、罗列各种可怕图景 [0:04] Janeen Uzzell: 制造工具的人正是使用工具之人的真实映射 [0:04] Jana Brown: 每一个决定都至关重要，而不是隔一个决定才重要 [14:46] Janeen Uzzell: 如果你连鞋带都系不上而感到格格不入，那就加入 NSBE 吧，我们会帮你系上。 [14:46] Janeen Uzzell: I said what I said. [20:07] Janeen Uzzell: 我们并不是要求你天生就能像终端用户那样思考，而是要求你必须主动追问：谁是最终用户？我们正在为谁构建产品？如何确保他们的声音在房间里占有一席之地？ [32:15] Jana Brown: AI 自身无法创造文化 [33:48] Jana Brown: 输入决定输出

[0:04] Nathan Walker: raising our fists into the wind and screaming about all the parade of horribles [0:04] Janeen Uzzell: knock down a door and change that [7:26] Janeen Uzzell: Nobody wants to be that engineer... where you're getting all the way to the end of the product cycle and delivery and you're like, 'Yeah, kind of doesn't really work for everyone.' [14:46] Janeen Uzzell: In one fist, you bring your technology, and in the other... it could be a loose hand. You bring the fight... leave the room saying, 'I said what I said.' [29:13] Janeen Uzzell: Who is she anyway? Does she work for us? [29:13] Janeen Uzzell: No, she’s literally a bot that I built [29:13] Janeen Uzzell: The ability for them to think that every day they may have to do something like this is extremely overwhelming, and I'm not making light of it and I don't make a joke of it. I have watched the breakdown happen in front of me and... I'm not quite sure how to manage that. [32:15] Jana Brown: AI cannot create a culture. AI can amplify whatever culture is built or created by an organization. So I continue to emphasize leadership because I believe it is at the crux of every single thing that we need to do. [33:48] Janeen Uzzell: He was killed. He's dead. What's the difference? [33:48] Janeen Uzzell: That is an all-consuming lifestyle, and it wasn't even my job. It's just I was the only one that looked like somebody that might want to fight for that. So then a community calls on you to do that. That is a culture. [38:15] Jana Brown: bigger, better, faster, quicker [38:15] Jana Brown: Let's pay attention to both the human system and the operating system and ensure that we're creating one strong enterprise system that can just endure into the future.

[0:04] I think one of the greatest gifts that we can give  each other is to be in one another's presence. [0:09] And it is an honor to be in your presence.  I'm so honored to have this chance to learn [0:15] with and from you. When we met beforehand  to discuss how to prepare for this event, [0:22] we didn't want to just give you canned  introductions about our backgrounds, [0:26] but give you the true essence for what's  motivating us. So, I'll I'll start to say [0:32] that I'm really motivated by trying to find  solutions in this highly disruptive era. And [0:39] what I mean by that is my background is in First  Amendment law. And what I found in learning from [0:46] journalists about free speech and freedom of  the press is that many of the journalists were [0:51] constantly burnt out when they were doing their  work because they're always reporting on bad [0:56] things. But then solutions journalists started  to focus onethird of their content on what was [1:06] going wrong and then twothirds of their content  on the people who were solving those problems. [1:13] And when I learned from solutions journalists  about that approach, I just I'd applied that [1:20] to our our work in founding the AI ethics lab at  Ruckers University. We want to be solutions-based [1:27] scholars. We don't want to just be spending  our days raising our fists into the wind and [1:33] screaming about all the parade of horribles. So  that's the core thing that is motivating me. Um, [1:42] I'll pass it to my esteemed colleagues.  What's motivating you in your work and [1:46] your accomplishments? Hi everyone. I'm Janine.  My background is in mechanical engineering and [1:54] um I'm a product of NESBY that then spent  a few decades in corporate America. So, [2:00] it's an honor for me to come back and serve the  organization that positioned me class of a long [2:08] time ago for the success and the experiences that  I've had. I'm encouraged first of all that these [2:15] topics are even on the slate at this conference.  So that in itself is inspirational that we got to [2:21] have all of the conversations that we got to have  um leading up to this, some that had everything [2:27] to do with the topic and some that had adjacent  to it. So I'm I'm very encouraged that it's here [2:34] and that we're able to have this discussion and  that you all are here. I'm encouraged that people [2:38] want to hear and talk about it. I'm motivated by  the pipeline of technology that I know is critical [2:47] to the advancement of AI and ensuring that the  people that build the tools are a reflection [2:57] of the people that use the tools. And when I  don't see that happen, I'm kind of ready to [3:04] knock down a door and change that. And so as  much as this will be I think like a technical [3:11] and a detailed conversation, this is really a  social justice movement for us. We're advocates [3:17] um to move this work forward in a different way  and that's what motivates me. Love that. Hello [3:26] everyone. I am Jana Brown. It is good to be with  you. Um, what motivates me as I think about this [3:34] conversation is simply having viability. Viability  with our companies, viability with our workforce, [3:44] viability