People frequently share their work-related experiences and perceptions on social media, communicating abundant career messages—ideas, norms, and expectations about different occupations or workplaces—that may reinforce popular narratives about work, inform how people think about and approach their careers, and ultimately socialize people into the working world. This paper outlines the content and construction of career messages on TikTok based on a qualitative content analysis of 1,049 short-form TikTok videos using socio-technical rhetorical strategies and social cognitive career theory as analytical frameworks. The analysis revealed three ways messages functioned : (1) affective representations to share reflections and evaluations of work; (2) immersion to situate viewers within the depicted environments and routines; and (3) support for explicit informational guidance or requests. Visual and rhetorical devices facilitated these functions, including repackaging existing material, external audio, and strategic recording angles, which often coalesced into recognizable social media genres , such as vision boards, memes, and Day-in-the-Life videos. Altogether, functions, facilitators, and genres shape what and how career messages are disseminated and may influence how viewers cultivate sensibilities about work-life, normalize particular workplaces or routines, and position certain forms of career guidance as credible and accessible. We offer a framework on digitally-mediated vocational socialization , which highlights how platform features, affordances, and user practices may co-construct career meaning-making. This work provides a basis for advancing scholarship on how online environments like TikTok might influence career development processes and calls for future work that considers social media as an influential factor in career formation.
The rapid adoption of AI in customer-facing services, such as digital livestreaming, promises efficiency gains but introduces profound socio-technical frictions, including algorithm aversion and trust erosion. This paper develops a two-stage dynamic optimization model of AI-human co-learning over a finite horizon, endogenizing knowledge accumulation and algorithmic trust penalties in service delivery. In the first stage, the system determines the initial knowledge allocation to the AI; in the second, it dynamically paces the sampling of human interaction data to balance marginal learning benefits against endogenous reputational costs. We derive closed-form optimal daily sampling intensities and an implicit solution for initial allocation, establishing a threshold condition for the human-AI handover when algorithmic capability surpasses human knowledge. Analytical results demonstrate that an optimal policy strictly front-loads data sampling to capture compounded learning benefits, while quadratic trust penalties mathematically enforce conservative exploitation in later stages. These findings bridge mechanism design and social computing, formalizing the dynamic frictions often overlooked in static automation models. The framework provides actionable guidelines for algorithmic platform governance, enabling designers to calibrate AI initialization, manage user trust trade-offs, and pace data extraction. Extensions incorporating declining market response, heterogeneous data quality, and platform scale effects confirm the model’s robustness in complex socio-technical contexts.
As large language model (LLM) agents become routine participants in public online conversation, understanding how their replies reshape human-to-human interaction structure is a central challenge for social computing and platform governance. Yet most post-deployment evaluations focus on toxicity or engagement volume, leaving thread-level conversational structure—a dimension critical to deliberative quality, social capital formation, and equality of voice—largely unmeasured. We address this gap through a large-scale structural evaluation of @CommentR , a production LLM agent serving millions of users on Weibo, China’s leading microblogging platform. Modeling each thread as a directed reply graph, we match over 216,000 real post-deployment threads on strictly pre-anchor covariates and estimate effects on human-only conversational structure using a doubly robust estimator. Because agent replies often arrive before any human-to-human interaction is observed, we introduce lifecycle-aware estimands that distinguish early-stage formation effects from mature-thread rewiring effects. Under conditional ignorability, agent replies reduce reciprocity, increase branching, and reduce geographic homophily in early-stage threads, while degree-corrected bridging remains unchanged—consistent with a hub-and-spoke shift from dialogic exchange toward one-off commenting around a focal reply. In mature threads, eligibility for the incumbent-rewiring analysis is itself reduced by the agent—sustained incumbent exchange becomes less likely—so we report the mature-thread rewiring estimates as bounded rather than point-identified and treat the formation regime as the one in which our structural evidence is secure. The magnitude of the reshaping depends on how the agent answers: more comprehensive, factual replies produce a larger focal shift. Reply-target analysis confirms that a substantial share of human comments redirect toward the agent, reducing lateral human-to-human exchange. These findings demonstrate that public AI agents can reshape not only what people say but how people talk to each other , with direct implications for platform governance, conversational agent design, and the structural monitoring of AI-mediated public discourse.
