arXiv subjects

Prasanta Bhattacharya

Publications and source records attributed to Prasanta Bhattacharya.

14 recordsLinked to original sources

arXiv2026

ValueGraph: Value-Signal Guided Graph Pre-training for Contextualized User Representation

Value signals are aggregated user-level moral representations that capture users' inferred value-related tendencies from their online discourse. User behavior on social media is shaped not only by what users say or whom they interact with, but also by the value signal through which they express attitudes. Existing user representation methods largely miss this value-relevant dimension. We propose ValueGraph, a graph pre-training framework that uses automatically inferred moral-value signals as noisy auxiliary signals for contextualized user representation. From post-reply graphs, ValueGraph learns semantic and structural representations and further aligns users through relative value similarity with contrastive and clustering objectives. Rather than treating inferred values as gold psychological labels, ValueGraph uses them as soft constraints for representation learning. Experiments on stance detection and twitter bot detection show consistent gains over strong text-based, graph-based, and text-only LLM baselines, highlighting value-signal guidance as a useful inductive bias for socially informed user modeling.

cs.CL↗

arXiv2025

When Incentives Backfire, Data Stops Being Human

Progress in AI has relied on human-generated data, from annotator marketplaces to the wider Internet. However, the widespread use of large language models now threatens the quality and integrity of human-generated data on these very platforms. We argue that this issue goes beyond the immediate challenge of filtering AI-generated content -- it reveals deeper flaws in how data collection systems are designed. Existing systems often prioritize speed, scale, and efficiency at the cost of intrinsic human motivation, leading to declining engagement and data quality. We propose that rethinking data collection systems to align with contributors' intrinsic motivations -- rather than relying solely on external incentives -- can help sustain high-quality data sourcing at scale while maintaining contributor trust and long-term participation.

cs.CY↗

arXiv2025

Investigating the effectiveness of multimodal data in forecasting SARS-COV-2 case surges

The COVID-19 pandemic response relied heavily on statistical and machine learning models to predict key outcomes such as case prevalence and fatality rates. These predictions were instrumental in enabling timely public health interventions that helped break transmission cycles. While most existing models are grounded in traditional epidemiological data, the potential of alternative datasets, such as those derived from genomic information and human behavior, remains underexplored. In the current study, we investigated the usefulness of diverse modalities of feature sets in predicting case surges. Our results highlight the relative effectiveness of biological (e.g., mutations), public health (e.g., case counts, policy interventions) and human behavioral features (e.g., mobility and social media conversations) in predicting country-level case surges. Importantly, we uncover considerable heterogeneity in predictive performance across countries and feature modalities, suggesting that surge prediction models may need to be tailored to specific national contexts and pandemic phases. Overall, our work highlights the value of integrating alternative data sources into existing disease surveillance frameworks to enhance the prediction of pandemic dynamics.

q-bio.QM↗

arXiv2025

What Contributes to Affective Polarization in Networked Online Environments? Evidence from an Agent-Based Model

Affective polarization, or, inter-party hostility, is increasingly recognized as a pervasive issue in democracies worldwide, posing a threat to social cohesion. The digital media ecosystem, now widely accessible and ever-present, has often been implicated in accelerating this phenomenon. However, the precise causal mechanisms responsible for driving affective polarization have been a subject of extensive debate. While the concept of echo chambers, characterized by individuals ensconced within like-minded groups, bereft of counter-attitudinal content, has long been the prevailing hypothesis, accumulating empirical evidence suggests a more nuanced picture. This study aims to contribute to the ongoing debate by employing an agent-based model to illustrate how affective polarization is either fostered or hindered by individual news consumption and dissemination patterns based on ideological alignment. To achieve this, we parameterize three key aspects: (1) The affective asymmetry of individuals' engagement with in-party versus out-party content, (2) The proportion of in-party members within one's social neighborhood, and (3) The degree of partisan bias among the elites within the population. Subsequently, we observe macro-level changes in affective polarization within the population under various conditions stipulated by these parameters. This approach allows us to explore the intricate dynamics of affective polarization within digital environments, shedding light on the interplay between individual behaviors, social networks, and information exposure.

cs.MA↗

arXiv2025

Rethinking stance detection: A theoretically-informed research agenda for user-level inference using language models

