改革开放40余年中国幸福观的内容维度与历史沿革——基于词嵌入技术的计算文本分析

The Content Dimensions and Historical Evolution of the Chinese Concept of Happiness in More Than 40 Years Since the Reform and Opening up——Computational Text Analysis Based on Word Embeddings Technology

改革开放四十余年以来,快速变迁的社会环境对中国社会的文化心理产生了深远影响,中国社会对“何为幸福”的认知也经历着时代变迁。文章首先通过梳理相关文献和采用深度访谈方法,提炼出兼具民族特色和时代特征的五大幸福观维度——“物质满足”“人际支持”“自我实现”“自然和谐”和“家国情怀”。在此基础上,文章基于1979-2021年《人民日报》上的新闻文本数据,采用词嵌入(word embeddings)技术(一种刻画词汇间语义关联的机器学习方法)刻画中国幸福观的历史演变。结果发现,对于“何为幸福”这一问题,“自我实现”的重要性呈上升趋势,以“家国情怀”为中心的传统幸福理念也逐渐得到巩固;自千禧年以来,“物质满足”基础幸福观维度整体式微,“自然和谐”幸福观维度的地位迎来了社会的重视。通过构建幸福变化指数发现,近40年来五个历史时期(1982-1984年、1991-1993年、2000-2002年、2013-2014年、2017-2018年)里的经济社会变动对中国幸福观的变化产生了较大影响。文章不仅从横向(研究范式)与纵向(跨时段视角)拓展了社会文化心理的相关研究,而且对构建国民幸福指标体系和发展幸福产业具有指导意义。

This paper investigates that an emerging computerized text analysis technique,word embeddings,can capture cul-tural and concept change.In this paper,we utilize this technique to quantify temporal dynamics of the concept of happiness be-tween 1979—2021 in China.We first conducted in-depth interviews and thus formed five dimensions that generally represent Chinese people's conception of happiness,including physical need,social support,self-actualization,environment harmony and patriotism.We then used word embeddings to analyze the associations of five dimension with well-being.Results show that the correlation between happiness and physical need declines over time while happiness's correlations with self-actualization and with patriotism go upward obviously.What's more,individualism("self-actualization")and collectivism-oriented("patriotism")dimension shows a consistent trend,this means that traditional value centerring on family and country interest has also been preserved at present age.Although the harmonious relationship with natural environment has not gotten enough at-tention in general,it tends to go upward these years.By comparing diachronic changes among five dimensions,this research captures the social economic shifts,e.g-,the Reform and Opening-up in the end of 1970s,economic reform since 1992,new millennium of 2000,the new era after 2013.Our method further proves to be valid by showing that changes in the embed-ding track closely with a series of domestic social surveys.Our research demonstrates the powerful utility of computational text a-nalysis and would facilitate intersection between machine learning and social science.