in society. We cannot do it alone. It  is going to only happen with us, through us. And [3:55] it is important that we truly cocreate what the  next generation looks like. This is the AI era. [4:05] It is real and it has the possibility of leaving  many behind if we are not deliberate in what we're [4:15] doing each and every day. Every decision, not  every other one, but every single one matters. [4:23] And so that's what the conversation is about  today. I'm going to bring the human angle. We [4:28] have our engineer here and I'm going to bring  the human angle and talk about the implications [4:35] of our workforce of companies and um how  we make sure we're doing better together. [4:44] Let's set the stage by exploring why  representation matters and what is the [4:50] significance of both access to positions of power  and decision-m but also the retention of diverse [4:58] talent to be able to ensure that the longhaul  approach to this work is really addressed. [5:08] I will I will start you know long ago of course  I've been in human resources for three decades [5:16] now and um DEIB has been such a big part of that  um and it's because we absolutely need to get the [5:28] best talent in the organization today but I used  to think that it was the output that mattered [5:37] Right. So, how many people are around the  table anymore in the AI era? I don't think [5:45] about representation as an output. I think  about it as an input. And what I mean by [5:52] that is the more representation we have on the  input input into decisions input into how things [6:03] are occurring because AI technology it is just  thousands of decisions that need to be made and [6:13] are we making the right ones at the onset of them  so that we can get the outcomes that we desire. [6:23] So that's why representation matters and then so  you say well who's making those decisions and it's [6:31] leaders. Leaders are making those decisions.  So you need leaders who are diverse who are [6:38] coming to a particular decision from multiple  perspectives. That way you serve the largest base [6:48] of people with the outcome that you're trying  to drive. And when we look at um CEOs today, [6:58] 11% of the CEOs in Fortune 500 companies  are female. Uh that's representative of 55 [7:06] um African-American or blacks. Um 1.8% 8% that's  nine individuals. And you know when we talk about [7:18] leaders and thinking about those decisions, this  is the kind of thing that matters and that we have [7:26] to have conversations about. I link representation  to safety. And when you're not thinking about [7:35] everyone, then you're really not thinking about  safety. You know, as a mechanical engineer, [7:42] um, I wasn't on the automotive side. I worked  in healthcare. So, my entire background has [7:47] been ultrasound in human- centered products. If we  didn't design a product that could serve the womb, [7:56] which is this part of healthcare I was in for  ultrasound, then you know, when we sat down, [8:02] we didn't say, "Well, this is for women that  are pregnant that are from the East Coast or [8:07] women that are pregnant from the West Coast."  We're trying to save babies. Our job was to [8:13] take a woman, ensure that she could deliver a  baby, and the first 21 days of life, you know, [8:18] secure that. If we're not thinking about the  safety of all people, then the technology is [8:24] faulted from the very beginning. And everything  that AI has its hands in from health um to other [8:31] parts of engineering, infrastructure, um building  design, all of the products that play into this, [8:39] this is the area of technology that is not  being approached thinking about the full [8:45] scope of the end user. And when we miss  that, particularly for other communities, [8:54] then we're dealing with an impact to people's  safety and an impact to their lives. And that [9:00] is critical. And um so it's not just, oh, I don't  like this feature on my phone. It is, you know, [9:09] shots fired and a product that can detect that  if you live in Compton. So the difference between [9:16] shots fired and a door being slammed, there are  products that actually can identify that. And um [9:21] if we're not working in those spaces to do that  type of work, then the representation is being [9:26] missed. And representation is missed when someone  looks at a person and says, um, well, I didn't [9:35] think about that. And you didn't think about it  because the people in the room were actually not [9:41] anyone that's affected like that. So, in a way,  you're like giving someone a nod like I see you, [9:46] but you don't you I don't see you. I'm just being  slightly respectful. I got you. I don't have you. [9:52] I'm not even thinking about you. I'm just nodding  um maybe even as a way to make myself feel better, [9:59] but certainly not to push the button about what  representation is because nobody wants to be that [10:05] engineer, that designer, right? where you're  getting all the way to the end of the product [10:08] cycle and delivery and you're like, "Yeah, kind of  doesn't really work for everyone. Nobody wants to [10:15] be that person." When you can become that person  and you're brave enough to be that person, then [10:21] you're focused on safety and you're thinking