The analysis used the public Key Propagation Graph Generator/CUHK Weibo propagation corpus, assembled from parent–child records between November 2019 and March 2022. The full catalog contained approximately 4,174 source events, represented by roughly 960,000 microblogs, including reposts and comments, in the dataset authors’ Weibo22 release. Researchers excluded trees with fewer than 20 or more than 250 nodes, leaving 1,013 eligible cascades. They reconstructed directed trees from repost relationships, then calculated each event’s maximum breadth-first-search depth and its relative breadth, defined as the number of direct recipients of the seed divided by the cascade’s total size. These two structural features were standardized and separated with two-means clustering, yielding a silhouette coefficient of 0.513.

Before sampling, the clustering procedure assigned 399 events to a deep–narrow group and 614 to a shallow–broad group. The researchers then drew a balanced sample of 40 cascades from each regime, stratifying by cascade-size tertiles to reduce the chance that one size range would dominate the comparison. The resulting dataset represented about 7.9% of eligible events and contained 6,527 nodes connected by 6,447 directed repost edges. The sample was therefore designed for contrasting cascade structures, but it cannot establish the natural prevalence of either regime on Weibo. The 20–250-node eligibility window was selected before analysis to avoid near-degenerate small trees and very large events whose breadth-first-search truncation could distort depth and breadth measures.

Two prespecified hypotheses received confirmatory support. The first predicted that deep–narrow cascades would contain more opportunities for sequential, chain-relayed influence, reflected in longer average path lengths. The observed mean was 3.235 for deep–narrow events and 2.218 for shallow–broad events. A Welch two-sample test with Holm–Bonferroni adjustment produced an adjusted p-value below 0.001 and Cohen’s d of 2.71, indicating an exceptionally large separation. The second hypothesis predicted greater modularity in deep–narrow cascades, consistent with fragmented clusters in which locally connected users might encounter repeated peer endorsement. Mean modularity was 0.551 versus 0.200, also with an adjusted p-value below 0.001 and Cohen’s d of 2.95. These were effects on network geometry, not direct measurements of the psychological processes that might accompany the geometry.

Not every network measure distinguished the regimes. Average density was 0.0177 in deep–narrow cascades and 0.0231 in shallow–broad cascades, a difference that did not remain significant after multiplicity correction. Global clustering was effectively zero in both groups because parent–child repost trees rarely contain closed triangles. The selective result matters: the contrast was not a generic consequence of one group containing denser or more tightly interconnected networks, but centered on the distance between users and the separation of local communities. The authors retained density and clustering as confirmatory completeness checks rather than treating every structural statistic as evidence of a different influence regime.

Hop-by-hop analysis showed that the structural contrast developed beyond the initial burst of sharing. Shallow–broad cascades had more direct recipients of the seed, with a mean out-degree of 62.3 compared with 49.4 for deep–narrow cascades. Yet the pattern reversed at the next layer: deep–narrow nodes had a mean out-degree of 0.516, compared with 0.124 in shallow–broad cascades. At hop two, deep–narrow cascades contained 1,019 participating nodes, compared with 308 in shallow–broad cascades; at hop three, the counts were 330 and 64. By hop four, only seven shallow–broad nodes remained, while 131 deep–narrow nodes were still active. Deep–narrow events continued, with declining counts, through hops five to nine, whereas shallow–broad events effectively ended by hop five. These profiles describe continued opportunities for local influence, not a measured trajectory of anxiety, emotion, persuasion or uncertainty reduction.

An exploratory analysis also found higher betweenness centrality in deep–narrow cascades across all three branching-role proxies. Mean values were 0.003781 versus 0.000216 for high-branching nodes, 0.003215 versus 0.001377 for medium-branching nodes, and 0.004603 versus 0.000722 for low-branching nodes. The roles were structural classifications based on out-degree, not verified categories of “influencers,” media outlets or ordinary users. Because these comparisons were not included in the multiplicity-controlled confirmatory family, the result is treated as hypothesis-generating evidence that deep–narrow networks may distribute brokerage across several classes of nodes rather than concentrate it in one identifiable group. The interpretation remains descriptive and does not establish that a particular account or class of users controlled the flow of information.

Together, path length, modularity and hop-layer branching offer a meso-level picture of how sharing behavior is organized. In deep–narrow cascades, separated modules coexist with elongated relay chains, creating multiple locations where information can pass between branches and where local social proof could potentially accumulate. In shallow–broad cascades, many recipients appear close to the seed or to early distributors, producing a compact pattern dominated by early broadcast exposure. The study therefore distinguishes opportunities for psychological processes from proof that those processes occurred. No questionnaire, self-report, emotion coding, personality measure or attitude assessment was available to show whether users actually felt uncertain, sought reassurance or changed their views. The researchers also did not use the corpus’s text features to score emotion, topic or veracity.

Several limitations constrain the conclusions. The analysis is observational, so topology may reflect unmeasured factors such as the seed account’s characteristics, topic, timing, audience size, platform recommendations, external events, temporal change across 2019–2022 or other factors unavailable in the edge-list archive. The data also span years that may include changes in Weibo’s platform environment, and the two-cluster solution was not presented as a uniquely natural taxonomy; alternative clustering features, size windows or numbers of clusters could shift membership at the margin. Average path length is mathematically related to depth, one of the features used to define the regimes, so the large contrast partly characterizes the clusters rather than independently validating them. Although a continuous regression related path length to depth and breadth with an R² of 0.756, that relationship likewise cannot establish a causal psychological mechanism or prove that deep–narrow cascades were more emotional, persuasive or truthful.

For media psychology, the study’s principal advance is a measurable middle layer between isolated forwarding choices and platform-wide outcomes such as virality. Cascade geometry can be treated as a set of indicators for future research: path length can represent sequential opportunity, modularity can capture local clustering, betweenness can describe brokerage and hop-layer branching can track where diffusion opportunities persist. Pairing those structural measures with participant surveys, coded content, personality data or experimental attitude measures could test whether users experience the two regimes differently. The current evidence supports a narrower conclusion: within this selected observational sample of Weibo cascades, observed sharing aggregated into at least two structurally distinct influence regimes—one built around extended, modular chains and the other around early, hub-centered exposure—without demonstrating the intrapersonal psychological mechanisms behind either structure.