Stance detection has emerged as a popular task in natural language processing research, enabled largely by the abundance of target-specific social media data. While there has been considerable research on the development of stance detection models, datasets, and application, we highlight important gaps pertaining to (i) a lack of theoretical conceptualization of stance, and (ii) the treatment of stance at an individual- or user-level, as opposed to message-level. In this paper, we first review the interdisciplinary origins of stance as an individual-level construct to highlight relevant attributes (e.g., psychological features) that might be useful to incorporate in stance detection models. Further, we argue that recent pre-trained and large language models (LLMs) might offer a way to flexibly infer such user-level attributes and/or incorporate them in modelling stance. To better illustrate this, we briefly review and synthesize the emerging corpus of studies on using LLMs for inferring stance, and specifically on incorporating user attributes in such tasks. We conclude by proposing a four-point agenda for pursuing stance detection research that is theoretically informed, inclusive, and practically impactful.

cs.CL↗

arXiv2024

Estimating Policy Effects in a Social Network with Independent Set Sampling

Evaluating the impact of policy interventions on respondents who are embedded in a social network is often challenging due to the presence of network interference within the treatment groups, as well as between treatment and non-treatment groups throughout the network. In this paper, we propose a novel empirical strategy that combines network sampling based on the identification of independent sets with a stochastic actor-oriented model (SAOM) to infer the direct and net effects of a policy. By assigning respondents from an independent set to the treatment, we are able to block direct spillover of the treatment among the treated respondents for an extended period of time, during which the direct effect of the treatment can be isolated from the associated network interference. We empirically demonstrate this using a simulation-based evaluation of a fictitious policy implementation using both real-life and generated networks, and use a counterfactual approach to estimate the treatment effect of the policy. Our results highlight the effectiveness of our proposed empirical strategy, and notably, the role of network sampling techniques in influencing the evaluation of policy effects. The findings from this study have the potential to help researchers and policymakers with planning, designing, and anticipating policy responses in a networked society.

cs.SI↗

arXiv2024

From traces to measures: Large language models as a tool for psychological measurement from text

Large language models are increasingly being used to label or rate psychological features in text data. This approach helps address one of the limiting factors of digital trace data - their lack of an inherent target of measurement. However, this approach is also a form of psychological measurement (using observable variables to quantify a hypothetical latent construct). As such, these ratings are subject to the same psychometric considerations of reliability and validity as more standard psychological measures. Here we present a workflow for developing and evaluating large language model based measures of psychological features which incorporate these considerations. We also provide an example, attempting to measure the previously established constructs of attitude certainty, importance and moralization from text. Using a pool of prompts adapted from existing measurement instruments, we find they have good levels of internal consistency but only partially meet validity criteria.

cs.HC↗

arXiv2024

Enhancing Stance Classification on Social Media Using Quantified Moral Foundations

This study enhances stance detection on social media by incorporating deeper psychological attributes, specifically individuals' moral foundations. These theoretically-derived dimensions aim to provide a comprehensive profile of an individual's moral concerns which, in recent work, has been linked to behaviour in a range of domains, including society, politics, health, and the environment. In this paper, we investigate how moral foundation dimensions can contribute to predicting an individual's stance on a given target. Specifically we incorporate moral foundation features extracted from text, along with message semantic features, to classify stances at both message- and user-levels using both traditional machine learning models and large language models. Our preliminary results suggest that encoding moral foundations can enhance the performance of stance detection tasks and help illuminate the associations between specific moral foundations and online stances on target topics. The results highlight the importance of considering deeper psychological attributes in stance analysis and underscores the role of moral foundations in guiding online social behavior.

cs.CL↗

arXiv2024

Predicting User Stances from Target-Agnostic Information using Large Language Models

We investigate Large Language Models' (LLMs) ability to predict a user's stance on a target given a collection of his/her target-agnostic social media posts (i.e., user-level stance prediction). While we show early evidence that LLMs are capable of this task, we highlight considerable variability in the performance of the model across (i) the type of stance target, (ii) the prediction strategy and (iii) the number of target-agnostic posts supplied. Post-hoc analyses further hint at the usefulness of target-agnostic posts in providing relevant information to LLMs through the presence of both surface-level (e.g., target-relevant keywords) and user-level features (e.g., encoding users' moral values). Overall, our findings suggest that LLMs might offer a viable method for determining public stances towards new topics based on historical and target-agnostic data. At the same time, we also call for further research to better understand LLMs' strong performance on the stance prediction task and how their effectiveness varies across task contexts.

cs.CL↗

arXiv2022

A Discussion on Building Practical NLP Leaderboards: The Case of Machine Translation

Recent advances in AI and ML applications have benefited from rapid progress in NLP research. Leaderboards have emerged as a popular mechanism to track and accelerate progress in NLP through competitive model development. While this has increased interest and participation, the over-reliance on single, and accuracy-based metrics have shifted focus from other important metrics that might be equally pertinent to consider in real-world contexts. In this paper, we offer a preliminary discussion of the risks associated with focusing exclusively on accuracy metrics and draw on recent discussions to highlight prescriptive suggestions on how to develop more practical and effective leaderboards that can better reflect the real-world utility of models.

