Unbiased-AI: removing bias from AI systems to facilitate more effective moderation of hate speeches.

无偏见的人工智能：消除人工智能系统中的偏见，以促进更有效地调节仇恨言论。

• 批准号: 10076166
• 金额: $ 6.35万
• 依托单位: CREATIVE INNOVATIONS HUB LIMITED
• 依托单位国家: 英国
• 项目类别: Grant for R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10076166
• 关键词: Unbiased; AI; removing; bias; systems

Unbiased-AI is an AI-based tool that can analyse digital content and detect hate speech across text, image and video formats.It would be used by those operating online websites that allow User Generated Content (UGC), such as forums, social media, games, reviews, and any site allowing commenting or other user content. Operators would be able to scan content before it is posted, thus preventing the publication of hate speech.The operator would be able to customise the tool according to the type of content they want to prevent and the level of risk they want to accept (e.g. they may choose to impose more restrictions for users who are under 18 than for adults). Unbiased-AI can automatically block content according to the operator's requirements, or can flag certain content for human moderation before posting.

Unbiased-AI 是一种基于人工智能的工具，可以分析数字内容并检测文本、图像和视频格式的仇恨言论。它将由那些允许用户生成内容 (UGC) 的在线网站（例如论坛、社交媒体）运营者使用、游戏、评论以及任何允许评论或其他用户内容的网站。运营商将能够在发布内容之前对其进行扫描，从而防止仇恨言论的发布。运营商将能够根据他们想要阻止的内容类型和他们想要接受的风险级别来定制工具（例如，他们可能会选择对 18 岁以下的用户施加比成人更多的限制）。 Unbiased-AI可以根据运营商的要求自动阻止内容，或者可以在发布之前标记某些内容以供人工审核。