Subject of Research: Observational analysis of structural regimes in Weibo repost cascades

Article Title: Social influence aggregation on Weibo: psychological process expectations and meso-level cascade regimes in media psychology

Article References: Wang, L. (2026). Social influence aggregation on Weibo: psychological process expectations and meso-level cascade regimes in media psychology. Frontiers in Psychology, 17, Article 1953616. https://doi.org/10.3389/fpsyg.2026.1953616

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1953616

Keywords: Weibo, social influence, information cascades, media psychology, network topology, reposting, online diffusion, social proof, digital media, Social, influence, aggregation

The research team, led by Kamron Naher of Presidency University in Dhaka alongside colleagues from institutions across Bangladesh and China, built their investigation on a fusion of two classic frameworks in information systems research. The first is the Technology Acceptance Model, or TAM, which holds that two beliefs govern whether people adopt a technology: perceived usefulness, the sense that a tool improves performance, and perceived ease of use, the sense that it requires little effort. The second is the DeLone and McLean IS Success Model, which evaluates systems through quality dimensions such as the calibre of information delivered, the responsiveness of interactions, and the technical reliability of the platform itself. By stitching these frameworks together and adding social influence from the Unified Theory of Acceptance and Use of Technology, the researchers created a model capable of testing how system-level qualities and social pressures jointly shape student attitudes.

The data collection ran from January to April 2025 using a structured online questionnaire distributed through Facebook, Messenger, WhatsApp, and email networks, with QR codes shared across academic communities. Respondents were screened to confirm prior experience with ChatGPT, ensuring that every participant could evaluate the tool from direct use rather than hearsay. Of 657 initial responses, 640 survived rigorous screening for completeness and consistency. The sample skewed male at 55.63 percent, with nearly half of participants aged between 21 and 30, and just over half enrolled at public universities. Business students formed the largest disciplinary group at 25 percent, followed by respondents from engineering and information technology fields.

Analysis proceeded through partial least squares structural equation modeling, a statistical technique well suited to testing networks of relationships among latent psychological constructs. The researchers first verified their measurement model, confirming internal consistency with Cronbach’s alpha and composite reliability values above accepted thresholds, and establishing convergent and discriminant validity through average variance extracted, the Fornell-Larcker criterion, and heterotrait-monotrait ratios. Common method bias was ruled out using the full collinearity approach, with all variance inflation factors falling below the 3.3 cutoff. The structural model was then tested with 10,000 bootstrap subsamples, and the resulting coefficient of determination reached 0.91 for the endogenous variables, indicating that the integrated framework explains an exceptionally large share of the variance in the outcomes it models.

Every one of the eleven hypotheses was supported. Information quality, meaning the accuracy, relevance, timeliness, and clarity of ChatGPT’s outputs, emerged as a powerful driver of perceived usefulness. This finding carries particular weight in Bangladesh, where formal academic support can be scarce and students often depend on self-directed learning. When the chatbot returns answers that are correct, well-structured, and pertinent to coursework, students come to see it as a genuine performance enhancer rather than a novelty. Interaction quality, defined as the responsiveness, coherence, and conversational richness of the exchange, significantly boosted perceived ease of use, confirming that natural, context-aware dialogue lowers the friction of learning to work with generative AI.

System quality, encompassing reaction speed, dependability, and simplicity of the platform, influenced both perceived usefulness and perceived ease of use, echoing earlier findings from e-learning research. Perceived usefulness and perceived ease of use were mutually reinforcing, and both fed directly into students’ attitudes toward ChatGPT. Attitude, in turn, proved to be the single most consequential predictor of actual adoption, with an importance-performance map analysis quantifying its total effect on adoption at 0.47. The researchers also applied Necessary Condition Analysis, a complementary method that identifies whether predictors are prerequisites rather than mere contributors. Effect sizes ranged from 0.22 to 0.40, with social influence showing a medium effect and the remaining constructs showing large effects, and all permutation tests reached significance at the 0.05 level.

The headline discovery, however, concerns the moderating role of social influence. In Bangladesh’s collectivistic academic culture, the positive relationship between attitude and adoption was significantly strengthened by the encouragement of peers, the endorsement of early users, and supportive institutional environments. Students did not simply assess ChatGPT’s quality in isolation and decide for themselves; they calibrated their judgments against the visible enthusiasm of friends, classmates, faculty, and online communities. The authors argue that this finding challenges Western, individualistic models of technology acceptance, in which adoption is typically portrayed as a solitary cognitive decision. In a developing South Asian educational ecosystem, systemic quality metrics do not automatically translate into use. They are heavily filtered through social structures before becoming behaviour.

The theoretical contribution lies in demonstrating that TAM, the DeLone and McLean model, and UTAUT are not redundant alternatives but complementary lenses whose integration is conceptually necessary. The study also refines the meaning of interaction quality for generative AI specifically, showing that prompt-dependent, conversational systems expand the traditional boundaries of technology acceptance theory in higher education. Earlier work by Almulla in 2024 had established that quality parameters matter for smart learning systems, but the Bangladeshi study goes further by modeling adoption as a socially certified phenomenon and by shifting the analytical lens to a developing economy where peer momentum acts as a catalyst amplifying every direct pathway in the model.

The practical implications are equally pointed. The authors urge university administrators to abandon top-down mandates in favor of peer-led learning circles and collaborative workshops where students collectively model ethical and productive AI use. Institutions are advised to craft clear, context-specific AI guidelines and robust data-handling frameworks to address student anxieties about privacy and academic integrity, transforming conversational AI into a trusted learning assistant rather than a source of institutional risk. Developers, meanwhile, should prioritize onboarding guidance and intuitive interfaces, since early friction discourages novices. The study also acknowledges persistent counterweights: hallucinated information, plagiarism concerns, and data privacy worries remain active background constraints even when social momentum accelerates uptake.

The authors are candid about limitations. The purposive, non-probability sampling design, while necessary given the absence of any centralized database of AI-using students in Bangladesh, restricts generalizability to non-users and other demographics. The cross-sectional design captures attitudes at a single moment and cannot confirm long-term causal trends, and the sample over-represents public universities and business, engineering, and IT students. Roughly 37 percent of the variance in adoption remains unexplained, pointing to factors such as digital literacy, institutional support, prior AI experience, and internet access that future models should incorporate. Longitudinal tracking, cross-cultural comparisons across South Asia, and studies including non-adopters are flagged as priorities. Even so, the message of the research is unmistakable: in the classrooms of emerging economies, the future of AI in learning will be decided not just by the quality of the algorithms, but by the quality of the conversations students have about them with each other.

