Soyun Jeon | Searcharxiv

arXiv2026

Analyzing Human Heuristics and Strategies in Everyday Decision-Making Conversations for Conversational AI Design

Conversational AI increasingly supports everyday decision-making, yet most systems rely on data-centric reasoning rather than the heuristic and interactional strategies people use in natural conversation. To ground design in actual human practice, we analyze 955 real-world Korean conversations (15,476 utterances) involving food and travel decisions, applying a decision-making codebook through an LLM-assisted coding pipeline. Our findings reveal that people prioritize satisficing over optimization, relying heavily on internal knowledge and interactional strategies to manage cognitive load. Critically, we identify a frequency-efficiency mismatch: the most prevalent heuristics sustain conversational flow during exploration, whereas infrequent, rule-based strategies are highly effective at driving resolution during exploitation. By mapping how these patterns transfer across the spectrum of human-AI interaction, this work provides empirical grounding consistent with cognitive theories of decision-making and offers design implications that align AI systems with human heuristic processes.

cs.HC↗

arXiv2025

Copula-based spatio-temporal modeling of air pollutant data incorporating covariate dependencies

Elevated levels of PM10 are known to cause severe respiratory and cardiovascular diseases, and, in extreme cases, cancer and mortality. Despite various reduction policies implemented across different sectors, PM10 concentrations in South Korea continue to exceed the annual recommended limits set by the World Health Organization. Spatio-temporal PM10 concentrations may exhibit both spatial and temporal dependencies. Additionally, interactions between PM10 and environmental factors can further influence the variability in PM10. Therefore, this study proposes a method that incorporates the spatio-temporal neighbors of covariates alongside those of PM10 by adopting an approach that explains spatio-temporal interactions through spatio-temporal neighbors. Vine copulas are used to integrate the pairwise dependence structures between a given location and its surrounding spatio-temporal neighbors. We applied the model to weekly average PM10 data from South Korea in 2019, using PM2.5 and CO as covariates. Given that all three variables exhibited skewness, we assumed the Gumbel and Generalized Extreme Value distributions as marginal distributions. The proposed model outperformed a traditional Bayesian spatio-temporal model, a kriging method, and an alternative copula-based approach, particularly in predicting the top 5% of extreme values, by effectively capturing tail dependencies crucial for extreme value analysis. This study highlights the importance of utilizing vine copulas to effectively model diverse dependency structures in spatio-temporal data while simultaneously accommodating spatial and temporal dimensions, including spatio-temporal dependencies among covariates. The results underscore the broader applicability of the proposed approach to other fields where complex dependency structures are present.

stat.AP↗