The ubiquity of generative AI (GenAI) requires social computing scholarship to examine how such extensive AI mediation may (re)shape the emotional and analytic qualities of human communication, both of which are central to audience engagement, trust, and information integrity. This study conducts a large-scale evaluation of 11 custom and fine-tuned LLMs, including DeepSeek, Llama (Meta), Mistral, and Gemma (Google), by asking them to rewrite the complete set of content, published by a local news outlet over a 12-year span (2011–2023). Using transformer-based sentiment models (i.e., RoBERTa fine-tuned on GoEmotions) and lexicon-based measures (i.e., LIWC, NRCLex), we compare the emotional expressiveness and analytic style of human-written versus AI-adapted content across short-form (titles/headlines) and long-form communication (article bodies/content). Analyzing 3,623 original news articles, nearly 40,000 primary-prompt AI-generated rewrites, and an additional generic-prompt ablation set, we estimate linear mixed-effects models that account for article-level clustering and control for model and prompt heterogeneity, communication form, verbosity, and readability. Results show that, under the corpus and prompting conditions examined here, LLMs tend to intensify emotional expression relative to human-written texts. Meanwhile, LLM rewrites tend to preserve, and in many cases increase, linguistic markers associated with analytic style relative to human benchmarks. Interestingly, even under the simpler generic prompt, LLMs in our sample display a shift toward more positive emotional categories, including joy, excitement, admiration, optimism, and gratitude, whereas human writers in this corpus display a broader affective range and more nuanced expression. To assess risks of clickbait-like headline framing, we fine-tune a BERT model for clickbait detection and examine linguistic cues in titles and headlines, yet we find minimal evidence that LLMs’ emotional amplification devolves into sensational and hyperbolic framing. Lastly, to enhance validity, we conduct a blinded human evaluation study on a randomly sampled subset of the corpus. Across 17 small experiments, human judges ( n = 170) independently evaluated analytic style, reasoning depth, and clickbait characteristics. These human evaluations provide convergent evidence that AI-generated texts were perceived as more analytically structured and slightly higher in perceived reasoning depth, while AI-generated headlines were also rated as less clickbait-oriented. Overall, our findings suggest that GenAI and LLMs have the potential to enhance the affective intensity of communication without necessarily compromising its informational quality. However, as human communication becomes increasingly and inevitably AI-mediated, current LLMs may shift the affective register, narrow the emotional spectrum, and introduce tonal biases in public communication that could subtly influence how information is disseminated, and how language is traditionally interpreted.
As large language models (LLMs) become deeply embedded in daily life, the urgent need for safer moderation systems that distinguish between naive and harmful requests while upholding appropriate censorship boundaries has never been greater. While existing LLMs can detect dangerous or unsafe content, they often struggle with nuanced cases such as implicit offensiveness, subtle gender and racial biases, and jailbreak prompts, due to the subjective and context-dependent nature of these issues. Furthermore, their heavy reliance on training data can reinforce societal biases, resulting in inconsistent and ethically problematic outputs. To address these challenges, we introduce GuardEval, a unified multi-perspective benchmark dataset designed for both training and evaluation, containing 106 fine-grained categories spanning human emotions, offensive and hateful language, gender and racial bias, and broader safety concerns. We also present GemmaGuard (GGuard), a Quantized Low-Rank Adaptation (QLoRA), fine-tuned version of Gemma3-12B trained on GuardEval, to assess content moderation with fine-grained labels. Our evaluation shows that GGuard achieves a macro F1 score of 0.832, substantially outperforming leading moderation models, including OpenAI Moderator (0.64) and Llama Guard (0.61). We show that multi-perspective, human-centered safety benchmarks are critical for mitigating inconsistent moderation decisions. GuardEval and GGuard together demonstrate that diverse, representative data materially improve safety, and adversarial robustness on complex, borderline cases.
