Discourse, sentiment, and resonance of workplace burnout on Chinese social media: A computational analysis

Background and theoretical context: Occupational burnout has evolved from an individual psychological concern into a systemic challenge to organisational sustainability. Traditional burnout research relies predominantly on structured scales or small-scale interviews, making it difficult to capture the spontaneous, socially interactive emotional expressions that employees generate in authentic contexts. Against the backdrop of a discursive shift in Chinese digital workplace culture from individual attribution to structural attribution, social media has become a critical arena in which employees express burnout, seek resonance, and produce collective critical discourse, thereby offering a new perspective on burnout research that goes beyond the scale-based paradigm. Research objectives: This study aims to systematically examine the discourse structure of contemporary Chinese workplace burnout (how employees talk about burnout, and which themes appear frequently and are invested with meaning) as well as its resonance mechanism (which emotional expressions are more likely to elicit collective online engagement), with a view to revealing the dual structure of workplace burnout discourse. Methods: Adopting a cross-sectional observational computational social science paradigm with a descriptive and exploratory design, and integrating the theoretical frameworks of emotional labour and affective publics, this study collected texts from five major Chinese social media platforms (Rednote, Bilibili, Weibo, Douyin, and Douban). After data cleaning and line-by-line manual review, 1,276 valid texts were retained. Grounded theory three-layer coding (open coding, axial coding, and selective coding; Fleiss’ Kappa = 0.792) was used to identify discourse themes. A dual-model PCA fusion of RoBERTa and SnowNLP scores was employed to extract sentiment scores, supplemented by human annotation (n = 300; ICC(2,1) = 0.870) for cross-validation. OLS log regression and Poisson pseudo-maximum likelihood estimation (with covariates including follower count, platform, and year) were applied to examine the association between sentiment polarity and like-based engagement. Results: At the discourse level, grounded coding identified 6 axial themes and 3 core categories. The two highest-frequency themes were Labour Time Deprivation and Intensity Overload (22.7%) and Wage Deprivation and Employment Hardship (21.4%). Workers move beyond individualised narratives of emotional exhaustion, directly naming structural oppression through discourse frames such as Organisational Oppression and Institutional Critique (18.0%) and Passive Resistance and Collective Self-deprecation (16.5%), highlighting a paradigm shift from passive endurance to active critique in contemporary Chinese workplace burnout discourse. At the resonance level, follower count was found to exert a suppression effect on the sentiment-like relationship: after controlling for account exposure, negatively valenced content was significantly associated with higher like counts (PCA: beta = -0.403, p