Analytics, Information, Operations research insights

Citation: Ahuja, V., Alan, Y., & Arikan, M. (2025). The role of route-level decisions in the efficiency and resilience of airline operations: Evidence from the Wright Amendment repeal. Manufacturing & Service Operations Management. 27 (2): 339–678.

KU Business faculty co-author: Mazhar Arikan

Research objective

This study examines how airlines' route planning and flight scheduling decisions influence passenger travel outcomes. The authors focus on efficiency, measured through scheduled travel times, and resilience, measured through the likelihood of long travel delays.

The study uses the repeal of the Wright Amendment as a natural setting to observe how changes in route offerings affect these outcomes. Rather than focusing solely on whether nonstop service is added, the research seeks to understand why impacts differ across destinations. The goal is to identify route-level features that explain variation in passenger experiences.

What the researchers find

The repeal of the Wright Amendment led to substantial changes in Southwest Airlines' service patterns at Dallas Love Field, but the effects on passengers varied widely by destination. Some routes experienced improved efficiency or fewer long delays, while others saw little change or even worse outcomes. Adding nonstop service alone does not fully explain these differences.

The authors show that route inefficiency and route resilience, two route-level measures capturing itinerary structure and buffer times, help explain the uneven effects. They also find that small adjustments in how connecting passengers are assigned to itineraries can meaningfully reduce long delays with little impact on overall travel time.

Why it matters

The findings highlight that airline performance depends not only on headline decisions like adding nonstop flights, but also on deeper route design choices. For airline managers, the study shows how targeted reallocation of connecting passengers can improve resilience without sacrificing efficiency.

For regulators, the results suggest that passenger-level data can reveal important operational dynamics that are invisible in standard flight-level statistics. More broadly, the study clarifies how efficiency and resilience trade-offs can be actively managed rather than treated as unavoidable. This insight is valuable for improving passenger experiences while maintaining operational performance.

Citation: Banerjee, T., & Sharma, P. (2025). Nonparametric empirical bayes prediction in mixed models: Statistics and Computing, 35(5), 145.

KU Business faculty co-author: Trambak Banerjee

Research objective

This study examines how predictions can be improved in mixed models when the usual assumptions about random effects are unrealistic. Mixed models are commonly used to analyze repeated observations on the same subjects, such as firms or individuals, over time.

Standard prediction methods assume that unobserved subject-level effects follow a Normal distribution with a fixed structure. The authors aim to develop a prediction approach that does not rely on this restrictive assumption. The goal is to produce more accurate predictions when the true distribution of random effects is unknown or irregular.

What the researchers find

The authors introduce a new empirical Bayes prediction framework that estimates subject-specific effects without assuming a specific distribution for them. This approach generalizes the commonly used predictor that is optimal only under Normality. Simulation results show that the proposed method consistently delivers more accurate predictions than existing approaches when the random effects distribution departs from Normality.

The performance gains are especially pronounced in settings with skewed or multimodal unobserved heterogeneity. The method is also shown to extend naturally to more complex models, including dynamic panels and models with multiple sources of unobserved effects.

Why it matters

Many applied researchers rely on mixed models for prediction in fields such as economics, finance, health, and management. When standard distributional assumptions are violated, commonly used predictors can perform poorly without users realizing it. This study provides a practical alternative that improves predictive accuracy while remaining broadly applicable.

By reducing reliance on strong modeling assumptions, the approach helps analysts generate more reliable subject-level predictions. This insight is particularly valuable in settings where predictions inform decision-making, such as forecasting firm performance or financial outcomes.

Citation: Gang, B., & Banerjee, T. (2025). Large-scale multiple testing of composite null hypotheses under heteroskedasticity. Biometrika. 112 (2): asaf007.

KU Business faculty co-author: Trambak Banerjee

Research objective

This study examines how to conduct large-scale multiple-hypothesis testing when test statistics exhibit different levels of variability. In many applications, the variance of a test statistic differs across units and may also be related to the underlying parameter being tested. Common practices such as standardizing test statistics can unintentionally reduce power or distort error control in these settings.

