[论文解读] Understanding the Social Cascading of Geekspeak and the Upshots for Social Cognitive Systems

Michał B. Paradowski, Łukasz Jonak|arXiv (Cornell University)|Nov 29, 2011

Language and cultural evolution参考文献 19被引用 4

一句话总结

本文通过一个微博平台研究了新词（'极客语'）如何在社交网络中传播，运用复杂性科学建模级联信息扩散。研究揭示了人类社交认知中的暴露阈值与系统性动态，为从多个社交源而非二元互动中学习的人工认知系统提供了设计原则。

ABSTRACT

Barring swarm robotics, a substantial share of current machine-human and machine-machine learning and interaction mechanisms are being developed and fed by results of agent-based computer simulations, game-theoretic models, or robotic experiments based on a dyadic communication pattern. Yet, in real life, humans no less frequently communicate in groups, and gain knowledge and take decisions basing on information cumulatively gleaned from more than one single source. These properties should be taken into consideration in the design of autonomous artificial cognitive systems construed to interact with learn from more than one contact or 'neighbour'. To this end, significant practical import can be gleaned from research applying strict science methodology to human and social phenomena, e.g. to discovery of realistic creativity potential spans, or the 'exposure thresholds' after which new information could be accepted by a cognitive agent. The results will be presented of a project analysing the social propagation of neologisms in a microblogging service. From local, low-level interactions and information flows between agents inventing and imitating discrete lexemes we aim to describe the processes of the emergence of more global systemic order and dynamics, using the latest methods of complexity science. Whether in order to mimic them, or to 'enhance' them, parameters gleaned from complexity science approaches to humans' social and humanistic behaviour should subsequently be incorporated as points of reference in the field of robotics and human-machine interaction.

研究动机与目标

• 理解新颖语言表达（'极客语'）如何在社交网络中传播。
• 识别人类在社会级联中信息采纳的系统性模式与阈值。
• 从人类社交认知中提取参数，以改进人工认知系统。
• 弥合人类社交动态的洞见与多智能体学习系统的设计。

提出的方法

• 分析真实微博数据，追踪自创词汇的传播。
• 应用复杂性科学方法，从局部互动中建模涌现的全局动态。
• 使用基于代理的模拟，复制并研究信息扩散模式。
• 测量暴露阈值——认知代理接受新信息的临界点。
• 将研究发现整合到机器人与人工智能的社会认知系统建模中。
• 将观察到的社会级联与集体行为的理论模型进行比较。

实验结果

研究问题

• RQ1 在何种临界阈值下，新语言创新会获得社会认可？
• RQ2 局部互动如何引发信息扩散中的大规模系统性模式？
• RQ3 多个信息源在塑造认知代理决策中发挥何种作用？
• RQ4 社会级联动态如何指导社会智能人工认知系统的设计？
• RQ5 从人类社交认知中可测量的哪些参数可增强多智能体环境中的机器学习？

主要发现

• 新词通过社交网络以级联模式传播，其根源在于局部的低水平互动。
• 存在一个临界的'暴露阈值'，超过该阈值后，新信息更可能被认知代理采纳。
• 极客语的传播遵循与复杂性科学中集体行为模型一致的系统性动态。
• 多个信息源显著影响新颖语言结构的接受度与传播。
• 微博中的社会级联模式与真实世界认知过程相吻合，为人工系统提供了可测试的参数。
• 研究结果表明，多源信息整合对于设计社会适应性认知代理至关重要。