Subject of Research: Factors influencing ChatGPT adoption among university students in Bangladesh, including the moderating role of social influence

Article Title: Factors affecting ChatGPT adoption among Bangladeshi university students with the moderating role of social influence

Article References: Naher, K., Siddiqi, K. O., Al Mamun, M. A., Chakma, S., Rashid, F., Arju, U. H., Alam, S., & Tanchangya, T. (2026). Factors affecting ChatGPT adoption among Bangladeshi university students with the moderating role of social influence. Discover Education, 5(1), Article 993. https://doi.org/10.1007/s44217-026-02075-8

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02075-8

Keywords: ChatGPT, technology acceptance model, social influence, Bangladesh, higher education, information quality, interaction quality, system quality, PLS-SEM, generative AI, university students, DeLone and McLean model

The missing ingredient is what the authors call multi-scale local structure, a form of clustering that operates not just at a single scale but across many of them simultaneously. In everyday terms, real social networks are not merely collections of friends who know one another in tight triangles. They contain overlapping communities of many sizes, from small cliques of close contacts to larger neighborhoods, workplaces, and social circles nested inside one another. The researchers show that when you sample real interaction networks from empirical data, this layered, nested clustering is pervasive, and that common baseline models used to generate synthetic comparison networks, such as the Erdős–Rényi random graph and the Chung-Lu model, simply do not reproduce it.

The distinction matters because epidemic behavior turns out to be remarkably sensitive to it. In simulations, the team found that epidemics spreading on real network samples behave very differently from epidemics spreading on synthetic models built to match the same basic statistics. The multi-scale local structure acts as a kind of natural firebreak. Clusters of densely connected individuals tend to contain outbreaks locally, so that even when the virus penetrates one community, it struggles to leap into the next. In the real network samples, local quarantining interventions could stop epidemic spread outright. In the simplified synthetic models of those same networks, the same interventions failed, and the epidemic marched on.

This is a striking reversal of a common intuition. Network scientists have long known that clustering, the tendency of your friends to know each other, affects transmission, and many modelers have treated it as a second-order refinement. The new results suggest that the triangle-based clustering captured by standard metrics is only part of the story. The authors demonstrate explicitly that the intervention-sensitivity they observe results from more than just local triangle structure. It is the multi-scale organization, the way dense pockets are themselves arranged into larger dense pockets across a range of sizes, that determines whether an outbreak can be contained by targeting individuals.

Technically, the researchers approached the problem by comparing epidemic dynamics on empirical samples of interaction networks against matched synthetic baselines. The Erdős–Rényi model connects every pair of nodes with the same probability, producing networks with no meaningful community structure at all. The Chung-Lu model improves on this by preserving the degree distribution, the fact that some people have many more contacts than others, but it still wires edges essentially at random given those degrees, wiping out higher-order organization. By running the same epidemic processes and the same quarantine strategies on both real and synthetic networks, the team could isolate precisely which structural features drove the differences in outcomes.

The findings carry practical weight for intervention design. Local quarantining, the practice of identifying an infected individual and cutting off their immediate contacts, is one of the most tractable public health tools available during an outbreak, since it does not require sealing off entire cities or vaccinating entire populations. The study shows that the effectiveness of this tool depends heavily on the underlying network fabric. On networks with widespread multi-scale local structure, quarantining is easier to apply successfully, because the structure itself cooperates with the intervention, confining the pathogen to neighborhoods that can be severed from the rest of the graph. On the flattened synthetic models, the same strategy cannot halt spread, because there are no natural boundaries for the quarantine to reinforce.

Beyond mitigation, the analysis yields a second, equally valuable insight: it characterizes which nodes are ultimately unlikely to be infected. In a network riddled with multi-scale clustering, the epidemic does not reach everyone. Certain individuals, by virtue of their position in the layered community structure, sit in pockets that the outbreak never penetrates. Identifying such nodes in advance could help public health officials understand the natural limits of spread, refine risk estimates, and prioritize surveillance on the individuals and bridges between communities where transmission is actually likely to occur, rather than spreading resources uniformly across a population whose members face very different levels of risk.

Where does this multi-scale structure come from in the first place? The authors illustrate processes that plausibly generate it. One candidate is homophily, the well-documented tendency of people to form ties with others who are similar to them, whether in age, location, occupation, or interest. Homophily naturally produces nested layers of similarity-based grouping, from households to neighborhoods to professional communities. Another candidate is social influence, through which behaviors and connections propagate and consolidate within groups over time. The researchers also show that random walks on a network can be used to isolate and expose this multi-scale organization, providing a clean analytical handle for separating intervention sensitivity that arises from multi-scale structure from sensitivity attributable to simpler features like degree heterogeneity.

The methodological lesson for the modeling community is pointed. If epidemic simulations are calibrated against synthetic networks that lack multi-scale local structure, the predictions they produce may be systematically pessimistic, overstating both the final size of an outbreak and the difficulty of controlling it. Conversely, models that faithfully preserve the layered clustering of real contact data may reveal that interventions are more effective than simpler analyses suggest. The choice of network generator is therefore not a technical footnote but a first-order determinant of the conclusions drawn from any simulation study of transmission dynamics.

More broadly, the work adds to a growing recognition that higher-order network structure, the organization of connections beyond simple counts of neighbors, shapes dynamical processes in ways that first-order statistics cannot capture. For epidemic modeling specifically, the message is that the texture of human contact patterns, with its nested communities forged by homophily and social influence, is not noise to be averaged away but the very feature that determines whether an outbreak smolders locally or sweeps through a population. As the authors demonstrate, capturing that texture in synthetic models, or working directly with empirical network samples, may be essential for forecasting epidemics accurately and for designing containment strategies that work in the world as it is actually wired.

Subject of Research: The impact of multi-scale local network structure on epidemic spread and intervention effectiveness

Article Title: Multi-scale local network structure critically impacts epidemic spread and interventions

Article References: Eldaghar, O., Mahoney, M. W., & Gleich, D. F. (2026). Multi-scale local network structure critically impacts epidemic spread and interventions. PLOS Complex Systems, 3(9), e0000116. https://doi.org/10.1371/journal.pcsy.0000116

Image Credits: AI Generated

DOI: 10.1371/journal.pcsy.0000116

Keywords: epidemic modeling, network science, multi-scale structure, quarantining, Erdős–Rényi model, Chung-Lu model, clustering, homophily, social influence, random walks, complex systems, intervention strategies

The trap rests on two well-studied mechanisms of moral self-regulation. The first is vicarious moral licensing. When people observe someone else performing a morally commendable act, the observed virtue can be absorbed into their own self-concept, generating a sense that their moral account is already in credit. That surplus then licenses a reduction in personal effort. The second mechanism is vicarious moral identification. When people observe someone close to them behaving badly, they may mentally identify with the transgressor, and that identification justifies their own inaction: if someone like me is not bothering, why should I? In the environmental domain, this means a friend’s diligent composting can quietly erode one’s own motivation, while a friend’s flagrant wastefulness can serve as an excuse for doing nothing at all.

