SGER: Multiple Agents and Distributed Language Dynamics

这是一项为期12个月的SGER奖。异构性是构建分布式(例如多代理)信息系统以及一般通信中的一个主要问题。为了获得对询问的有效答复或发展富有成效的持续信息互动，参与者必须在各自的信息和活动的本地表示之间建立和保持全球对应关系。实现这一点的主要方法(表示和交换的标准，以及与公共基础国际语言之间的翻译)取决于预先确定的语言和表示和交换信息的规则。更重要的是，支撑这些方法的理论依赖于这些结构和标准的先验存在。因此，大多数努力都花在了设计标准和规则上，而不是质疑这些方法作为一个整体的有效性，或者寻找一个真正新的、更灵活的替代方案。这个项目研究一种新的、探索性的、动态的方法，基于交互语言的发展。其想法是扭转上述主导图景，并将交互用作在信息世界之间创建新的、动态的桥梁的工具。最终目的是为词汇、句法、语法、语义甚至更高层次的社会结构，如组织形式和文化(被视为信息处理媒介)在分布式交互中的共同进化提供可靠的理论和技术。新的方法将改进对开放性的处理，并将详细说明语言演变的机制和限制，因此我们知道如何在有限的资源或时间内弥合异质性。这项研究将a)研究人类协作环境中的语言生成/进化，b)调查语言进化的极限和机制的计算理论，c)假设有界的机制来解释观察到的和理论上可能的过程，d)建立初步的计算模型作为概念的证据，并探索在全自动环境中有用的实际机制。

This is a twelve month SGER award. Heterogeneity is a major issue in building distributed (e.g. multi-agent) information systems, and in communication in general. To get effective answers to queries or to develop a productive ongoing information-based interaction, participants have to build and maintain global correspondences between their respective local representations of information and activities. The dominant approaches to this (standards for representation and interchange, and translation to and from common base interlingua) depend on pre-established languages and rules for representing information and exchanging it. More importantly, theories that underpin these approaches depend on the prior existence of these structures and standards. Consequently, most effort goes into designing the standards and rules, rather than questioning the validity of these approaches as a whole or looking for a truly new and more flexible alternative. This project investigates a new, exploratory, dynamic approach, based on the evolution of interaction languages. The idea is to reverse the dominant picture described above, and to use interaction as the vehicle for creating emergent, dynamic bridges among information worlds. The eventual aim is to provide robust theories and techniques for the co-evolution of lexicons, syntaxes, grammars, semantics and eventually even higher-order social structures such as organizational forms and cultures (treated as information processing media) in distributed interaction. The new approach will improve the handling of openness, and will detail mechanisms and limits of language evolution, so we know just how heterogeneity will be bridgeable with limited resources or time. The research will a) investigate language generation/evolution in human collaborative settings, b) investigate computational theories of the limits and mechanisms of language evolution, c) hypothesize bounded mechanisms to explain the observed and theoretically possible processes, and d) build preliminary computational models as proofs of concept and to explore actual mechanisms that can be useful in fully-automated contexts.