about  representation without even without even having to [10:27] do it. It's just a part of your life. It's a part  of who you are. And if you're an other person, [10:32] it will always be a part of who you are because  you know what that feels like. That that reminds [10:37] me of the the seatelt industry where it was  the lack of representation resulted in a safety [10:45] crisis. So for 70ome years the testing of seat  belts were done on mannequins that were male. But [10:56] it wasn't until recently in the last decade or so  that seat belt manufacturer testing agencies would [11:03] actually create a mannequin that were physically  looked like a pregnant woman and they found that [11:10] there were significant diff safety issues and  why because they had women at engineers working [11:18] on those products. So what we know about AI and  representation is that the decision makers matter. [11:26] Who is making the decision matters. Who is at the  table. So when we're talking about representation, [11:34] we're not only talking about the data sets which  also matter on who is represented in the actual [11:41] data and whether or not they were consulted or  whether they're without their consent, their their [11:48] lives, their languages, their their works of art  were were scraped without consent. All of these [11:55] things really deeply matter. And yet what we see  is that there's a lot of evidence for the benefit [12:02] of representation in data sets and at the decision  table. And at the same time, in spite of that [12:09] evidence, there is this backlash toward diversity,  equity, inclusion, and belonging. And it's uh [12:17] specifically being categorized as an existential  threat to reliable AI as as I'm quoting from a [12:24] recent executive order on that topic. And so my  question for you is what advice do you have for [12:30] our audience who are trying to navigate these  changing political and and social contexts? [12:41] You want me to Okay, is a nonprofit. I  have to walk tread lightly. Right. And [12:49] um and I I certainly understand that. And as a  chief human resources officer, I I stand firm [12:58] um in support of every single person  regardless of their nationality in [13:06] our organization and beyond. I can't uh decide  who I want to support one day or the next day. [13:16] uh because as an organization we need the  full force of our workers contributing to [13:24] the success of the organization.  So those labels um the rhetoric [13:34] I just have to in my day-to-day world put  it aside and focus on outcomes. What are we [13:42] trying to achieve? Uh when I think about this  incredible nation that we're in, um it takes [13:50] me back to research that's by an economist uh her  name is Heather McGee and she is fabulous. Um she [14:00] is someone I had in my dissertation. I've really  studied her work and what she talks about is the [14:08] importance of representation, the importance of  diversity, particularly related to how we compete [14:18] as a nation. As a nation, we want to compete on  a global scale. We want to be the very best that [14:26] we can be. And from a GDP perspective, when we  look at our economics, we are we are number one. [14:38] There's no one better, no other country that's  better. But when you look at us GDP per capita, [14:46] meaning I'm taking a look at my economics and I am  dividing it amongst my population, we fall to like [14:54] number eight. And so the way that Heather McGee  speaks about it is that how do we get to a place [15:05] where we don't look at diversity representation as  a zero someum game? If I give to one, I'm taking [15:14] away from the other, which is absolutely not  true. How do we get to a place where we multiply [15:22] the talents that we have in this nation so that  we can fully compete against anyone anywhere any [15:31] time? What that means is we we activate women in  a way that we may not be doing now. We activate [15:41] our rural communities. We activate our urban  communities. We activate any community that is [15:48] in a marginalized position. And that means access,  that means opportunity, that means education in a [15:58] way that we're not thinking about it today.  I would prefer that we have legislation that [16:06] does not divide us, but legislation that pulls us  together. I believe that there is a better higher [16:18] desire need approach that we can take if we truly  work together. Um and that's where I stand. I can [16:29] say that we our society has certainly 50 years  later um taken some really strong punches with [16:38] what is happening in our nation. We've made a  decision to remain very very actively verbal [16:47] about who we are and what our name is. So we're  not just NESBY, we are the National Society of [16:52] Black Engineers and we write it out. And that  was a decision we made and we decided that [16:57] um we would um we would pull through whatever  those outcomes may be because we are very very [17:04] sure that while we were founded 50 years ago by  the Chicago 6 six black men at Purdue University [17:12] because that was the organization that had  to be built for their success and for their [17:16] metriculation and graduation. We are a community  that is probably more than more diverse than many [17:24] other