What makes the new work distinctive is that it does not simply re-document the trap. The central question the authors posed was how it can be broken. Their answer, tested across the four studies, involves two psychological buffers: situational openness, which concerns whether the witnessed behavior is public or private, and environmental self-identity, which concerns how deeply green values are woven into a person’s sense of who they are. The findings identify boundary conditions for the vicarious effect and offer preliminary evidence for an identity-driven pathway that could inform interventions built on social norms and self-concept.

Study 1, with 97 participants, established the basic phenomenon and probed an important boundary: the type of relationship. Participants recalled either a close friend’s or a family member’s environmental behavior and then reported their own subsequent pro-environmental intentions and actions. The results were selective. Recalling a close friend’s environmental actions reduced participants’ subsequent pro-environmental behavior, whereas recalling a family member’s actions did not produce the same dampening effect. The authors interpret this asymmetry through the lens of self-construal and comparison processes. Friends are typically perceived as similar others against whom we implicitly calibrate our own conduct, making their behavior especially potent raw material for vicarious licensing. Family members, embedded in roles of obligation and interdependence, may not trigger the same comparative accounting.

Study 2, with 180 participants, tested the first buffer. Situational openness was manipulated by varying whether the recalled behavior of the close other was framed as publicly visible or private. The reasoning draws on accountability dynamics: behavior that is public carries reputational stakes, both for the actor and for the observer who now knows about it. When the friend’s environmental behavior was made public, the licensing effect vanished. Openness, in other words, eliminated the trap entirely. The practical implication is provocative. Environmental virtue that circulates openly, through visible commitments, community recognition, or shared reporting, may protect against the complacency that private virtue can breed, because public behavior invites scrutiny and signals norms rather than supplying moral credits.

Studies 3 and 4 turned to the second, and arguably deeper, buffer: environmental self-identity, the extent to which being an environmentally responsible person is central to an individual’s self-concept. Study 3, again with 180 participants, measured environmental self-identity and found that participants high in this trait showed complete immunity to the vicarious trap. For them, a friend’s good deeds did not license slacking, and a friend’s bad deeds did not justify inaction. The mechanism makes theoretical sense. Moral licensing operates by shifting the balance of one’s moral self-worth; if one’s identity as a green person is stable and central, a single observed act by someone else cannot easily tip that balance. Identity, in this account, functions as ballast against the vicarious winds of other people’s conduct.

Study 4 provided the most methodologically ambitious test. Sixty-six participants completed a brief identity-enhancement intervention and then kept a diary for 30 days, recording their daily environmental self-identity and their daily pro-environmental behavior. The authors describe the evidence from this study as preliminary and exploratory, but the pattern was consistent with an identity-driven pathway. Daily environmental self-identity predicted next-day pro-environmental behavior even with autoregressive controls, meaning that the statistical model accounted for each person’s own prior behavior and identity levels. Mediation analyses further supported identity as the causal mechanism linking the intervention to behavior. In plain terms, strengthening the green self-concept appeared to translate into more sustainable action on subsequent days, suggesting that identity is not merely a correlate of behavior but a moving force behind it.

Taken together, the four studies sketch a coherent psychological architecture. The vicarious trap is real, it is strongest in the context of close friendships rather than family ties, and it can be dismantled by two distinct routes. The first route is situational: make environmental behavior visible and open, so that it functions as a social signal rather than a private moral credit. The second route is dispositional: cultivate environmental self-identity so deeply that external examples, positive or negative, lose their power to regulate one’s own effort. The two buffers operate at different levels, one reshaping the social context in which behavior is witnessed and the other reshaping the self that does the witnessing, which is precisely why the authors present them as complementary rather than competing.

The findings arrive at a moment when social-norm campaigns are a mainstay of environmental policy. Utility companies send households comparisons with their neighbors’ energy use; cities publicize recycling rates; apps gamify low-carbon choices. The vicarious trap suggests a hidden cost to such approaches that designers rarely consider: a well-crafted message about a model citizen’s green feats could, for some recipients, function as a license to ease off. The new results indicate that the risk is not uniform. It concentrates among people whose environmental identity is weak or unformed, and it can be neutralized when the exemplar’s behavior is public. Campaign planners might therefore pair norm-based messaging with identity-building components, encouraging people to internalize environmental action as part of who they are rather than as a ledger that others’ good deeds can balance on their behalf.

The authors are careful about the limits of their evidence. The samples were modest, the diary study was small and explicitly framed as exploratory, and the identity intervention was brief, so the durability of identity-driven change remains an open question. The relationship-type finding from Study 1 raises further questions about cultural variation, since the relative weight of friend versus family comparison differs across societies. Still, the convergence of experimental, correlational, and longitudinal methods within a single program is unusual in this literature, and the practical upshot is clear enough to act on. If we want other people’s virtue to inspire rather than anesthetize, we should make that virtue visible and work to make environmental identity personal. The vicarious trap thrives in the dark, on private virtue and fragile self-concepts; openness and identity are how it gets broken.

Subject of Research: Vicarious moral licensing and identification in pro-environmental behavior and buffers against close others' environmental influence

Article Title: Breaking the vicarious trap: two buffers against close others’ environmental influence

Article References: Xiao, X., Zhan, Y., & Zhong, Y. (2026). Breaking the vicarious trap: two buffers against close others’ environmental influence. BMC Psychology. https://doi.org/10.1186/s40359-026-05363-y

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05363-y

Keywords: pro-environmental behavior, vicarious licensing, moral identification, environmental self-identity, situational openness, social influence, moral self-regulation, diary study, social norms, environmental psychology, self-concept, behavioral intervention

The research, published in Discover Psychology, was led by Riketa Parmar and Vipul Patel of Ganpat University’s V. M. Patel College of Management Studies, together with Mruga H. Mehta and Jigneshkumar P. Desai of Parul University. The team set out to build and test what they call the Green Banking Consumer Adoption Model for India, or GB-CAM-India. Rather than relying on a single behavioural theory, the model deliberately stitches together three of the most widely used frameworks in consumer and environmental psychology: the Technology Acceptance Model, the Theory of Planned Behaviour, and the Value-Belief-Norm theory. To these it adds constructs tailored to the Indian market, including perceived trust, perceived risk and customer satisfaction.

