An explainable dual-attention transformer for predicting the sociocultural impact of global sports events

Why big sports moments matter beyond the scoreboard

When billions of people tune in to the Olympics or the World Cup, the effects go far beyond medals and TV ratings. These mega‑events shape how nations see themselves, how they are seen by others, and how issues like gender equality, diversity, and environmental responsibility move into the global spotlight. This paper introduces a new artificial‑intelligence tool that tries to measure those hard‑to‑grasp social ripples as a single, comparable score, offering governments, organizers, and the public a clearer picture of what we really gain—or lose—from the world’s biggest games.

Turning complex social reactions into one clear score

The authors start by defining a “Social Impact Score” for each major sports event held worldwide between 2000 and 2024. Instead of looking only at money or TV audiences, they combine multiple dimensions: cultural engagement, diversity, gender representation, public mood, environmental sustainability, and media attention. Each of these ingredients is first normalized so that very different kinds of data—such as survey sentiment, social media activity, or green‑event indicators—can be fairly compared. Carefully chosen weights and smoothing over time then turn this mix into a continuous score between low and high impact, meant to reflect how strongly an event resonates socially and culturally rather than claiming any strict cause‑and‑effect.

Building a smarter lens for event data

To predict this Social Impact Score, the researchers assemble a large dataset of about 70,000 event‑level records, covering competitions like the Olympics, FIFA World Cup, Asian Games, and regional tournaments. Each record includes economic background (such as national income), cultural indicators (like diversity measures and public participation in cultural programs), media coverage and social‑media volume, gender balance in participation and coverage, and sustainability metrics. They enhance these raw inputs with “composite” features that combine factors—for example, a joint cultural index, a sustainability‑plus‑media signal, and measures of how sentiment interacts with gender balance—so that the model can pick up subtle patterns that simple counts would miss.

A dual‑attention AI that learns context as well as numbers

Standard statistical models struggle with the messy, non‑linear way these ingredients interact. The authors therefore design a new deep‑learning architecture called the Sociocultural Dual‑Attention TabTransformer. In plain terms, this model treats each feature (such as diversity or media reach) as a token and learns how they relate to one another. One “attention” pathway looks at the numerical variables and how they influence each other—for instance, how public sentiment and media coverage move together. A second pathway injects context from categorical information like event type, host region, and sport. Together, these paths let the model grasp both fine‑grained feature interactions and broader cultural setting. Additional training tricks gently encourage reasonable behavior—for example, ensuring that higher sustainability or greater gender equality should not reduce the predicted social impact.

Checking reliability, fairness, and what really drives impact

The new model is tested against familiar tools including linear regression, nearest‑neighbor methods, and a strong gradient‑boosted tree model. Across a battery of error measures and rank‑based comparisons, the dual‑attention transformer consistently predicts Social Impact Scores more accurately and with tighter confidence intervals. Time‑based tests (training on earlier years and predicting later ones) and “leave‑one‑region‑out” experiments show that the approach remains robust when applied to new years or different parts of the world. To open the black box, the authors distill the transformer’s behavior into a simpler model and use SHAP, a popular explainability method, to see which features matter most. Consistently, cultural engagement, diversity, gender representation, public sentiment, and sustainability emerge as the strongest positive contributors, while purely economic indicators and raw viewership play a supporting, but less decisive, role.

What this means for the future of global games

Overall, the study suggests that the social value of mega‑events is shaped less by sheer size or wealth and more by how inclusive, uplifting, and responsible they are. Events that actively promote diverse participation, gender balance, environmentally friendly practices, and a positive public mood tend to achieve higher Social Impact Scores, especially in regions where such themes are already gaining traction. At the same time, the authors stress that their model captures patterns, not causes: it cannot prove that any single policy change will transform society. Still, by turning a tangle of cultural signals into a transparent, testable index, their framework offers a powerful starting point for organizers and policymakers who want to design sports events that genuinely strengthen communities rather than simply entertain them.

Citation: Chen, W., Syed Ali, S.K.B., Zulnaidi, H. et al. An explainable dual-attention transformer for predicting the sociocultural impact of global sports events. Sci Rep 16, 12812 (2026). https://doi.org/10.1038/s41598-026-43247-8

Keywords: sports analytics, social impact, artificial intelligence, public sentiment, cultural diversity