DECIDE-SIM: Simulation-Based Decision Analysis

• DECIDE-SIM is a framework that formalizes simulation scenario selection by partitioning the configuration space into Simulation Configuration Classes (SCCs) for comprehensive model validation.
• It automates validation workflows using constraint solvers and state sequencing to ensure rigorous, repeatable simulation analysis in complex systems.
• It incorporates entropy-based model fitting and modular, multiscale simulation architectures to support decision analysis in domains such as healthcare and engineering.

DECIDE-SIM refers to principled methodologies, frameworks, and tools for guiding, automating, and validating simulation-based decision-making in complex systems. Across diverse domains—healthcare, epidemiology, engineering, quantum foundations, and computational modeling—DECIDE-SIM emphasizes rigorous scenario selection, dynamic evidence synthesis, agent-based simulation design, and outputs suitable for empirical policy analysis or critical theoretical inference.

1. Formal Criteria for Simulation Scenario Selection

A foundational challenge in simulation-driven decision-making is the selection of representative scenarios from a vast, often infinite configuration space. In the context of Discrete Event System Specification (DEVS) model validation, DECIDE-SIM as described in (Hollmann et al., 2014) formalizes this process using a family of simulation criteria that partition the input space into finite Simulation Configuration Classes (SCCs). The criteria include:

• Case-based transition functions: A scenario for each conditional branch within internal/external transitions.
• Extensional and intentional sets: Explicit enumeration or DNF-based partitioning for states or events defined by logical predicates.
• Operator and domain partitions: Sub-domains for algebraic operators, propagated through composite expressions.
• Temporal intervals: Key points and intervals with respect to simulation clocks, yielding distinct SCCs per logical timing aspect.

This partitioning allows for systematic, disciplined coverage of all relevant behavioral classes and supports semi-automated scenario generation using constraint solvers (e.g., SMT or set-based logic systems). One simulation per SCC is typically sufficient under the uniformity hypothesis, greatly increasing model validation confidence compared to ad hoc case selection.

2. Automated Validation Methodologies

DECIDE-SIM frameworks enable partial or full automation of the simulation validation process. As implemented in (Hollmann et al., 2014), key steps include:

1. Parsing the mathematical specification of the simulation model.
2. Applying formal partition criteria to generate SCCs.
3. Using constraint solvers to select scenario representatives per SCC.
4. Sequencing scenarios via state intersections to optimize simulation reuse.
5. Abstracting and comparing simulation outputs against formal requirements.

Automation reduces manual workload and ensures rigorous, repeatable coverage. Extension to coupled models and the development of standardized formal grammars are identified as future directions, with potential for tool integration spanning parsing, partitioning, and simulation code generation.

3. Data Synthesis and Entropy-Based Model Fitting

Where simulation models must operate in domains with incomplete or heterogeneous real-world data—particularly in healthcare epidemiology—DECIDE-SIM methodologies incorporate novel synthesis strategies. In the chronic disease decision support model (Green et al., 2010), the absence of comprehensive longitudinal data motivates an entropy-based fitting approach:

$$M^* = \arg \min_{M \in \mathcal{M}} \sum_{j=1}^N w_j \cdot d\left( \tilde{\varphi}^j(t_j, \Delta_j), \hat{\varphi}^j(t_j, \Delta_j; M) \right)$$

Here, $d(\cdot, \cdot)$ denotes Jensen–Shannon distance, and constraints can be derived from heterogeneous sources. Simulated annealing is used for optimization, allowing the fitted survival densities to flexibly integrate expert judgement and diverse dataset features. This approach yields competing hazards models suitable for policy evaluation under scenarios of limited, fragmentary evidence.

4. Multiscale and Modular Simulation Architectures

DECIDE-SIM emphasizes scalability and modularity across simulation system design. The SICO infection control framework (Pine et al., 2023) illustrates this principle via:

• Agent-based modeling capable of simulating population sizes from local to national scale.
• Modular separation of disease transmission, testing regimes (including pooled testing), isolation dynamics, and vaccination modeling.
• Parameterization of intervention strategies through flexible configuration files, enabling policy-makers to explore trade-offs (e.g., between test frequency, cost, and outcome metrics).

Similarly, the integration of modular submodels for disease, cost, and utility in health economic simulation (hesim (Incerti et al., 2021)) leverages object-oriented structures (R6 and S3 classes) for rapid assembly and extension.

5. Simulation-Driven Decision Analysis and Policy Support

The operational outputs of DECIDE-SIM frameworks inform decision-making by producing interpretable metrics for policy evaluation. In health economics, this is typified by simulation-derived estimates of costs, QALYs, and cost-effectiveness, subject to uncertainty propagation via probabilistic sensitivity analysis (PSA). Decision rules are formalized:

• Net Monetary Benefit (NMB): $NMB(j, \theta) = e_j(\theta) \cdot k - c_j(\theta)$ for treatment $j$ and parameter set $\theta$.
• Incremental cost-effectiveness ratios (ICER): for pairwise treatment comparison.

Simulation agents (Kleiman et al., 19 May 2025) extend this paradigm by connecting simulation backends with LLM interfaces, enabling non-technical users to pose natural language queries, automatically modify simulation parameters, and interpret complex outputs for strategic or operational decision support.

6. Application Domains and Empirical Impact

DECIDE-SIM methodologies have demonstrated significant empirical impact in multiple domains:

• Healthcare: Quantified risk reduction of interventions like statin and beta blocker uptake, with direct support for policy decisions in coronary heart disease (Green et al., 2010), rigorous cost-effectiveness analysis via hesim (Incerti et al., 2021).
• Infectious Disease Control: Evaluation of pooled testing strategies, optimal isolation policies, and vaccination scenarios in COVID-19 response (Pine et al., 2023).
• Engineering: Interactive simulation with Lattice Boltzmann Methods (Wang et al., 2021) for real-time exploration and decision-making.
• Distributed Systems: Large-scale distributed simulation architectures to hide computational effort and optimize resource usage (Ciprian et al., 2011).
• Quantum Foundations: Simulation-based experimental designs to determine which foundational assumption—Locality, Realism, or Ergodicity—is violated in Bell test scenarios by analyzing the temporal evolution of randomness rejection rates (Hnilo, 2020).
• Supply Chain Optimization: Closed feedback integration of generative simulators and dual-aware decision models, enhancing timely delivery and profit under realistic operational constraints (Bai et al., 10 Jul 2025).

7. Future Directions and Open Challenges

Ongoing and future research in DECIDE-SIM includes:

• Standardization of input model grammars to enable full toolchain automation.
• Extension to coupled models and other simulation paradigms.
• Integration with advanced AI agents for richer output interpretation and adaptive scenario selection.
• Development of novel synthesis approaches for evidence aggregation under sparsity and heterogeneity.
• Empirical benchmarking and validation across increasingly complex and uncertain domains.

Efforts to advance the theoretical and practical foundations of DECIDE-SIM are poised to further close the gap between high-fidelity simulation and real-world decision efficacy, supporting robust, accountable, and efficient policy choices in diverse fields.