The logic behind the integration is worth unpacking. The Technology Acceptance Model, developed originally to explain how people come to use new technologies, focuses on two perceptions: that a system is useful and that it is easy to use. The Theory of Planned Behaviour adds the role of attitudes, subjective norms, meaning the social pressure we feel from people around us, and perceived behavioural control, our sense that we are capable of performing the behaviour. The Value-Belief-Norm theory, rooted in environmental sociology, traces pro-environmental action back through a chain of personal values, ecological beliefs and the activation of moral norms. Green banking sits at the intersection of all three: it is a technology, a planned consumer behaviour and an environmentally significant act at once, which is precisely why the authors argued that no single theory could capture the whole picture.

To test the model empirically, the researchers collected primary survey data from 760 respondents spread across five regions of Gujarat, a western Indian state with a large and diverse banking population. For each theoretical construct, they computed composite, or summed, scale scores and then subjected the data to a battery of statistical checks. Exploratory factor analysis was used to confirm that the survey items clustered into the intended dimensions, and reliability analysis confirmed internal consistency. The psychometric credentials of the measures were strong: the Kaiser-Meyer-Olkin measure of sampling adequacy reached 0.906, a value well above the conventional 0.6 threshold, every composite reliability value was at least 0.829, and all average variance extracted values were at least 0.545, calculated from the standardised loadings produced by the factor analysis. In plain terms, the survey instruments measured what they claimed to measure, and they did so consistently.

With the constructs validated, the team turned to multiple regression to test eight hypotheses linking the predictors to green banking adoption behaviour. The model performed respectably, explaining 46.8 percent of the variance in adoption behaviour, with an F-statistic of 82.51 on 8 and 751 degrees of freedom and a p-value below 0.001. Customer satisfaction dominated the results, with a standardised beta of 0.464, far ahead of any other predictor. Perceived ease of use came second at 0.193, followed by social influence at 0.083, both also significant at the 0.001 level. Perceived risk worked in the opposite direction, exerting a significant negative effect with a beta of minus 0.085 and a p-value of 0.002. The message for banks is that making green services feel safe, simple and satisfying matters more than preaching their environmental virtues.

One of the most technically interesting parts of the study concerns how satisfaction connects upstream perceptions to downstream behaviour. Using bootstrapped indirect-effect tests with 5,000 resamples and bias-corrected 95 percent confidence intervals, the researchers found that customer satisfaction carries significant indirect effects from perceived usefulness, perceived trust and environmental attitude through to adoption behaviour. Crucially, the corresponding direct paths from those three constructs to adoption were statistically nonsignificant. That pattern means satisfaction does not merely supplement these influences; it is the channel through which they operate. The authors are careful with terminology here, characterising these as indirect-only effects rather than partial mediation, a distinction that matters for how future researchers model the pathway.

The study also probed how the satisfaction-adoption relationship varies across the population. Income level emerged as a significant moderator of that path: higher-income consumers showed reduced sensitivity to satisfaction signals when deciding whether to adopt green banking. In other words, for wealthier customers, a pleasant service experience moves the adoption needle less than it does for lower-income consumers, a finding with clear implications for how banks segment their green marketing. The researchers additionally observed that mean adoption scores differed significantly across Gujarat’s five regions, with an F-statistic of 9.88 on 4 and 755 degrees of freedom and a p-value below 0.001. They are careful to note, however, that this is a group-difference finding rather than a formally tested regional moderation effect, an honest caveat that reflects the exploratory, composite-score design of the analysis.

The practical implications ripple outward from these statistics. For banks, the results suggest that investments in service quality, complaint resolution and overall customer experience may do more to accelerate green banking uptake than awareness campaigns alone, because satisfaction is the conduit through which usefulness, trust and environmental attitudes are converted into action. Reducing perceived risk, whether fear of fraud in digital channels or uncertainty about green products, should also pay dividends, given its significant negative coefficient. For policymakers working under the Reserve Bank of India’s Green Finance Framework, the regional differences in adoption scores across Gujarat hint that a one-size-fits-all national rollout may underperform, and that income-sensitive design could matter, since wealthier consumers respond differently to satisfaction cues than lower-income ones.

The authors are equally candid about the limits of their approach. Because the analysis relied on composite scores, exploratory factor analysis and multiple regression rather than a full structural equation model, the causal architecture of GB-CAM-India remains a proposal awaiting confirmatory testing. They explicitly flag directions for future SEM-based research that could estimate the full path model with latent variables, test the regional moderation hypothesis formally, and extend the sample beyond Gujarat to other Indian states. The survey design is also cross-sectional, so the direction of influence, while theoretically motivated, is inferred rather than observed over time.

Even with those caveats, the study lands at a propitious moment. As India’s financial sector aligns itself with sustainable development goals, understanding the psychology of the consumer becomes as important as the engineering of the green products themselves. What this research demonstrates is that the road to sustainable finance in India runs through the everyday experience of the bank customer: a useful app that is trusted, a service that satisfies, and a social environment that quietly nudges people toward greener choices. If satisfaction is the engine of green banking adoption, then the banks that win India’s sustainable finance transition may be those that treat environmental ambition and customer experience as a single, inseparable project.

Subject of Research: Consumer adoption of green banking services in India, modelled through an integrated TAM-TPB-VBN framework

Article Title: An integrated TAM–TPB–VBN framework with empirical validation from Gujarat for green banking consumer adoption in the Indian market

Article References: Parmar, R., Patel, V., Mehta, M. H., & Desai, J. P. (2026). An integrated TAM–TPB–VBN framework with empirical validation from Gujarat for green banking consumer adoption in the Indian market. Discover Psychology. https://doi.org/10.1007/s44202-026-00883-5

Image Credits: AI Generated

DOI: 10.1007/s44202-026-00883-5

Keywords: green banking, consumer adoption, Technology Acceptance Model, Theory of Planned Behaviour, Value-Belief-Norm theory, India, Gujarat, customer satisfaction, sustainable finance, perceived risk, social influence, survey research

The study was motivated by a persistent puzzle in environmental psychology. Human activities account for roughly 72 percent of total greenhouse gas emissions, and virtually every daily choice—what to eat, how to travel, what to buy—contributes to the problem. Yet motivating individuals to change their behavior remains notoriously difficult. Climate change is an extensive, protracted, and uncertain problem, which can evoke feelings of powerlessness or a sense of inequity when tackled at the individual level. One promising remedy is to reframe climate change as a collective social issue, harnessing the power of social norms: people constantly compare their own behavior to that of others and adjust accordingly, a process psychologists call social conformity.

