Interpretable Markov-Based Spatiotemporal Risk Surfaces for Missing-Child Search Planning with Reinforcement Learning and LLM-Based Quality Assurance
arXiv:2603.08933v1 Announce Type: new Abstract: The first 72 hours of a missing-child investigation are critical for successful recovery. However, law enforcement agencies often face fragmented, unstructured data and a lack of dynamic, geospatial predictive tools. Our system, Guardian, provides an end-to-end decision-support system for missing-child investigation and early search planning. It converts heterogeneous, unstructured case documents into a schema-aligned spatiotemporal representation, enriches cases with geocoding and transportation context, and provides probabilistic search products spanning 0-72 hours. In this paper, we present an overview of Guardian as well as a detailed description of a three-layer predictive component of the system. The first layer is a Markov chain, a sparse, interpretable model with transitions incorporating road accessibility costs, seclusion preferences, and corridor bias with separate day/night parameterizations. The Markov chain's output predict
arXiv:2603.08933v1 Announce Type: new Abstract: The first 72 hours of a missing-child investigation are critical for successful recovery. However, law enforcement agencies often face fragmented, unstructured data and a lack of dynamic, geospatial predictive tools. Our system, Guardian, provides an end-to-end decision-support system for missing-child investigation and early search planning. It converts heterogeneous, unstructured case documents into a schema-aligned spatiotemporal representation, enriches cases with geocoding and transportation context, and provides probabilistic search products spanning 0-72 hours. In this paper, we present an overview of Guardian as well as a detailed description of a three-layer predictive component of the system. The first layer is a Markov chain, a sparse, interpretable model with transitions incorporating road accessibility costs, seclusion preferences, and corridor bias with separate day/night parameterizations. The Markov chain's output prediction distributions are then transformed into operationally useful search plans by the second layer's reinforcement learning. Finally, the third layer's LLM performs post hoc validation of layer 2 search plans prior to their release. Using a synthetic but realistic case study, we report quantitative outputs across 24/48/72-hour horizons and analyze sensitivity, failure modes, and tradeoffs. Results show that the proposed predictive system with the three-layer architecture produces interpretable priors for zone optimization and human review.
Executive Summary
This study presents Guardian, an end-to-end decision-support system for missing-child investigations and early search planning. The system leverages Markov-based spatiotemporal risk surfaces, reinforcement learning, and large language model (LLM) quality assurance to provide probabilistic search products spanning 0-72 hours. A three-layer predictive component is detailed, including a Markov chain incorporating various factors, reinforcement learning for search plan optimization, and LLM validation of search plans. The study reports quantitative results from a synthetic case study, demonstrating the system's ability to produce interpretable priors for zone optimization and human review. The proposed system has the potential to enhance missing-child search planning, but its effectiveness in real-world scenarios remains to be seen.
Key Points
- ▸ Guardian is an end-to-end decision-support system for missing-child investigations and early search planning.
- ▸ The system uses Markov-based spatiotemporal risk surfaces, reinforcement learning, and LLM quality assurance.
- ▸ A three-layer predictive component is proposed, incorporating a Markov chain, reinforcement learning, and LLM validation.
Merits
Strength in Spatiotemporal Modeling
The proposed system effectively captures the complexities of spatiotemporal relationships in missing-child investigations, enabling more accurate search planning.
Interpretable Results
The use of a Markov chain and reinforcement learning ensures that the system's predictions are interpretable, facilitating human review and decision-making.
Demerits
Limited Real-World Evaluation
The study relies on a synthetic case study, limiting the system's evaluation in real-world scenarios and potential biases in the data.
Dependence on LLM Quality
The system's reliance on LLM quality assurance may introduce limitations, particularly if the LLM is not adequately trained or validated.
Expert Commentary
While the Guardian system presents a promising approach to missing-child search planning, its limitations, particularly the reliance on synthetic data and LLM quality assurance, necessitate further research and evaluation. A more comprehensive evaluation of the system's performance in real-world scenarios, incorporating diverse data sources and human review, would be essential. Additionally, the study's focus on technical aspects of the system overlooks the social and organizational contexts in which it will be deployed, highlighting the need for future research to address these critical factors.
Recommendations
- ✓ Future studies should focus on evaluating the Guardian system in real-world scenarios, incorporating diverse data sources and human review.
- ✓ Researchers should investigate the social and organizational contexts in which the system will be deployed, addressing potential implementation challenges and biases.