The authors aim to design a testing procedure that explicitly accounts for heteroskedasticity without collapsing the data through rescaling. The objective is to maintain valid false discovery rate control while improving detection power.

What the researchers find

The authors develop a new multiple testing procedure that directly incorporates variance information into the testing process. Rather than standardizing test statistics, the method uses a nonparametric empirical Bayes deconvolution approach to model the joint behavior of the signal and its variance.

Theoretical results show that the proposed procedure achieves valid and asymptotically optimal false discovery rate control. Simulation evidence demonstrates that the method delivers substantially higher power than existing approaches across a wide range of heteroskedastic settings. An application to mobile game data shows that the approach improves the detection of highly engaged users.

Why it matters

Large-scale testing problems arise in economics, finance, biology, and digital platforms, where variability often differs systematically across observations. Methods that ignore heteroskedasticity or handle it through simple standardization can miss meaningful signals or produce misleading inferences. This study offers a practical framework that uses available variance information to improve discovery while preserving statistical validity.

The results guide researchers and practitioners who need reliable multiple-testing tools in complex, data-rich environments. By improving power without sacrificing error control, the approach enhances decision-making based on large-scale statistical evidence.

Citation: Luo, J., Banerjee, T., Mukherjee, G., & Sun, W. (in press). Empirical Bayes estimation with side information: A nonparametric integrative Tweedie approach. Statistica Sinica.

KU Business faculty co-author: Trambak Banerjee

Research objective

This study examines how to improve the estimation of compound means from many normal distributions when rich side information is available for each unit. Standard empirical Bayes methods typically rely only on the primary outcomes and often ignore auxiliary data that may contain valuable structural information.

The authors aim to develop a flexible empirical Bayes approach that integrates multivariate side information in a principled way, without imposing parametric assumptions on the underlying distributions. The goal is to achieve more accurate estimation in high-dimensional settings where conventional methods become inefficient.

What the researchers find

The authors develop a nonparametric integrative Tweedie (NIT) approach that extends classical Tweedie-based empirical Bayes estimation to settings with side information. The method directly estimates the gradient of the log marginal density via convex optimization, allowing structural constraints from auxiliary data to be incorporated naturally. Theoretical results characterize the asymptotic risk of the estimator and show that NIT converges to the oracle estimator at an explicit rate.

The analysis also shows a clear trade-off: higher-dimensional side information can reduce estimation risk while slowing convergence. Simulation studies and real data applications demonstrate that NIT consistently outperforms existing empirical Bayes and integrative methods in terms of estimation accuracy.

Why it matters

Many modern applications, such as genomics, neuroimaging, finance, and large-scale experimentation, generate outcome data alongside rich auxiliary information. Ignoring this side information can lead to inefficient or biased estimates, especially when the number of parameters is large.

This study provides a general, nonparametric framework for integrating auxiliary data into empirical Bayes estimation while maintaining theoretical guarantees. The results offer practical guidance for researchers seeking to exploit side information to improve inference, and they broaden the scope of empirical Bayes methods in data-rich environments.

Citation: Sharma, P., & Banerjee, T. (in press). Do financial regulators act in the public's interest? A Bayesian latent class estimation framework for assessing regulatory responses to banking crises. Journal of the Royal Statistical Society: Series A. 

KU Business faculty co-author: Trambak Banerjee

Research objective

This study examines whether financial regulators act in the public interest when resolving bank failures during systemic crises. Regulators face a core trade-off between preserving short-run financial stability, which may require bailouts, and limiting moral hazard, which often favors liquidation.

The authors aim to evaluate regulators’ overall decision frameworks, rather than isolated, high-profile cases, by comparing observed resolution decisions with decision rules recommended by economic theory. A key challenge addressed is that regulators’ internal classification of economic conditions, which determines which rule they apply, is unobserved.