But social norms carry what the researchers describe as a double-edged power. Observing neighbors adopt greenhouse gas-reducing behaviors can encourage imitation, and surveys show that people increase pro-climate behavior when they perceive such norms around them. Conversely, social influence can also inhibit sustainable action: visible littering, for example, increases the probability that others will litter, in line with the broken windows effect. Because there is no single, unambiguous social norm about climate behavior, understanding the mechanisms of conformity in both pro-climate and anti-climate directions—and whether those effects last—became the central question of the new work.

To answer it, a team led by Dahye Yeon and Ji-Bum Chung of Ulsan National Institute of Science and Technology, together with colleagues at Hanyang University, recruited 50 students and analyzed data from 47 after exclusions. Participants rated their intention to perform 113 pro-climate behaviors drawn from the UN’s 170 climate change actions and Korean government guidance, expressed on a continuous scale from 0 to 100 percent. Immediately after each rating, a fictitious group opinion—presented as the average intention of fellow UNIST students—appeared alongside their own response. The group opinion was algorithmically manipulated in real time along two dimensions: deviation from the participant’s initial answer (congruent, moderately incongruent at about 15 percent, or highly incongruent at about 35 percent) and direction (pro-climate, meaning the group was more climate-friendly, or anti-climate, meaning it was less).

Participants then repeated their ratings twice: once immediately afterward, allowing the researchers to compute immediate conformity as the change from the first to the second response, and once seven days later, capturing sustained conformity as the change from the first to the third response. This three-session design, rare in environmental psychology, allowed the team to distinguish transient compliance from genuine attitude change. Linear mixed-effects models with crossed random intercepts for participants and stimulus items analyzed 5,311 valid trials, and additional analyses controlling for regression to the mean confirmed that the effects could not be explained by participants simply drifting back toward average responses.

The behavioral results were clear. Immediate conformity increased systematically with greater deviation from the norm, regardless of direction: in the anti-climate condition, ratings shifted by −1.50 under congruent norms, −3.09 under moderately incongruent norms, and −5.43 under highly incongruent norms, while the pro-climate condition produced mirror-image positive shifts up to +5.31. The dramatic difference emerged one week later. Conformity toward highly incongruent pro-climate norms was amplified over time, with the estimated effect growing to 12.91 points, whereas conformity toward anti-climate norms remained comparatively weak and did not strengthen. According to the authors, immediate conformity likely reflects normative influence—complying with group pressure to avoid disapproval—while sustained conformity reflects informational influence and internalization, in which norms are accepted as credible evidence about what is genuinely appropriate.

The electroencephalographic recordings added a crucial layer. EEG was captured during the first two sessions from 31 electrodes, and the team analyzed two event-related potentials linked to conformity in prior research: feedback-related negativity (FRN), a frontocentral deflection peaking roughly 200 to 400 milliseconds after feedback that signals error detection and social conflict, and the late positive potential (LPP), appearing from about 300 to 400 milliseconds onward and associated with the perception of social approval and elaborative evaluation. The prediction, grounded in social norm theory and the Value-Identity-Personal (VIP) framework, was that because pro-climate behavior is morally positive, discrepancies with anti-climate norms would register as errors, whereas discrepancies with pro-climate norms would feel consistent with participants’ moral self-concept.

The neural data largely fit that profile. FRN modulation was observed only in anti-climate conditions, with a significant contrast between congruent and highly incongruent trials, suggesting that large deviations toward anti-climate group opinions were weakly processed as error signals or negative social feedback, while pro-climate discrepancies produced no comparable conflict response. The LPP, measured at the Pz electrode in a 500 to 700 millisecond window, showed a significant deviation-by-direction interaction: under anti-climate pressure, larger discrepancies elicited greater LPP amplitudes, consistent with heightened cognitive engagement with norm-incongruent information, whereas under pro-climate pressure, LPP amplitudes were attenuated at higher discrepancy levels—possibly because participants accepted pro-climate norms as socially desirable standards rather than processing them as errors. The authors caution that these ERP effects were limited and exploratory, but they illustrate how neurophysiology can reveal processing differences invisible to self-report, which is vulnerable to social desirability and experimenter demand biases.

The study also examined whether personal predispositions moderate conformity, using a questionnaire based on the VIP theory assessing biospheric values, environmental identity through the nature relatedness scale, and personal norms of moral obligation. Demographics influenced baseline intentions—women and supporters of non-conservative parties reported higher pro-climate intentions—but neither sex nor political affiliation affected conformity itself, suggesting norm-based interventions can work across demographic groups. Personal norms emerged as the most consistent moderator: individuals with strong personal norms resisted social pressure and defended their initial positions, particularly against pro-climate norms, indicating they had already internalized those attitudes. Biospheric values similarly predicted resistance to anti-climate pressure over the week-long delay.

The authors acknowledge limitations, including a homogeneous sample of Korean science and technology students with likely elevated climate awareness, the absence of a no-norm control condition, and the treatment of all pro-climate behaviors as roughly equivalent in cost and effort. Cross-cultural replication is needed, especially since South Korea is a context of heightened sensitivity to collective consensus—making the durability of pro-climate conformity there all the more telling. Still, the policy implications are provocative: messages emphasizing anti-climate descriptive norms, such as how many people are failing to act, may generate only fleeting compliance, while communicating pro-climate behaviors as socially validated, credible standards can produce lasting shifts in intention. As the authors conclude, presenting pro-climate social norms and highlighting when individuals fall short of them may be a more reliable route to sustainable behavior than moral appeals alone.

