Abstract-Upon concluding a meeting, participants can occasionally leave with different understandings of what had been discussed. Detecting inconsistencies in understanding is a desired capability for an intelligent system designed to monitor meetings and provide feedback to spur stronger shared understanding.In this paper, we present a computational model for the automatic prediction of consistency among team members' understanding of their group's decisions. The model utilizes dialogue features focused on the dynamics of group decision-making. We trained a hidden Markov model using the AMI meeting corpus and achieved a prediction accuracy of 64.2%, as well as robustness across different meeting phases. We then implemented our model in an intelligent system that participated in human team planning about a hypothetical emergency response mission. The system suggested topics that the team would derive the most benefit from reviewing with one another. Through an experiment with 30 participants, we evaluated the utility of such a feedback system, and observed a statistically significant increase of 17.5% in objective measures of the teams' understanding compared with that obtained using a baseline interactive system. c 2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.Index Terms-Consistency of understanding, intelligent agent participation, adaptive review, human-computer interaction, dialogue acts, hidden Markov models.

Electronic health records (EHRs) represent a rich data source to support precision medicine, particularly in disorders with small and heterogeneous populations where longitudinal phenotypes are poorly characterized. However, the impact of EHR data is often limited by incomplete or imperfect source documentation and the inability to leverage unstructured data. Here, we address these shortcomings through a computational analysis of one of the largest cohorts of developmental and epileptic encephalopathies (DEEs), representing 466 individuals across six genetically defined conditions. The DEEs encompass debilitating pediatric-onset disorders with high unmet needs for which treatment development is ongoing. By applying a platform approach to data curation and annotation of 18 clinical data entities from comprehensive medical records, we characterize variation in longitudinal clinical journeys. Assessments of the relative enrichment of phenotypes and semantic similarity analysis highlight commonalities and differences between the six cohorts. Evaluation of medication use reflects unmet needs, particularly in the management of movement disorders. We also present a novel composite measure of seizure severity that is more robust than existing measures of seizure frequency alone. Finally, we show that the attainment of developmental outcomes, including the ability to sit independently and the ability to walk, is correlated with seizure severity scores. Overall, the combined analyses demonstrate that patient-centric real world data generation, including structuring of medical records, holds promise to improve clinical trial success in rare disorders. Applications of this approach support improved understanding of baseline disease progression, selection of relevant endpoints, and definition of inclusion and exclusion criteria.