What the researchers find

The authors develop a Bayesian latent-class estimation framework that models regulators’ unobserved assignment of failed banks to distinct economic states, such as high versus low economic distress. Each latent class is associated with a different resolution rule, linking bank characteristics to the probability of bailout, sale, or liquidation. Applying the framework to U.S. banking crises in the 1980s, the study finds sharp contrasts across regulators. The Federal Deposit Insurance Corporation’s decisions closely align with theoretical recommendations, assisting primarily during periods of economic and industry-wide distress.

In contrast, the Federal Savings and Loan Insurance Corporation did not systematically distinguish between high- and low-distress environments and instead appeared more responsive to political considerations. These deviations are consistent with the FSLIC’s eventual insolvency and the significant taxpayer losses it incurred.

Why it matters

Public debates about bank bailouts often focus on individual decisions, rather than the broader decision logic regulators follow. This study provides a disciplined statistical framework for evaluating whether regulators’ overall behavior aligns with public interest objectives.

The approach allows policymakers, researchers, and oversight bodies to detect structural weaknesses in regulatory decision-making before crises become fiscally catastrophic. Beyond banking regulation, the framework is broadly applicable to settings where decision rules vary across unobserved states, such as healthcare policy, consumer protection, and financial supervision.

Citation: Lee Y., Chen, A. N. K., & Wang, W. (2025). Push it cross the finish line – Designing online interface to induce choice closure at the post-decision pre-purchase stage. Information Systems Research, 36(3), 1821-1845.

KU Business faculty co-author: Andrew N. K. Chen

Research objective

This study examines why consumers often abandon online purchases after selecting a product but before completing the transaction. The authors focus on the post-decision, pre-purchase stage, when consumers have chosen but have not yet purchased.

The research seeks to understand whether online interface designs can reduce consumers’ lingering doubts and help them feel that their decision is complete. The goal is to explain how interface cues can foster a sense of choice closure and increase satisfaction with the decision. The study emphasizes reducing dissonance rather than only boosting positive evaluations.

What the researchers find

The authors show that consumers frequently experience cognitive dissonance after making an online choice, driven by continued comparison with forgone alternatives. Carefully designed interface cues can reduce this dissonance and increase perceived choice closure.

Both direct reinforcement cues, such as messages affirming the chosen option, and social reinforcement cues, such as signals that others made similar choices, are effective. Reduced cognitive dissonance increases perceived choice closure, which in turn increases decision satisfaction. These effects occur even though consumers cannot test or consume the product at this stage.

Why it matters

Online retailers lose substantial revenue when consumers abandon shopping carts after choosing a product. This research highlights the importance of addressing consumers’ unsettled thoughts after choice, rather than focusing only on making products or deals more attractive.

The findings offer actionable guidance for designing online interfaces that help consumers feel confident and settled in their decisions. More broadly, the study extends the concept of choice closure from physical retail settings to digital environments and shows how decision-making can be shaped through interface design.

Citation: Dey, D., Lahiri, A., & Mukherjee, R. (2025). Polarization or bias: Take your click on social media. Journal of the Association for Information Systems. 26 (3): 850–878.

KU Business faculty co-author: Debabrata "Deb" Dey

Research objective

This study examines whether profit-driven social media platforms have economic incentives to create polarization or inject bias among users. The authors aim to understand how a platform’s user-targeting and content-recommendation strategies shape social divisions when engagement drives advertising revenue.

A central goal is to analyze how polarization and bias enter the platform’s profit calculus. The study also explores how policymakers might intervene, and whether such interventions produce unintended consequences.

What the researchers find

Using a microeconomic model of a social media platform with heterogeneous users and content providers, the authors find that profit motives can lead platforms to adopt strategies that increase polarization. Polarization and bias both raise engagement and profits, but they function as substitutes in the platform’s profit function.