cs.CL↗

arXiv2022

Task Preferences across Languages on Community Question Answering Platforms

With the steady emergence of community question answering (CQA) platforms like Quora, StackExchange, and WikiHow, users now have an unprecedented access to information on various kind of queries and tasks. Moreover, the rapid proliferation and localization of these platforms spanning geographic and linguistic boundaries offer a unique opportunity to study the task requirements and preferences of users in different socio-linguistic groups. In this study, we implement an entity-embedding model trained on a large longitudinal dataset of multi-lingual and task-oriented question-answer pairs to uncover and quantify the (i) prevalence and distribution of various online tasks across linguistic communities, and (ii) emerging and receding trends in task popularity over time in these communities. Our results show that there exists substantial variance in task preference as well as popularity trends across linguistic communities on the platform. Findings from this study will help Q&A platforms better curate and personalize content for non-English users, while also offering valuable insights to businesses looking to target non-English speaking communities online.

cs.CL↗

arXiv2021

Exploring the contextual factors affecting multimodal emotion recognition in videos

Emotional expressions form a key part of user behavior on today's digital platforms. While multimodal emotion recognition techniques are gaining research attention, there is a lack of deeper understanding on how visual and non-visual features can be used to better recognize emotions in certain contexts, but not others. This study analyzes the interplay between the effects of multimodal emotion features derived from facial expressions, tone and text in conjunction with two key contextual factors: i) gender of the speaker, and ii) duration of the emotional episode. Using a large public dataset of 2,176 manually annotated YouTube videos, we found that while multimodal features consistently outperformed bimodal and unimodal features, their performance varied significantly across different emotions, gender and duration contexts. Multimodal features performed particularly better for male speakers in recognizing most emotions. Furthermore, multimodal features performed particularly better for shorter than for longer videos in recognizing neutral and happiness, but not sadness and anger. These findings offer new insights towards the development of more context-aware emotion recognition and empathetic systems.

cs.MM↗

arXiv2018

A Co-evolution Model of Network Structure and User Behavior in Online Social Networks: The Case of Network-Driven Content Generation

With the rapid growth of online social network sites (SNS), it has become imperative for platform owners and online marketers to investigate what drives content production on these platforms. However, previous research has found it difficult to statistically model these factors from observational data due to the inability to separately assess the effects of network formation and network influence. In this paper, we adopt and enhance an actor-oriented continuous-time model to jointly estimate the co-evolution of the users' social network structure and their content production behavior using a Markov Chain Monte Carlo (MCMC)- based simulation approach. Specifically, we offer a method to analyze non-stationary and continuous behavior with network effects in the presence of observable and unobservable covariates, similar to what is observed in social media ecosystems. Leveraging a unique dataset from a large social network site, we apply our model to data on university students across six months to find that: 1) users tend to connect with others that have similar posting behavior, 2) however, after doing so, users tend to diverge in posting behavior, and 3) peer influences are sensitive to the strength of the posting behavior. Further, our method provides researchers and practitioners with a statistically rigorous approach to analyze network effects in observational data. These results provide insights and recommendations for SNS platforms to sustain an active and viable community.

cs.SI↗

arXiv2017

Characterizing and Predicting Supply-side Engagement on Crowd-contributed Video Sharing Platforms

Video sharing and entertainment websites have rapidly grown in popularity and now constitute some of the most visited websites on the Internet. Despite the active user engagement on these online video-sharing platforms, most of recent research on online media platforms have restricted themselves to networking based social media sites, like Facebook or Twitter. We depart from previous studies in the online media space that have focused exclusively on demand-side user engagement, by modeling the supply-side of the crowd-contributed videos on this platform. The current study is among the first to perform a large-scale empirical study using longitudinal video upload data from a large online video platform. The modeling and subsequent prediction of video uploads is made complicated by the heterogeneity of video types (e.g. popular vs. niche video genres), and the inherent time trend effects associated with media uploads. We identify distinct genre-clusters from our dataset and employ a self-exciting Hawkes point-process model on each of these clusters to fully specify and estimate the video upload process. Additionally, we go beyond prediction to disentangle potential factors that govern user engagement and determine the video upload rates, which improves our analysis with additional explanatory power. Our findings show that using a relatively parsimonious point-process model, we are able to achieve higher model fit, and predict video uploads to the platform with a higher accuracy than competing models. The findings from this study can benefit platform owners in better understanding how their supply-side users engage with their site over time. We also offer a robust method for performing media upload prediction that is likely to be generalizable across media platforms which demonstrate similar temporal and genre-level heterogeneity.

cs.HC↗