Self-Organising Multi-Agent Systems

自组织多代理系统

• 批准号: 2767542
• 金额: --
• 依托单位: Imperial College London
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2767542
• 关键词: Self; Organising; Multi; Agent; Systems

The wicked societal problems that arise from the transition to the 'Digital Age' present the challenge to engineer increasingly complex socio-technical systems to combat them. These systems must unite human and computational intelligence, and rely on accurate, analogue-world information about its users for establishing the accessibility and ownership of information. Many existing systems are custom owned, yet non-customisable, meaning that users must wait for system updates from the developers, instead of being able to actively make the changes themselves. Furthermore, the information that users supply is stored by the system owner, which prevents users from having autonomy over their own data. It is therefore important that a generative system is engineered, allowing for users to personalise and customise their experience at will, with an architecture that allows for the local storage of data. We envision a system with an architecture based around local server instantiation (with a local database) for data autonomy, and the ability to download various plug-ins for customisability. A meta-platform provides a repository for the server code and set of plug-ins which can be downloaded onto a variety of media, such as a Raspberry Pi, mobile phone, or home tablet. This allows for users to choose where to host the server, by purchasing a relevant piece of technology, and download the appropriate plug-ins for their experience. Users may also invite others to their network to create an environment, where multiple users share a platform with an owner to interact with their plug-ins.This variety of plug-ins affords generativity, as it empowers the userbase to use the software to create innovative platforms which were not imagined by the original platform's designers. The meta-platform repository is both open source and non-restrictive as any users may upload any plug-in for use on any platform. We offer three examples of plug-ins that can be useful for a sound socio-technical system: firstly, a plug-in that is able to codify deep social knowledge for example by creating digital governance. Secondly, we consider a plug-in for a lightweight blockchain alternative - one where an 'economy' of favours/tokens can be tracked between users. Thirdly, we envision a locally hosted machine learning (ML) plug-in, where the output is for the benefit of the user, not the owner, as any data embedding is obscured from everyone but the user themselves.

从过渡到“数字时代”所产生的邪恶的社会问题提出了挑战，工程师越来越复杂的社会技术系统来解决这些问题。这些系统必须将人类智能和计算智能结合起来，并依赖于关于其用户的准确的模拟世界信息，以建立信息的可访问性和所有权。许多现有的系统是自定义的，但不可定制，这意味着用户必须等待开发人员的系统更新，而不是能够主动进行更改。此外，用户提供的信息由系统所有者存储，这阻止了用户对自己的数据拥有自主权。因此，重要的是设计一个生成系统，允许用户随意个性化和定制他们的体验，并具有允许本地存储数据的架构。我们设想了一个系统，其架构基于本地服务器实例化（使用本地数据库）以实现数据自治，并能够下载各种插件以实现可定制性。元平台为服务器代码和一组插件提供了一个存储库，这些插件可以下载到各种媒体上，例如Raspberry Pi、移动的电话或家用平板电脑。这允许用户通过购买相关的技术来选择在哪里托管服务器，并下载适当的插件以获得他们的体验。用户也可以邀请其他人加入他们的网络来创建一个环境，在这个环境中，多个用户与一个所有者共享一个平台，与他们的插件进行交互。这种插件的多样性提供了生成性，因为它使用户群能够使用软件来创建原始平台设计者没有想到的创新平台。元平台存储库是开源和非限制性的，因为任何用户都可以上传任何插件以在任何平台上使用。我们提供了三个对健全的社会技术系统有用的插件的例子：首先，一个能够通过创建数字治理来编纂深层社会知识的插件。其次，我们考虑一个轻量级区块链替代方案的插件-其中可以在用户之间跟踪优惠/令牌的“经济”。第三，我们设想了一个本地托管的机器学习（ML）插件，其输出是为了用户的利益，而不是所有者的利益，因为任何数据嵌入都对除了用户自己之外的所有人都是模糊的。