When unconstrained, platforms tend to favor polarization rather than bias. However, when policymakers impose penalties or regulations aimed at reducing polarization, platforms may respond by shifting toward biased content strategies. The results also show that higher public awareness of the social costs of polarization and bias can reduce platforms’ incentives to rely on either mechanism.

Why it matters

This research clarifies why regulating social media platforms is more complex than simply penalizing harmful outcomes. Interventions aimed at reducing polarization may inadvertently increase bias, potentially worsening social divisions in different ways.

The findings suggest that public education and awareness can play an essential role alongside regulation in shaping platform behavior. For policymakers, the study highlights the need to account for strategic platform responses when designing policies intended to protect social welfare.

Citation: Dey, D., & Lahiri, A. (in press). "Extortionality" in ransomware attacks: A microeconomic study of extortion and externality. Information Systems Research.

KU Business faculty co-author: Debabrata "Deb" Dey

Research objective

This study examines ransomware payments as a source of a negative externality that increases future cyberattack risk for all organizations. The authors aim to understand how firms’ decisions to pay or not pay ransoms affect attackers’ incentives over time.

A central objective is to analyze whether standard policy tools, such as taxes, subsidies, or bans on ransom payments, can mitigate this externality. The study also explores how outcomes depend on whether ransomware attackers behave strategically or nonstrategically.

What the researchers find

The authors develop a multiperiod game in which firms face repeated ransomware threats and must decide whether to pay ransoms, knowing their decisions affect future attack probabilities. Paying ransom benefits the breached firm in the short run but increases attackers' incentives to attack, raising risks for all firms.

When attackers are nonstrategic and set ransom demands exogenously, fiscal interventions such as taxes or subsidies can reduce payments and mitigate the externality. However, when attackers are strategic and adjust ransom demands in response to policy, fiscal tools may backfire by increasing firms’ willingness to pay. In such cases, a ban on ransom payments becomes the only effective policy lever to reduce the externality.

Why it matters

Ransomware has become a significant threat to firms, governments, and national security, yet there is little consensus on whether banning ransom payments is desirable or effective. This study shows that policy effectiveness critically depends on attacker behavior, a factor often ignored in standard externality analysis.

The findings caution policymakers against relying solely on fiscal interventions without considering attackers' strategic responses. More broadly, the concept of “extortionality” provides a framework for evaluating policies in settings where externalities arise from extortion rather than conventional market activity.

Citation: Amaya, J. & Reed, S. (2025). Space management policy for urban last-mile parking infrastructure: A demand-oriented approach. Transportation Research Part E: Logistics and Transportation Review, 200:104185.

KU Business faculty co-author: Sara Reed

Research objective

This study examines how urban on-street loading zones for last-mile freight deliveries can be managed more efficiently by designing parking policies that explicitly account for delivery driver behavior and demand patterns.

The authors aim to evaluate whether flexible parking sessions, reservations, and pricing schemes outperform the common first-come, first-served policy with fixed 60-minute parking sessions, in terms of utilization, illegal parking, and revenue.

What the researchers find

Using a Markov Decision Process to model a loading-zone reservation system, the authors show that current static 60-minute parking sessions result in low efficiency, with only about 27% of awarded parking minutes actually used. Introducing flexible parking sessions that better match observed delivery durations substantially improves utilization, raising it to about 70% of awarded minutes on average.

Flexible sessions combined with reservations can further improve space management without reducing overall revenue. When paired with increasing block pricing and reservation holding fees, the proposed system generates revenues comparable to current policies while using curb space more effectively. Extending the model with a multinomial logit framework, the authors find that flexible policies also reduce illegal parking by making legal loading zones more attractive to drivers.

Why it matters

Urban freight parking shortages contribute to congestion, emissions, delivery inefficiencies, and illegal parking. This study shows that better policy design, rather than costly infrastructure expansion, can significantly improve curbside performance.

