The study addresses the urgent need to understand and promote green consumption in China, particularly in light of the "dual-carbon goal" of carbon neutrality. While government, businesses, and social organizations are actively promoting green consumption through information strategies, the effectiveness of these strategies in changing public attitudes and behaviors remains unclear. The study focuses on the role of social media as a significant source of information influencing consumer perceptions and decisions regarding green consumption. Businesses and governments recognize the value of social media text data in understanding consumer sentiment and preferences. Consumers, in turn, utilize social media to gather information about products and their environmental impact, leading to informed purchasing choices. The inherent virality of social media platforms amplifies positive consumer experiences, making social influence a major driver of consumer behavior change and impacting green consumption. This research aims to explore how information strategies from different entities can influence green consumption attitudes through a large-scale analysis of social media text data, validating the Elaboration Likelihood Model (ELM) theory in a real-world setting. A deeper understanding of public attitudes towards green consumption and the factors driving them is of great practical significance for policymakers and business managers.

Existing research attributes willingness to engage in green consumption to both subjective motivations (behavioral perception, self-efficacy) and external environmental influences (laws, regulations, social and technological developments). Public attitude is considered a crucial antecedent of consumer behavior. The Elaboration Likelihood Model (ELM) suggests that attitude formation depends on the extent of information processing. While prior studies have used questionnaires or empirical methods, these are often limited by sample size and representativeness. This study leverages text data mining to overcome these limitations, extracting valuable information from extensive social media data. Various theories (Theory of Planned Behavior, Self-Determination Theory, Technology Acceptance Model, Use and Satisfaction Theory, Diffusion of Innovations Theory) have been applied to understand consumer behavior. However, these often focus solely on either high-level or low-level cognitive processing. The ELM offers a unique advantage by considering both, explaining information processing, selection, reception and the subsequent impact on attitudes. This study uses the ELM framework for quantitative analysis to understand consumer attitude changes after exposure to social media information, along with text mining to analyze the influence of information released by different sources on public attitudes.

The study employs a novel methodological framework integrating the Elaboration Likelihood Model (ELM) and text mining techniques. This approach addresses the limitations of previous studies by combining quantitative and qualitative analysis methods. The study uses social media data (news posts, online comments) on green consumption to analyze public attitudes. The data collection involves retrieving text data from official news sources and online comments, followed by preprocessing steps to clean and prepare the data for analysis. Text mining techniques are then applied using Convolutional Neural Network (CNN) for subject classification (government, business, media), Latent Dirichlet Allocation (LDA) for topic classification, and Long Short-Term Memory (LSTM) for sentiment analysis. The geographical location and gender of commenters are analyzed to understand the demographic factors influencing attitudes. The ELM framework is used to quantitatively analyze the changes in consumer attitudes, using a Linguistic Inquiry and Word Count (LIWC) dictionary and statistical unsupervised algorithms to quantify factors affecting attitudes. The Generalized Linear Mixed Model (GLMM) is employed to test the influence of various factors (information content, attention, participation, policy environment, economic environment, gender, region, time) on public attitudes toward green consumption. The study tests several hypotheses related to the influence of information strategies on consumer attitudes and the various ways consumers process information and engage in information diffusion. The figure illustrates the implementation procedure of the proposed framework based on ELM and text analysis.

The study reveals that women and individuals in economically developed regions (East China) show greater concern for green consumption. Sentiment analysis shows positive public emotions toward government policies and corporate actions, but negative reactions toward media campaigns. Topic modeling identifies specific themes related to government policies (environmental protection, waste reduction, low-carbon travel), corporate measures (express delivery carbon reduction, green home appliances), and media promotion (advocating low-carbon lifestyles). The public's responses were categorized and quantified into positive and negative sentiments. Sentiment analysis revealed positive public attitudes toward government policies and corporate measures, indicating support for initiatives promoting green consumption. Conversely, negative sentiments were prevalent regarding media campaigns. Analysis of commenter demographics showed higher participation from women and those in East China. The study's GLMM analysis revealed that while public attention and understanding of information positively influence participation, the impact on attitudes varies depending on the source (government, business, media). The economic environment (CCI) negatively impacts attitudes toward green consumption, while the policy environment had no significant effect. For government policy topics, increased attention and understanding positively influenced participation in information dissemination. However, this didn't directly translate to changes in attitudes. For corporate topics, increased attention to corporate initiatives was associated with greater participation and led to improved attitudes toward green consumption. The consumer confidence index negatively correlated with public engagement and attitudes. For media topics, increased attention and understanding of information, along with active participation, positively correlated with attitude change, despite the initially negative public sentiment towards media campaigns.

The findings address the research question by showing how different information strategies influence public attitudes towards green consumption. The significance of the results lies in demonstrating the power of social media data in understanding consumer behavior and the value of combining text mining and theoretical frameworks like ELM for this analysis. The relevance to the field includes contributing a new methodology for studying consumer attitudes towards sustainability, highlighting the differential impact of information sources, and emphasizing the need for tailored strategies to promote green consumption. The study’s findings highlight the importance of tailoring information strategies to specific audiences and leveraging social media for effective communication. Enhancing consumer confidence and addressing the perceived low efficacy of green consumption behaviors are crucial in encouraging widespread adoption. Strengthening policy frameworks by creating clearer guidelines and offering incentives is recommended.

The study contributes a novel methodological framework combining ELM and text mining for analyzing public attitudes toward green consumption. It highlights the importance of understanding the differential impact of information sources (government, business, media) and the role of consumer confidence. Future research should expand the geographical scope, incorporate broader age groups, and refine the factor analysis to enhance the model’s accuracy and generalizability. The findings provide actionable insights for policymakers and businesses to design effective strategies for promoting green consumption.

The study's focus on Chinese online users limits the generalizability of the findings to other contexts. The reliance on social media data may not represent the views of all segments of the population, particularly those who are less active online. The factor analysis using LIWC dictionary and statistical unsupervised algorithm may not be exhaustive, and the choice of factors might influence the results. Further research should address these limitations by expanding the sample, exploring other data sources, and refining the factor analysis methodology.