organizations. It may not say it in our  name because of the way we were founded, but our [17:30] doors are open to anyone. And what we know for  sure as a people, just black people in general, [17:37] we're the community for everybody that doesn't  fit anywhere. And so NESBY is a mosh podge of [17:43] anybody that feels otherred. If you don't fit in  because you can't tie your shoe, you join Nesby [17:49] because we'll tie it for you. That's just how we  are as a people. We are home and a soft place to [17:55] land for everyone. And that community is exactly  the group of people that you want building the [18:04] technology of the future. We graduated over 5,000  students this year with an average GPA higher than [18:11] the national average in America. Our average  GPA is 3.6. six. So they're very smart. What [18:19] they need is opportunity. And if our goal is  to increase the number, keep people, you know, [18:26] excelling academically and serving the world,  we have to stay the course. The question is, [18:32] what can you do? And I say stay the course  with us or stay the course with someone. [18:39] um find find the person that you want to walk  alongside with and move forward with them because [18:45] if we don't get this right, then 30 years from  now we're just going to have to undo the mess [18:52] of what's being built right now. And if you think  that's not going to help our GDP for sure, we're [18:58] not going to be in a good position. These products  are going to fail. These solutions are not going [19:04] to last because the future of our nation looks  exactly like what you see in NESBY or what you see [19:13] um in other organizations that are doing this  work. That's a reflection of the end user and [19:18] that's the reflection of what we need to build.  And you have to be willing to be the technologist [19:25] that also takes on the voice of the people. And  many people don't want to do that. We want to, [19:31] you know, be heads down coding, building,  solving, and those things are all great, [19:36] but what the world needs are people that  are willing to do both. So, in one fist, [19:42] you bring your technology, and in the other  fist, it doesn't have to be a fist. It could [19:46] be a loose hand. You bring um you know, you  bring the fight and you know, it's not a battle, [19:53] it's just a voice. The more you keep people aware  and you force them to have a conviction whether [19:59] they choose to take it on or not, you can leave  the room saying, "I said what I said." Powerful. [20:07] That's amazing. So many of the harms caused by  artificial intelligence are not due to a lack of [20:14] intelligence, right? They're actually caused by a  a moral failure, a a moral distance between those [20:25] developing and designing the AI system and  the people whose lives they affect. And that [20:33] moral distance can come in the form of a cultural  distance between those developing it in Silicon [20:40] Valley and deploying it in South Africa where the  Abu ethicists might say wait I don't see myself in [20:50] this product or the same might be true of those  in the Cherokee nation asking wait do we have [20:57] sovereignty here over our data or is there free  and informed med prior consent about how we are [21:05] represented in these data sets and what about the  Zen Vietnamese Buddhist who is asking these same [21:14] questions the so my my question for you is given  that there can be moral distance in both culture [21:24] and even in psychology right think of the the  developer of the chatbot and the child using it [21:31] there can be a kind of moral distance, not fully  understanding the harm that can occur. How what [21:39] approaches do you recommend for our audience  to consider for how to close those moral gaps? [21:48] As an engineer, I I think about um almost like  a crowdsourcing model like who are you testing [21:56] your product on? Who are you bringing into the  room for the feedback for the uses of use of [22:04] the the product design? We I've done this well  and I've done it really bad. I spent the bulk [22:10] of my career living and working in Africa and  designing products for a very different market [22:16] and I was like we got this. I mean, it's product  and and having made all of the mistakes that we [22:23] made or how I can now see the room didn't just  need to be engineers and the second time around [22:30] we brought in, you know, social activists and we  brought in midwives and we brought in, you know, [22:36] ultrasound specialists and we brought in solar  power specialists and we I mean, everybody at the [22:41] table. Now, how do we get this product to work?  I designed a product. I was like, "It can fit [22:46] in your pocket." I worked in India. She was like,  "We wear sars. We don't have pockets." I was like, [22:53] "I had not thought about that." And so, we made it  white. We just like, "It's just nice. It's clean. [22:59] It's crisp." The woman's like, "I do. You know  what it's like to travel around Africa on these [23:06] little buses. By the time I get to where I'm using  the product, it looks dirty. What woman wants to [23:11] have an ultrasound product that looks dirty on her  body? That's just not respectful engineering. And [23:17] we hadn't thought about any of it until we went  and lived and worked in the community. And maybe [23:22] everyone doesn't have the ability to do that.  