By aligning parking rules with real driver behavior, cities can increase utilization, reduce externalities, and maintain revenue neutrality. The findings provide actionable guidance for planners and regulators seeking demand-oriented, behaviorally informed solutions for last-mile logistics in dense urban environments.

Citation: Reed, S. (in press). Parking in routing last-mile deliveries. To appear in Encyclopedia in Operations Management. Elsevier. 

KU Business faculty author: Sara Reed

Research objective

This chapter examines how parking decisions can be systematically incorporated into last-mile delivery routing models. The goal is to capture the operational trade-offs between vehicle movement and walking-based deliveries by explicitly modeling parking as part of the routing problem, rather than treating it as an exogenous constraint.

What the author finds

The chapter conceptualizes last-mile delivery routes as a two-echelon system. The first echelon represents vehicle routing decisions, including where and when to park. In contrast, the second echelon represents walking routes used by delivery personnel to serve customers from a given parking location. The author reviews three deterministic modeling approaches for integrating parking into routing decisions.

The cluster-first route-second approach groups customers to be served from the same parking location before optimizing vehicle routes. The weighted activity-based approach assigns relative weights to driving and walking activities to balance efficiency trade-offs. The parking time cost-based approach directly incorporates expected parking search time, driving time, and walking time into routing cost functions.

Each approach highlights different ways to trade off longer walking distances against the cost and disruption of moving the vehicle to find new parking.

Why it matters

Parking availability is a critical and often binding constraint in urban last-mile delivery, affecting travel time, congestion, and delivery efficiency. By formalizing parking as an integral component of routing decisions, this work lays the foundation for more realistic, implementable delivery optimization models.

The chapter offers researchers and practitioners structured ways to account for curb access and walking deliveries, helping logistics operators design routes that better reflect real-world urban conditions and constraints.

Citation: Sherwood, Ben, Li, Shaobo, and Maidman, Adam. (2025). rqPen: An R Package for Penalized Quantile Regression. The R Journal. 17(2): 146-175.

KU Business faculty co-authors: Ben Sherwood and Shaobo Li

Research objective

Quantile regression directly models a conditional quantile of interest. The R package rqPen provides penalized quantile regression estimation using the lasso, elastic net, adaptive lasso, SCAD, and MCP penalties.

It also provides extensions to group penalties, both for groups of predictors and grouping variable selection across quantiles. The paper presents the different approaches to penalized quantile regression and how to use the provided methods in rqPen.

What the researchers find

The authors demonstrate how to use rqPen, a package that supports lasso, elastic net, adaptive lasso, SCAD, MCP, and several group-based penalties, including penalties that enforce consistent variable selection across quantiles.

The package implements both linear programming methods and Huber-type approximations of the quantile loss to address the non-differentiability of the check function and the computational challenges posed by large data sets. The package, rqPen allows users to estimate multiple quantiles in a single call, select tuning parameters using cross-validation or information criteria, and visualize how coefficient estimates change across quantiles and penalty levels.

Why it matters

Quantile regression is a popular approach for robust estimation and modeling heterogenous effects. Both issues are important in high-dimensional analysis. A common approach for high-dimensional analysis is to use penalized methods. The package rqPen provides users with the ability to estimate quantile regression models using a variety of penalties that are available for mean regression methods.

In addition, it incorporates penalties that are specific to quantile regression, specifically a group penalty across quantiles. Finally, the package provides users with Huber-type approximations to the quantile loss which decreases the computational cost compared to using the non-differentiable quantile loss.

Citation: Jiang, J., Bandeli, K. K., & Srinivasan K. (2025). Dynamic model selection in enterprise forecasting systems using sequence modeling. Decision Support Systems. 193: 114439.

KU Business faculty co-author: Karthik Srinivasan

Research objective

This study develops a new approach to dynamic model selection in large-scale enterprise forecasting systems, where different forecasting models may perform better at various times. The objective is to formalize model flipping and introduce a scalable framework that automatically selects and updates forecasting models for each time series as data patterns evolve.

