TLTaesik Lee

Taesik Lee

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Taesik Lee has a total of 34 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Decision-Oriented Representations of Response Heterogeneity with Modular Finite Mixture Regression

    by Jungwoo Kim, Taesik Lee

    When decision makers choose interventions for different people, it is not enough to know that responses differ; they must know where those differences arise and whether they align with available actions. In addition, such representations must not impose excessive interpretive complexity. We propose Modular Finite Mixture Regression, a configurable framework that organizes attributes into modules and represents modulelevel response types. By making the unit and resolution of heterogeneity configurable, mFMR reframes mixture regression as a problem of decision-relevant representation design. We also develop a configuration discovery procedure that returns a fit-complexity frontier of parsimonious, interpretable, and empirically supported representations.

  • Open Access

    Modeling Experience Effects from Related but Non-Identical ChoiceContexts

    by Jihwan Bae, Jungwoo Kim, Taesik Lee

    A large strand of the choice modeling literature has examined how priorexperience affects subsequent choices in repeated-choice settings, where ex-perience is often represented as persistence toward previously chosen alter-natives. However, many real-world choices are cross-contextual. Decisionmakers face a choice context that is new in its exact form while drawing onprior experiences from related, but non-identical, contexts. In such settings,prior experience may not only create alternative-specific persistence, but alsoinform how observable attributes are interpreted and weighted in the currentchoice context.To address this gap, this paper develops a constructed-state approachfor incorporating fragmented cross-contextual histories into current choicemodels. The approach converts prior encounters into two decision-relevantstates: a patient-level general experience state, which allows prior experi-ence to moderate attribute valuation, and an alternative-specific momentumstate, which captures persistence toward previously experienced alternatives.We apply the framework to hospital choice among South Korean patients un-dergoing first-time total knee arthroplasty using National Health Insuranceclaims data. The constructed-history specification better explains patients’observed choices, demonstrating that constructed experience states providea systematic way to extend choice models to settings in which prior historiesare fragmented across related but non-identical contexts.

  • Open Access

    A Conceptual Framework for Spatial Accessibility: Aligning Metrics with Urban Service Determinants

    by Joongchan Ahn, Huijae Kim, Taesik Lee

    Spatial accessibility is a core indicator in urban planning and policy, yet its practical use is often complicated by the coexistence of multiple definitions and the absence of clear principles for matching services and metrics. This study proposes a conceptual framework that supports fitness-for-purpose metric selection by aligning accessibility measures with the operational objectives of urban services. We conceptualize spatial accessibility through three components: (i) travel impedance (reachability), (ii) reachable opportunities (opportunity volume within reach), and (iii) competition arising when multiple users share limited service capacity. Using the notion of a dominant accessibility determinant, we summarize urban services into three archetypes: Time-critical (impedance-dominant), Opportunity-driven (opportunity-dominant), and Capacity-constrained (competition-dominant). In parallel, we organize representative accessibility measures into three corresponding families: Proximity-based, Opportunity-based (threshold and distance-decay formulations), and Competition-adjusted (FCA- type) measures. A shared three-component simplex provides an interpretable visual basis for service–metric alignment. The framework clarifies what different measures encode as “accessibility” and offers a transparent guideline for selecting and interpreting indicators in urban policy diagnostics, reducing the risk of ambiguous or misleading conclusions when measures are applied outside their intended service contexts.

  • Open Access

    Component-wise Latent Class Choice Model

    by Jungwoo Kim, Taesik Lee

    Understanding individual preference heterogeneity is a central challenge in discrete choice modeling. While the Latent Class Choice Model (LCCM) offers interpretable segmentation by grouping individuals with similar tastes, it imposes a restrictive structure in which all utility coefficients vary jointly within a single latent class. When heterogeneity arises independently across multiple attribute dimensions, this structure requires a combinatorial number of classes, leading to parameter redundancy, information dispersion, and statistical inefficiency.To address these limitations, we propose the Component-wise Latent Class Choice Model (CLCCM). The CLCCM decomposes the utility function into analyst-defined components, each governed by its own latent class structure. Rather than assigning individuals to a single monolithic class, the model allows distinct class memberships for each utility component, with overall preference profiles formed as the Cartesian product of component-specific memberships. This structure parsimoniously captures multidimensional heterogeneity while preserving interpretability. We further develop a generalized EM algorithm with component-wise updates that guarantees monotonic likelihood improvement and stable estimation.Monte Carlo simulations demonstrate that the CLCCM achieves superior statistical efficiency, more accurate recovery of utility coefficients and memberships, and more stable convergence than conventional LCCM specifications spanning the same heterogeneity space. An empirical application to discrete choice experiment data on breast cancer screening shows that the CLCCM uncovers interpretable behavioral segments associated with specific parts of the utility function that are conflated in standard models. Overall, the CLCCM provides a flexible and scalable framework that balances interpretability and efficiency for modeling multidimensional preference heterogeneity in applied choice analysis.

  • Open Access

    Hypernetwork-based approach for optimal composition design in partially controlled multi-agent systems

    by Kyeonghyeon Park, David Molina Concha, Hyun-Rok Lee, Chi-Guhn Lee, Taesik Lee

    Partially Controlled Multi-Agent Systems (PCMAS) are comprised of controllable agents, managed by a system designer, and uncontrollable agents, operating autonomously. This study addresses an optimal composition design problem in PCMAS, which involves the system designer's problem, determining the optimal number and policies of controllable agents, and the uncontrollable agents' problem, identifying their best-response policies. Solving this bi-level optimization problem is computationally intensive, as it requires repeatedly solving multi-agent reinforcement learning problems under various compositions for both types of agents. To address these challenges, we propose a novel hypernetwork-based framework that jointly optimizes the system's composition and agent policies. Unlike traditional methods that train separate policy networks for each composition, the proposed framework generates policies for both controllable and uncontrollable agents through a unified hypernetwork. This approach enables efficient information sharing across similar configurations, thereby reducing computational overhead. Additional improvements are achieved by incorporating reward parameter optimization and mean action networks. Using real-world New York City taxi data, we demonstrate that our framework outperforms existing methods in approximating equilibrium policies. Our experimental results show significant improvements in key performance metrics, such as order response rate and served demand, highlighting the practical utility of controlling agents and their potential to enhance decision-making in PCMAS.

  • Open Access

    Effectiveness of Social Distancing under Partial Compliance of Individuals

    by Hyelim Shin, Taesik Lee

    Social distancing reduces infectious disease transmission by limiting contact frequency and proximity within a community. However, compliance varies due to its impact on daily life. This paper explores the effects of compliance on social distancing effectiveness through a "social distancing game," where community members make decisions based on personal utility. We conducted numerical experiments to evaluate how different policy settings for social distancing affect disease transmission. Our findings suggest several key points for developing effective social distancing policies. Firstly, while generally effective, overly strict policies may lead to noncompliance and reduced effectiveness. Secondly, the public health benefits of social distancing need to be balanced against social costs, emphasizing policy efficiency. Lastly, for diseases with low reinfection risk, a segmented policy exempting immune individuals could lessen both infections and socioeconomic costs.