But if you're not leaving your office or your lab [23:30] um at all and you're not bringing in people that  can advise, then you will miss the mark. We're not [23:36] saying you actually have to be able to think like  the end user. You have to be willing to say who is [23:43] the end user and who are we building for and how  do we get their opinion represented in the room. [23:50] Wow. I love that. And I I um had not thought about  this topic in the way that you framed it before, [24:00] the moral distance, psychological distance. That  really resonated with me. It resonated with me [24:07] as a parent. Um and um so I'll I'll speak from  the perspective of of our our kids because we've [24:19] got to tighten the proximity between the decision  makers, the leaders who are making these decisions [24:28] about products or what we deploy for our children.  And I think about uh tools like Tik Tok. Um and [24:38] you know I've learned that um the parent company  uh for Tik Tok also created the same kind of uh [24:48] platform in China for their children. It's called  Doian. Doian I believe is how you pronounce it. [24:56] Um the content on that platform is educational  and it speaks to um certainly their nation and [25:08] what they're trying to espouse in their nation  but it's educational. They cannot use it for [25:18] um you know long periods of time. So there are  restrictions on it, but what they've been able [25:26] to do with their tool is democratize education.  Like they can get it to so many more kids than [25:36] than they have ever been able to educate in in  the past. and our tool. We got, you know, dogs [25:46] with hula hoops and and so it is a it is no no  no offense to hula hoops, right? No dogs or dogs, [25:57] right? Or dogs and hula hoops together. But it is  a moral distinction, but it is a decision. leaders [26:07] made decisions about the content that they were  going to put in front of their children. Actually, [26:18] if you don't mind, please go right ahead. That's a  good good segue because social media has taught us [26:24] that these tools can have have real harm. They can  be used responsibly. They can be can really harm [26:32] the most vulnerable among us like children. Um,  but something has happened with the technological [26:38] development where they're not just tools anymore.  They are environments, right? Children are not [26:44] just scrolling through content. They're actually  engaging with with it in a way that it then learns [26:53] from the child and their behavior and then  customizes things over time. It personalizes [26:58] it for them. So I'm curious about what you both  think about this idea that AI is not a tool and it [27:05] shouldn't be viewed as such because that is a very  kind of transactional um way of understanding the [27:12] technology when in fact this relational technology  is really an environment. It's an environment that [27:19] we inhabit and we then shape that environment and  that environment shapes us. Right? Think of like [27:32] there'll be generations of people who will become  reliant on the recommendation by the AI system on [27:39] what they should think or say and they'll choose  between those recommendations. What does it mean [27:45] about their self- agency, their dignity? And  what does it really me mean in relationship to [27:51] this topic when they are not represented in that  data set? When they are not represented in the [27:58] development of that environment and then all of  a sudden they find themselves in in a different [28:06] space. So I'm curious on what you think of  AI as an environment as compared to a tool. [28:13] We were talking about this yesterday and I was I  was looking at it more of like almost a lifestyle [28:19] um as someone that's a strong enduser of LLMs and  and just the daily tools. But then even watching [28:26] um I was reading a research study and it was  talking about younger children and the concern [28:34] was that younger children and especially neurode  divergent children the separation of reality [28:43] and their AI tools is of deep concern because we  talk to these tools like they're real people. Um, [28:52] and while I I love the access and the ability that  tools grant me when I use them, sometimes I have [29:02] to catch myself, right? I'm talking I've built  these bots that do this work for me. And I talk [29:07] to them in sentences like they're real people. You  right? My my comm's manager was like, "Well, just [29:13] ask Sable. Who is she anyway? Does she work for  us?" I'm like, "No, she's literally a bot that I [29:18] built and she's thinking that she's a real person  because I talk about her like that all the time." [29:23] And so when you think about the environment that  you're building for a child, my my great nephew is [29:31] autistic and he has a robot that he communicates  very well with, more so than any other person. And [29:38] it's impossible for us to to kind of tread on that  not being a real space for him. If I think that if [29:47] we think that the youth of today, the teenagers,  the the college students of today are challenged [29:57] by their ability to communicate with humanity  