Subject of Research: Social conformity in pro-climate behavior examined through behavioral experiments and EEG measures of social norm influence

Article Title: Exploring social conformity in pro-climate behavior: behavioral and neurophysiological responses

Article References: Yeon, D., Kim, J., Lee, B., Shim, M., Sul, S., Kim, S.-P., & Chung, J.-B. (2026). Exploring social conformity in pro-climate behavior: behavioral and neurophysiological responses. Current Psychology, 45(18), Article 1524. https://doi.org/10.1007/s12144-026-10066-z

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10066-z

Keywords: social conformity, pro-climate behavior, social norms, EEG, event-related potentials, feedback-related negativity, late positive potential, environmental psychology, sustainability, climate change behavior, social influence, personal norms

The distinction matters because the two effects demand completely different responses. If opinions converge because agents influence one another through structured exchange, that is a dynamical phenomenon — something that depends on network topology, repeated contact, and the rules of communication. If opinions converge because every model shares the same training data and the same fine-tuning choices, that is a bias phenomenon, and no amount of network rewiring will fix it. Conflating the two leads researchers to draw false conclusions about social dynamics whenever they use language models to simulate human populations, a practice that has grown rapidly across computational social science.

Classical opinion dynamics models, from the DeGroot framework to bounded-confidence models such as Deffuant and Hegselmann–Krause, have long separated individual predispositions from interpersonal influence. A human agent starts with a prior position and then updates it as a weighted function of the neighbors’ positions. When all agents start from identical priors, any observed change must come from the interaction term. When agents never interact, any observed alignment must come from shared priors. The new study imports this clean separation into the world of language models, where it is far harder to achieve, because an LLM’s ‘prior’ is opaque and entangled with billions of parameters shaped by training corpora, alignment procedures, and decoding settings.

The methodological core of the work lies in designing controlled experimental conditions that isolate the two channels. In one condition, model agents are allowed to exchange opinions across multiple rounds, each agent seeing and responding to the outputs of others, so that interaction effects can accumulate. In a matched condition, agents are queried in complete isolation, with no exposure to each other’s outputs, so that any consistency in their answers reflects only the models’ intrinsic tendencies. By comparing the trajectories of these two conditions across topics, model families, and network structures, the researchers can estimate how much of the observed opinion dynamics is attributable to social influence and how much to shared bias.

The results carry an important warning for anyone using LLMs as synthetic participants. Because large commercial models are trained on broadly overlapping corpora and aligned toward similar helpfulness and safety objectives, they exhibit substantial baseline agreement: asked independently, they often cluster around the same positions on political, ethical, and policy questions. When such models are then placed in an interaction network, this shared prior acts like a strong external field, pulling the whole population toward the same attractor regardless of the communication structure. Apparent consensus in an LLM society can therefore be an artifact of homogeneity in the underlying models rather than an emergent product of deliberation — the opposite of what genuine social consensus formation looks like in human groups.

The study also quantifies when interaction does matter. Interaction effects become visible when agents begin from heterogeneous positions, when prompts are constructed to suppress the models’ default leanings, or when network structures channel information asymmetrically so that some agents act as hubs and others as peripheral listeners. Under these conditions, the trajectory of collective opinion diverges measurably from the isolated-query baseline, and classical dynamical concepts — anchoring, threshold effects, polarization into clusters — reappear in recognizable form. This suggests that the tools of decades of opinion dynamics research remain applicable to artificial agents, provided the bias floor is first accounted for.

For the broader scientific community, the findings arrive at a moment of intense debate about the validity of LLM-based social simulation. Several recent papers have shown that synthetic samples generated by language models can reproduce survey response patterns with striking fidelity, raising hopes for cheap, scalable, and ethically uncomplicated substitutes for human participants. Other work has cautioned that such fidelity is skin-deep: models reproduce the central tendencies of human populations while flattening minority viewpoints, amplifying majority biases, and failing to capture the contextual sensitivity of real respondents. The new disentangling framework gives this debate a sharper analytical instrument, allowing researchers to state precisely which portion of a simulated social outcome they are willing to trust.

The practical implications extend to platform governance as well. As AI agents increasingly populate recommendation feeds, comment sections, and automated moderation pipelines, the opinions they express are not neutral background noise; they actively shape the informational environment that human users experience. If those agents converge on uniform positions due to shared training biases rather than deliberative processes, they could function as an invisible consensus machine, nudging public discourse in directions no deliberative body ever chose. Understanding the bias-versus-interaction decomposition is therefore not only a matter of methodological hygiene for simulator designers but a governance question for the emerging mixed human–AI public sphere.

The study points toward concrete best practices. Researchers using LLM agents to model opinion dynamics should always run the isolation control: query each agent independently and establish the bias baseline before interpreting any collective behavior. They should diversify model families, prompt formulations, and initial conditions to prevent a single homogeneous prior from dominating the result. And they should report the decomposition explicitly, distinguishing influence-driven convergence from bias-driven agreement, so that downstream readers know whether an emergent consensus in silico says something about social process or merely about the models themselves. As artificial agents become permanent residents of our information ecosystems, tools like this one offer a way to keep the distinction between what agents say to each other and what they were built to say firmly in view.

Subject of Research: Disentangling interaction effects from training biases in the collective opinion dynamics of large language model agents

Article Title: Disentangling interaction and bias effects in opinion dynamics of large language models

Article References: Brockers, V. C., Ehrlich, D. A., & Priesemann, V. (2026). Disentangling interaction and bias effects in opinion dynamics of large language models. Nature Communications, 17(1), Article 10077. https://doi.org/10.1038/s41467-026-77340-3

Image Credits: AI Generated

DOI: 10.1038/s41467-026-77340-3

Keywords: large language models, opinion dynamics, AI agents, training bias, computational social science, agent-based simulation, social influence, polarization, consensus formation, Nature Communications, synthetic respondents, AI governance

Social influence estimation is a cornerstone problem for platforms that rely on word-of-mouth dynamics. Identifying influential users underpins targeted advertising, viral marketing campaigns, and behavior prediction models. Prior work has long suggested that similarity between users is a significant driver of influence: people tend to trust, imitate, and adopt behaviors from those who resemble them. Traditional approaches measured that resemblance through structural metrics such as common neighbors, Jaccard coefficients, or path-based distances within the social graph. More recent methods turned to representation learning, embedding users into continuous vector spaces where proximity reflects similarity. Yet, as the authors point out, these techniques capture only limited facets of user characteristics, typically ignoring the rich geographic dimension that location-based social networks uniquely provide.

The study addresses this gap with two progressive contributions. The first, called NSH, short for Node Embedding and Spatial Homophily, is a hybrid, unsupervised method that linearly combines embedding-based structural similarity with spatial homophily, the well-documented tendency of geographically close users to be socially connected and behaviorally alike. NSH is deliberately simple and interpretable: it takes a user similarity score derived from network embeddings and blends it with a spatial homophily score computed from check-in co-location patterns. The result is a transparent similarity measure that serves as a strong baseline for analyzing the relationship between similarity and influence, and one that practitioners can compute without heavy machinery.

