Page 37 - 《中国图书馆学报》2025年第1期

035
裘江南 高双燕 黄榕江 李梦洁:AIGC 背景下人智认知导向的信息检索范式———基于世界 3 理论的探究

leveraging sense-making and cognitive-developmental theories the study deeply analyzes the mechanisms

underlying the bidirectional cognitive construction process between humans and retrieval systems.

Specifically human cognitive construction involves the continuous improvement of knowledge structures

during the retrieval process. Conversely the retrieval systems cognitive construction entails continuous

generation and refinement of its knowledge structures through internal programmatic processing powered by

large language models. Finally the study examines the paradigms impact on traditional information retrieval

processes from the perspectives of information need acquisition cognitive construction processes and

information behaviors. It identifies challenges within this new paradigm and proposes directions for

improvement to enhance the capabilities of next-generation information retrieval system.

In summary the study makes significant contributions to the field of information retrieval. First the

proposed human-intelligence cognitive-oriented paradigm enriches existing research and opens new avenues

for the development of future paradigms. Second the exploration of the bidirectional cognitive construction

process between humans and retrieval systems extends the application of theories such as Poppers three

worlds sense-making and cognitive-developmental theories in the information retrieval domain. The study

offers valuable references for both future research and practical implementations of information retrieval

system. Future studies should consider integrating techniques and theories such as retrieval enhancement

technologies semantic networks and cognitive psychology to address the challenges within the new

paradigm. 3 figs. 2 tabs. 41 refs.

KEY WORDS

AIGC. Information retrieval. Three worlds theory. Cognitive construction. Large language model.

0 引言

随着 AI 算法、算力与算据 的不断突破 [1] ,人工智能生成内容(AI-Generated Content,AIGC)正逐

①
渐成为数智时代信息资源管理的重要影响因素 [2,3] 。 信息检索的发展开始受到 AIGC 技术的推动。

2023 年,美国微软公司推出新一代生成式检索系统 New Bing,它继承了原有的搜索引擎 Bing,并融合
[4] ,能够理解、学习和模仿人类的语言模式与人互动,并根据与人类互动的上下
了大语言模型 GPT-4
文来产出匹配人类需求的数字内容,响应多样化的信息检索需求。 Nature 评价此类生成式智能搜索

将会为传统检索系统的发展带来新的机遇和挑战 [5] 。
信息检索系统有三个基本要素:人、系统、信息,围绕不同的要素开展研究,就形成了不同的研究
范式 [6] 。 信息检索范式也可分为三种:系统导向、用户导向和用户认知导向 [7] ,信息检索系统的研究
重点相应从关注系统功能设计到用户信息需求再到用户认知建构。 传统的认知建构视角下,用户认
知导向的信息检索围绕人这一核心要素,强调人通过主动检索完成对自身知识结构的不断丰富和调
整。 而 AIGC 技术驱动下的生成式检索系统发展成为信息检索系统的高阶形式,正在加速信息检索
的研究重心转向“人”和“系统” 两个要素。 相较于强调“人类认知建构” 的用户认知导向的信息检

① 算据,也称数据,是人工智能的核心要素之一。 对于 AIGC 而言,大数据语料和训练集为 AIGC 创作提供了最
基本的算据支撑,是 AIGC 的基础“燃料” [2] 。

2025 年 1 月 January,2025