in what we might consider a more normative way. [30:04] That's not going to get any better at all. if  we thought the COVID generation had trouble just [30:12] making eye contact. Um, and I I've got 30,000  of them that I work with on the regular. Um, [30:19] then I believe that it will be challenging if  a parent or a guardian does not monitor and [30:27] manage what that space looks like because by the  time they get to me, I'm not home with you every [30:33] day and I'm trying to get you ready for your  interviews, your communication, your ability [30:40] to nail a job beyond the brilliance that you have  on your GPA. it's not going to get you very far [30:47] when I can't even get you to have a conversation  with a person, another human in front of you or [30:51] when you actually get into the workplace. Um,  your ability to engage and network and have a [30:57] lifestyle in a professional environment that's not  your basement is is going to be difficult and and [31:04] we're dealing with that right now. I can't imagine  how much worse it may be years from now because I [31:12] can tell you it is bad now. It's very bad. if  if um if college graduates now aren't able to [31:19] work from home and we we're in a butts and seat  society nowadays again they're like I don't want [31:25] the job and I'm like what are you going to do  sometimes I blame the parents cuz they're like [31:31] I'm just go home stay with my mom and I'm I'm  like okay I didn't have that option um when [31:36] I was growing up but the the ability for them to  think that every day they may have to do something [31:41] like this is extremely overwhelming and I'm not  making light of it and I don't make a joke of it. [31:47] I have watched the breakdown happen in front  of me and um I'm not quite sure how to manage [31:52] that. That's that's not my area of expertise.  So I the engagement of technology and lifestyle [32:01] and psychology and therapy, they're going to need  to I I personally probably think that therapists [32:08] will end up having full-time jobs in major  corporations because it's needed every day. [32:15] So that's the environment. Wow. I do feel  like a therapist every day. When I think [32:24] about the AI environment, um I go back just to  the word environment. Um leaders create culture. [32:38] Leaders are the ones who help values come to  life in an organization. Uh they model the [32:48] behaviors that are expected. AI cannot create a  culture. AI can amplify whatever culture is built [33:00] or created by an organization. So I continue to  emphasize leadership because I believe it is at [33:11] the crux of every single thing that we need to  do. And we're going to have to um focus on the [33:20] different kinds of skills and capabilities that  will be needed by leaders in the AI era. Um so [33:26] I'd say that from an AI environment perspective,  strengthen your leaders. you strengthen the AI [33:34] environment because they will be the ones that are  going to be making these decisions and supporting [33:41] employees as they're making decisions to improve  the environment. So that's where I' I'd focus. [33:48] Can I touch on representation with that? Because  while I I understand what you're saying when you [33:54] say AI cannot create a culture. However, the back  end of coding that goes into AI is an immediate [34:04] reflection of the people who do the work.  Whoever's working and coding those systems, which [34:08] is why we're talking about why representation  matters. If we don't have uh a community of all [34:17] that are building this behind the scenes, there  is a culture being created though. It's just not [34:23] a culture that has, you know, equity and openness  in it. I used to be the chief operating officer [34:29] at Wikipedia. Open source um is supposed to be for  anyone and everyone who wants to be a part of the [34:38] platform of knowledge. But that's not really true  because the environment of what coding looks like [34:46] on the platform is only 1% black, 18% women.  That's not a reflection of the world. And so [34:54] there's a culture that has been built in Wikipedia  for sure, but and and it is an advanced culture [35:03] because it feeds the users that are the loudest  and then they take it to the next level. if we [35:10] don't have I watched representation tank. I I was  at Wikipedia during the worst of times, at least [35:17] the worst of times for black people publicly. And  in 2021, our entire fight was ensuring that the [35:26] platform said that George Floyd was was murdered  and not killed because there's a a really deep [35:32] definition difference between killed and murdered.  