What the researchers find

The authors propose a framework, TimeSpeaks, that reframes dynamic model selection as a sequence-prediction problem akin to text completion in natural language processing. Instead of relying on exogenous inputs or full-feature sets used by candidate forecasting models, TimeSpeaks learns from historical sequences of the best-performing models. The framework is implemented using two sequence models, a BiLSTM-based approach and a transformer-based approach called TimeXer.

Across public benchmarks, including the M4 competition and store sales data, as well as two retail case studies, TimeSpeaks consistently outperforms traditional ensembles, Bayesian model averaging, direct-loss estimators, and several state-of-the-art global forecasting models. The results show especially strong gains in long-horizon forecasting and settings with heterogeneous time series.

Why it matters

Enterprise forecasting systems must operate at scale while adapting to changing demand patterns, seasonality, and market conditions. This study shows that dynamic model selection can be improved by focusing on the temporal evolution of model performance rather than repeatedly retraining or tuning models using full data streams. TimeSpeaks offers a practical, scalable solution that aligns well with enterprise MLOps workflows and reduces reliance on exogenous information.

For organizations managing thousands of forecasts simultaneously, the approach provides a structured way to improve accuracy, robustness, and long-term planning performance.

Citation: Rao S., Juma N., Srinivasan K. (2025). Textual analysis of sustainability reports: Topics, firm value, and the moderating role of assurance. Journal of Risk and Financial Management. 18(8), 463.

KU Business faculty co-author: Karthik Srinivasan

Research objective

This study examines how the specific topics firms emphasize in standalone sustainability reports relate to firm value and whether third-party assurance changes that relationship. Rather than treating sustainability reporting as a single aggregate construct, the authors focus on the content of disclosures and the role of assurance in shaping market responses.

What the researchers find

Using latent Dirichlet allocation, the authors identify six dominant sustainability topics in U.S. firms’ sustainability reports: environmental impact, sustainable consumption, daily necessities, socio-economic impact, healthcare, and operations. Panel regressions show that disclosures of healthcare and daily necessities are associated with immediate and persistent increases in firm value, as measured by Tobin’s Q.

In contrast, environmental and socio-economic impact topics are linked to positive effects that materialize with a lag, typically about two years after disclosure. The moderating role of assurance is mixed. In some cases, assurance strengthens the valuation effect by enhancing credibility, but in others, it is associated with lower firm value, particularly for environmental and socio-economic topics. Additional analyses using discrete Bayesian networks and SHAP values confirm that these relationships are non-linear and context-dependent.

Why it matters

The findings highlight that not all sustainability disclosures are valued equally by the market, and that the timing of valuation effects depends on the disclosed topic. The study also shows that assurance is not uniformly beneficial, suggesting a dual signaling effect in which assurance can both increase credibility and intensify investor scrutiny.

For managers, the results emphasize the importance of strategic topic selection, consistency in sustainability reporting, and careful consideration of when and how to apply assurance. For researchers and regulators, the study underscores the need to look beyond aggregate ESG measures and consider the content and credibility of disclosures when assessing their economic consequences.

Citation: Fisher, A., Srinivasan, K., Hillier, S. and Mago, V., 2025. HEAL-Summ: a lightweight and ethical framework for accessible summarization of health information. Frontiers in Public Health, 13, p.1619274.

KU Business faculty co-author: Karthik Srinivasan

Research objective

This study examines how large language models can be used to make complex health-related news more accessible to the general public. The authors aim to develop a summarization framework that reduces information overload while remaining sensitive to ethical concerns, including clarity, tone, and potential harm.

The focus is on creating summaries that are easier to read and emotionally appropriate without relying on expensive or resource-intensive technology. More broadly, the study seeks to understand how automated tools can support more transparent public health communication in constrained settings.

What the researchers find

The researchers find that lightweight language models can produce health news summaries that closely align with the meaning of the original articles while improving readability. Different models show distinct strengths, with some generating simpler, more accessible language and others offering greater emotional range and vocabulary variety. Across models, summaries generally remain consistent in meaning, suggesting limited risk of distortion.