The second and central innovation is GeoSocial2Vec, an end-to-end unified embedding model designed to learn a single representation for each user by intrinsically merging social ties and spatial co-visitation patterns. Rather than computing two separate similarity scores and averaging them, GeoSocial2Vec constructs a heterogeneous graph in which users, social links, and locations coexist as nodes of different types. An adaptive random walk mechanism then traverses this combined graph, generating sequences of nodes that alternate between social hops and spatial transitions. These walks are fed into a skip-gram style embedding procedure, so that the resulting vectors encode both who a user knows and where a user goes within one coherent geometric space. The adaptive component weights different transition types dynamically, allowing the model to balance social and spatial evidence according to the structure of the data rather than a fixed heuristic.

This design reflects a broader shift in representation learning toward heterogeneous graph embeddings, where multiple node and edge types are embedded jointly instead of being projected into separate spaces. In the context of location-based social networks, the payoff is substantial. Two users who never interact directly but repeatedly visit the same neighborhoods at similar times will land close together in the GeoSocial2Vec embedding, capturing a latent affinity that pure network metrics would miss. Conversely, two friends who live in different cities and never share locations will still be pulled together by their social tie, something a purely geographic model would overlook. The embedding, in effect, learns the full socio-spatial signature of each user.

To test whether these richer similarity measures actually track influence, the researchers evaluated the relationship between user similarity and social influence through Pearson correlation and regression analysis on geographically filtered segments of three real-world location-based social networks: Gowalla, Brightkite, and Foursquare. These datasets, drawn from check-in services that were popular in the early 2010s, remain standard benchmarks in this field because they pair explicit friendship graphs with millions of timestamped location records. Filtering the data geographically ensured that the spatial component of the analysis was meaningful, focusing on regions dense enough for co-visitation patterns to be statistically robust.

The results were striking. Both proposed methods, in which spatial homophily is integrated with embedding-based similarity, produced a significantly stronger correlation with observed influence patterns than classical approaches that rely on structure alone. The GeoSocial2Vec framework achieved the highest accuracy across the benchmarks, demonstrating that intrinsically coupled socio-spatial learning outperforms post-hoc fusion of separate signals. In other words, it matters not just that social and geographic information are both considered, but that they are woven together during representation learning itself, allowing the model to capture interactions between the two dimensions that a linear combination cannot express.

At the same time, the authors are careful not to declare NSH obsolete. While GeoSocial2Vec offers the best predictive performance, the NSH framework retains clear advantages in interpretability and computational simplicity. Because NSH is a linear fusion of two comprehensible scores, an analyst can inspect exactly how much of a predicted influence relationship stems from network proximity versus geographic overlap. This transparency matters in business applications where stakeholders must justify targeting decisions, and in settings where training a deep heterogeneous embedding model is impractical due to limited compute or data volume. The two frameworks thus occupy complementary niches: GeoSocial2Vec as the accuracy-oriented state of the art, and NSH as a practical, explainable tool for everyday analysis.

The implications extend well beyond academic benchmarking. For advertisers on check-in-driven platforms, influence estimation that accounts for where people go, not just who they know, could sharpen the selection of seed users in viral campaigns, ensuring that promotional content spreads through communities that are both socially and geographically primed to receive it. For behavior prediction, the fused embeddings offer a way to anticipate adoption of products, venues, or even public health behaviors by modeling the twin channels of social contagion and spatial co-presence. Urban planners and mobility researchers may also find the technique useful, since the same embeddings encode movement regularities that correlate with social structure.

The study also contributes to a long-running scientific conversation about homophily, the principle that similar individuals connect and influence one another. Spatial homophily has been documented repeatedly in location-based social networks, where the probability of friendship decays with physical distance and shared places act as meeting grounds for ties. By demonstrating that this geographic dimension measurably strengthens the link between similarity and influence, the research quantifies how much predictive power has been left on the table by purely structural methods. It suggests that future influence models, and by extension the recommendation and marketing systems built on them, should treat location not as metadata but as a first-class signal in the learning process.

Technically, the work sits at the intersection of network embedding, heterogeneous graph learning, and random-walk-based representation methods. The adaptive random walk at the heart of GeoSocial2Vec echoes the logic of metapath-guided walks in heterogeneous graphs, but with a twist: transition probabilities are tuned to the data, letting the model decide how often to follow social edges versus spatial co-visit links. This adaptivity is what allows a single embedding space to serve both dimensions faithfully, and it is likely the key reason the framework outperformed the linear NSH fusion. The evaluation methodology, combining Pearson correlation with regression analysis across three independent platforms, adds confidence that the findings are not an artifact of one dataset’s quirks.

Limitations and open questions remain, as they do in any study of this kind. The benchmark datasets are aging, and modern platforms may exhibit different spatial dynamics, particularly as privacy changes and check-in behavior declines in some services. Influence itself is a latent quantity, inferred from observed diffusion patterns rather than measured directly, so all correlation results inherit the assumptions of the influence estimation pipeline. The authors also note that the published version is an early-access, peer-reviewed accepted manuscript carrying a permanent DOI, subject to final editorial edits. Still, the core message is robust: when it comes to understanding who moves whom in location-based social networks, knowing both the social graph and the map beats knowing either alone.

As location-aware platforms continue to blend commerce, mobility, and social connection, methods like GeoSocial2Vec point toward a future in which influence is predicted not from a single flattened network, but from the rich, coupled reality of how people relate and where they roam. The research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors, and the authors declare no competing interests.

Subject of Research: Estimating social influence in location-based social networks using embedding-based similarity and spatial homophily

Article Title: Investigating social influence using embedding-based similarity and spatial homophily on the location based social networks

Article References: Akhavan-Hejazi, Z. S., Ghobaei-Arani, M., & Sajjadi, Z. S. (2026). Investigating social influence using embedding-based similarity and spatial homophily on the location based social networks. Journal of Big Data. https://doi.org/10.1186/s40537-026-01546-x

Image Credits: AI Generated

DOI: 10.1186/s40537-026-01546-x

Keywords: social influence, location-based social networks, node embeddings, spatial homophily, GeoSocial2Vec, heterogeneous graph, random walk, Gowalla, Brightkite, Foursquare, representation learning, targeted advertising