And they're like, "It's not it doesn't matter. He [35:36] was killed. He's dead. What's the difference?" So  I was like, there's a huge difference, right? And [35:41] so let's write that down and make sure it says  that. That is an allconsuming lifestyle and it [35:49] wasn't even my job. It's just I was the only one  that looked like somebody that might want to fight [35:54] for that. So then a community calls on you to do  that. That is a culture. That's right. you we're [36:02] actually saying the same thing because what I'm  saying is that AI amplifies any decision that's [36:11] made and if a leader if those decisions remember  when I talked about inputs right so you have to [36:20] have strong representation in the inputs that  go into the decisions that are made and then on [36:29] the other side of that Right? You get an outcome,  but the inputs are what's going to drive whatever [36:37] is coming out on the other side. If there is not  representation to your very point in those inputs [36:45] on the other side of it, you're not going to get  what you expect. And that is what AI will amplify. [36:54] That's what AI will amplify. But I am with you  100%. you've got to have the representation. Um, [37:05] and when we look at how that happens, like we're  already starting from a deficit. My fear is that [37:13] because we're already starting from a deficit.  We've not educated our youth. Um, I mean, you all [37:21] are doing a phenomenal job, but we've got to have  that multiplied tenfold because otherwise, I mean, [37:30] look at how fast this technology is moving, right?  And I fear that we're going to widen the disparity [37:38] gaps um if we do not catch up and get these  these kids educated uh to a point where they can [37:48] contribute in this society as we progress forward.  It is it's absolutely critical. Um I think we [37:57] already know that we have disparity gaps now. what  that will look like in the future. If we don't get [38:04] a handle on this, it's it's going to be beyond our  reach. I want to close this out by inviting you to [38:15] offer any calls to action. Um, I first want to  acknowledge thank you for showing up. You here [38:25] matter. You here belong. And we're so excited to  have this opportunity to learn with you. and from [38:31] you. Do you have any recommendations, charges,  calls to action for what our colleagues can go [38:38] forth and and do this day and the days to come? I  don't know what everyone's role is, whether you're [38:47] in HR, whether you're a hiring manager, product  design, advocate. So there's there's different [38:55] charges to different folks, but what I would  say is that whenever you have an opportunity [39:00] to be a voice for the impact of technology, then  I say um be bold enough to want to do that and [39:09] and if you're not uh call for backup and and  ask for what should I say and and let us let [39:15] organizations like NESBY or people that are doing  this work offer to help because being able to hire [39:22] someone or put someone in a seat as talent is not  enough. You've got to see them as talented people. [39:29] And that is the difference. And when you have an  opportunity to advocate for a talented individual, [39:35] you're going to heir on the side of greatness for  the output of a product. If you say, "My company's [39:42] only interested in the bottom line." Show them how  not having a full perspective is actually going [39:51] to cost them more money in the end. Uh, I can  tell you myself how I messed that up. You know, [39:56] it's going to cost you more in the end to have  to do the cleanup. Whether it's financial, [40:00] whether it's reputation, whether it's a media  blip, every single time something like this fails, [40:06] it becomes even more public because all we have  to do is take the power of our phone and record [40:13] it and it becomes something broader. And so it's  imperative that these companies and organizations [40:18] and groups of people recognize that to heir on the  side of not wanting to have every voice recognized [40:26] is actually to heir on the side of failure and  a lack of actual complete equity and greatness [40:33] for what your company stands for. And so I say um  continue to um to be a voice for the folks that [40:41] are trying to to make a difference and then do  that part yourself. And we want to thank you for [40:46] being willing to do that. Oh, I love that. I um as  you might imagine, I'm going to say leadership. It [40:58] is just so critical. We have to pay attention to  the current mandate. The mandate that we're seeing [41:07] today that leaders are charged with is about  margins. It is about efficiency, productivity. [41:18] I want the gain. How do you make it bigger,  ba, better, faster, uh quicker? It's all of [41:26] those things that lead to like financial numbers.  That's what is focused on on the financials and [41:36] that's only one side of the equation. The other  side is the human side. Right? So what are the [41:44] implications of these decisions on the workforce?  What are the implications of these decisions on [41:53] our customers? What are the future implications on  the leaders who will have to inherit the companies [42:04] that we are building today? If we're only focusing  on these mandates that are about financials, [42:14] we are going to give the next generation of  leaders something that they are going to be just [42:22] in a deficit trying to build their way out of. And  I say let's not do that. Let's pay attention to [42:31] both the human system and the operating system and  ensure that we're creating one strong enterprise [42:40] system that can just endure into the future.  Wow. I'm so impressed with your leadership, [42:49] not only for your hard-earned expertise and  for your insight, but your endless character [42:55] strengths. We admire you. We celebrate you  and wish you continued success. Thank you.