Human oversight studies
Experiments test whether people reviewing AI outputs catch errors or defer to the system, and how interface design changes that.
GCRI · 人工智能与自主系统
Social and behavioural science studies how people use, trust and are affected by AI. In this field it supplies surveys, experiments and qualitative research on 人工监督, public attitudes, labour effects and the experience of people subject to algorithmic decisions.
Social and behavioural science studies how individuals, groups and institutions think, act and organise. It explains vulnerability, 风险认知, compliance and collective response to shocks. In artificial intelligence and data and compute, social science identifies who is exposed and less able to cope, and why warnings are or are not acted upon. It informs equitable design of policies and interventions across interconnected risks. The canonical description is on STEM Commons.
Experiments test whether people reviewing AI outputs catch errors or defer to the system, and how interface design changes that.
Interviews and participatory research with people subject to algorithmic decisions document their experience and what accountability they need.
Surveys and field studies measure how workers use AI tools and how tasks, productivity and job quality change.
团队遵循该方法时会使用的界面:其指标、状态与决策。名称为示例,数值仅作示意。
AI and autonomous systems · Social science of AI · Run R-1790
稳健标准误
| 登记册中的模型 | 0.28 | 0.09 | [−0.1, 0.3] | 是 |
| 发布时通过的评估 | −0.04 | 0.02 | [−0.3, 0.2] | 否 |
| 打开人工智能事件 | 0.36 | 0.04 | [0.0, 0.7] | 是 |
| 需要批准的智能体行动 | 0.09 | 0.02 | [−0.1, 0.5] | 否 |
Datasets and records used by social and behavioural science on this platform, with the role each plays and the label its values carry. Sources are listed in 来源登记册.
| 数据 | 角色 | 标签 |
|---|---|---|
| Experimental data | Oversight behaviour | 观测 |
| Interview records | Affected experience | 已报告 |
| Worker surveys | 采用 | 已报告 |
模型、其假设、数据和不确定性,以科学家或工程师会核查的形式呈现。结构遵循已发表的方法。所取数值仅为示意,不是研究发现。
The model structure used for Social science of AI, the assumptions it makes and how each is tested.
Ri = f(Hi, Ei, Vi, Ci)危害、风险暴露、脆弱性和应对能力EAL = ∑k pk · Lk对年概率为 pkxt+1 = A xt + B ut + εtA carries the couplings, u the shock, ε the error| 假设 | 为何制作 | 如何检验 |
|---|---|---|
| 在一次事件中,危害与暴露相互独立 | 使损失积分保持可计算 | 与基于联合记录拟合的 copula 进行比较 |
| 耦合在工作点附近是线性的 | 让冲击通过 A 传播 | 在线性范围之外施加扰动并比较 |
| 脆弱性曲线可以在不同地点之间迁移使用 | 很少有本地曲线可用 | 在有观测损失的地方据此进行验证 |
Findings are tested through replication, triangulation of methods and pre-registered designs. Measurement instruments are checked for reliability and validity.
| 核查 | 如何开展 |
|---|---|
| Ethics review | Research with participants passes ethics review with informed consent |
| Pre-registration | Experimental hypotheses and analyses are registered in advance |
| Representative sampling | Surveys state their sampling frame and weighting |
Behaviour is context dependent and may change under stress. Self-reported data carry recall and social desirability biases.
Process explorer · Process
10 steps·4 个阶段
流程
PRISMA 2020 是系统综述的报告指南;各步骤依循其核查清单,而非强制顺序,且没有任何 Nexus 文件采纳它。
报告Outline of an externally published reporting guideline, not a Nexus process and not a record of operation. PRISMA 2020 mainly covers reviews of intervention effects. No Nexus page names it. Every programme is at service level R0.
社会与人类发展 stewards social and behavioural science for the Consortium and reviews how it is applied across platforms.
Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
切勿发送个人数据、敏感基础设施细节或您无权分享的任何内容。贡献在使用前会经过审查;发送贡献不赋予署名权或编辑权,赞助也绝不能换取研究结论。
本网站上的地图、数据和图表仅作为背景和一般信息展示。它们不是建议、预测或预警;信息的局限见免责声明和数据说明。
This site uses only strictly necessary cookies to work and to remember your privacy choices. Nothing loads from other companies, and we never sell or share your personal information.
This site uses only strictly necessary cookies to work and to remember your privacy choices. Nothing loads from other companies, and we never sell or share your personal information. We run no advertising. You can review your choices at any time through Privacy choices at the foot of every page.