The evaluation results also indicate that automated checks can help identify differences in tone, emotional alignment, and potentially harmful content. Overall, the framework demonstrates that low-resource models can balance accessibility, expressiveness, and ethical considerations.

Why it matters

The study provides insight into how artificial intelligence can help the public better understand dense, difficult-to-interpret health information. This insight is especially relevant for communities facing barriers to access, limited digital resources, or high-stakes health decisions.

By showing that lower-cost tools can still support ethical and accessible communication, the research informs practitioners and policymakers interested in responsible technology use in public health. The findings also highlight the importance of evaluating not only accuracy but also readability and emotional tone when summarizing health information for broad audiences.

Citation: Srinivasan, K., Currim, F. and Ram, S., 2025. A Reduced Modeling Approach for Making Predictions with Incomplete Data Having Blockwise Missing Patterns. INFORMS Journal on Data Science, 4(1), pp.85-99.

KU Business faculty co-author: Karthik Srinivasan

Research objective

This study examines how predictive models can be built when datasets contain large, structured gaps rather than a few scattered missing values. The authors focus on situations in which missing data occur in blocks, such as when information comes from multiple sources or from surveys with optional sections. They aim to develop an approach that avoids discarding large amounts of data or imputing missing values with estimates.

More broadly, the study seeks to understand how models can make reliable predictions using only the information that is actually observed.

What the researchers find

The researchers find that their proposed blockwise reduced modeling approach delivers stronger predictive performance than common alternatives. By training multiple models on overlapping subsets of available data and matching new observations to the most appropriate model, the approach limits information loss. The method performs well across both simulated and real datasets with blockwise missing patterns.

Results show improvements for both simpler and more complex prediction models, and the approach scales efficiently as data size grows. Overall, predictions remain stable even when the missing data are substantial.

Why it matters

Many real-world datasets contain large, systematic gaps that make standard analysis difficult or unreliable. This study offers practical insight into how organizations can still extract useful predictions without relying heavily on data deletion or imputation.

The findings are relevant for analysts and decision-makers working with survey data, integrated databases, or health and social data sources. By preserving more of the original information, the approach supports better decision-making in settings where complete data are rarely available.

Citation: Ortiz, Jose, Myers, Michael D., and Tripathi, Arvind K., 2025. Promoting societal development through digital activism: A case study of a Guatemalan tragedy. Journal of Information Technology

KU Business faculty co-author: Arvind Tripathi

Research objective

This study examines how activist organizations use digital communication technologies to mobilize public support and promote broader societal values in settings where traditional media and political expression are constrained. The authors focus on how activists communicate about a major public tragedy in a politically repressive environment.

The goal is to understand how online activism helps shape shared meanings around social issues and encourages collective action. Rather than evaluating the success of a specific protest, the study seeks to clarify how digital platforms can support value formation related to societal development.

What the researchers find

The study finds that activist organizations used social media to circulate messages that differed from official or dominant accounts of the tragedy. These alternative messages highlighted shared responsibility and the need for public accountability. Over time, activists combined diverse viewpoints into more unified narratives that reinforced common values. This combination of presenting alternative perspectives and then aligning them into a shared message helped amplify public attention and engagement.

The findings suggest that digital platforms supported both the expression of dissent and the consolidation of collective meaning, primarily where traditional channels were restricted.

Why it matters

The study helps clarify how digital activism can function in environments where freedom of expression is limited. It shows that online platforms can provide spaces for citizens to challenge official narratives and coordinate around shared values. This insight is relevant for policymakers, civil society groups, and observers interested in how technology affects civic participation and accountability.

The findings also offer a clearer understanding of how digital communication can contribute to societal development without relying on formal political institutions. More broadly, the research informs discussions about the role of technology in supporting democratic engagement under challenging conditions.