Gauri Kambhatla, Sanjana Gautam, Angela Zhang, Alexander Liu, Ravi Srinivasan, Junyi Jessy Li, Matthew Lease. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Juan Rodriguez, Todd Hay, David Gros, Zain Shamsi, Ravi Srinivasan. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

With the recent advances in human sensing, the push to integrate human mobility tracking with epidemic modeling highlights the lack of groundwork at the mesoscale (e.g., city-level) for both contact tracing and transmission dynamics. Although GPS data has been used to study city-level outbreaks in the past, existing approaches fail to capture the path of infection at the individual level. Consequently, in this paper, we extend epidemics prediction from estimating the size of an outbreak at the population level to estimating the individuals who may likely get infected within a finite period of time. To this end, we propose a network science based method to first build and then prune the dynamic contact networks for recurring interactions; these networks can serve as the backbone topology for mechanistic epidemics modeling. We test our method using Foursquare's Points of Interest (POI) smart phone geolocation data from over 1.3 million devices to better approximate the COVID-19 infection curves for two major (yet very different) US cities, (i.e., Austin and New York City), while maintaining the granularity of individual transmissions and reducing model uncertainty. Our method provides a foundation for building a disease prediction framework at the mesoscale that can help both policy makers and individuals better understand their estimated state of health and help the pandemic mitigation efforts.

Bayesian reward learning from demonstrations enables rigorous safety and uncertainty analysis when performing imitation learning. However, Bayesian reward learning methods are typically computationally intractable for complex control problems. We propose Bayesian Reward Extrapolation (Bayesian REX), a highly efficient Bayesian reward learning algorithm that scales to high-dimensional imitation learning problems by pre-training a low-dimensional feature encoding via self-supervised tasks and then leveraging preferences over demonstrations to perform fast Bayesian inference. Bayesian REX can learn to play Atari games from demonstrations, without access to the game score and can generate 100,000 samples from the posterior over reward functions in only 5 minutes on a personal laptop. Bayesian REX also results in imitation learning performance that is competitive with or better than state-of-the-art methods that only learn point estimates of the reward function. Finally, Bayesian REX enables efficient high-confidence policy evaluation without having access to samples of the reward function. These high-confidence performance bounds can be used to rank the performance and risk of a variety of evaluation policies and provide a way to detect reward hacking behaviors.

As infectious disease surveillance systems expand to include digital, crowd-sourced, and social network data, public health agencies are gaining unprecedented access to high-resolution data and have an opportunity to selectively monitor informative individuals. Contact networks, which are the webs of interaction through which diseases spread, determine whether and when individuals become infected, and thus who might serve as early and accurate surveillance sensors. Here, we evaluate three strategies for selecting sensors-sampling the most connected, random, and friends of random individuals-in three complex social networks-a simple scale-free network, an empirical Venezuelan college student network, and an empirical Montreal wireless hotspot usage network. Across five different surveillance goals-early and accurate detection of epidemic emergence and peak, and general situational awareness-we find that the optimal choice of sensors depends on the public health goal, the underlying network and the reproduction number of the disease (R0). For diseases with a low R0, the most connected individuals provide the earliest and most accurate information about both the onset and peak of an outbreak. However, identifying network hubs is often impractical, and they can be misleading if monitored for general situational awareness, if the underlying network has significant community structure, or if R0 is high or unknown. Taking a theoretical approach, we also derive the optimal surveillance system for early outbreak detection but find that real-world identification of such sensors would be nearly impossible. By contrast, the friends-of-random strategy offers a more practical and robust alternative. It can be readily implemented without prior knowledge of the network, and by identifying sensors with higher than average, but not the highest, epidemiological risk, it provides reasonably early and accurate information.

We present an interactive gesture recognition-based, immersive augmented reality system visualizing computational fluid dynamics (CFD) datasets of indoor environments. CFD simulation is used to predict the indoor environments and assess their response to specific internal and external conditions. To enable efficient visualization of CFD datasets in actual-space, an augmented reality system was integrated with a CFD simulation engine. To facilitate efficient data manipulation of the simulated post-processed CFD data and to increase the user-control of the immersive environment, a new intuitive method of human-computer interaction (HCI) has been incorporated. A gesture recognition system was integrated with the augmented reality-CFD structure to transform hand-postural data into a general description of hand-shape, through forward kinematics and computation of hand segment positions and their joint angles. This enabled real-time interactions between users and simulated CFD results in actual-space.

Demonstrates the potential of PCIS (Piece-based Construction Information System) as a visualization tool for technical cooperation, collaboration and communication by literally aligning the points of view of architectural, engineering and construction methodologies. There is always complexity during an actual construction or assembly, wherein the sequence is mostly always confused, and this leads to wastage of time, materials and manpower. The pictorial database PCIS, as a decision-aid tool, provides information in a sequence that is often forgotten or missed in construction. PCIS has three components. The first is an anatomical model, functionally organized according to ASTM UNIFORMAT II. This programmatic standard means the model can be sequentially constructed and deconstructed according to a standard work breakdown structure that is tailored to the requirements of almost any building. The second component of PCIS is the dataTheater that surrounds the anatomical model. The theater sets up a combination of 2D data planes that can be hypergraphically accessed to view layout dimensions, details and specifications. The third part of PCIS is the dataWorld, with cameras fixed to the intersections of its latitudes and longitudes. The cameras attached to the dataTheater and dataWorld control the visual display of the pictorial information. A software engine polls the cameras and publishes the images in a set of directories as files that are accessed by the hypergraphic interface. The information visible to these cameras can be changed through a standard survey matrix.