Selecting a bundle of items that collectively satisfies constraints is a fundamental task across databases, recommender systems, and text summarization. Unlike traditional retrieval that returns individual or top-k items, bundle retrieval is inherently combinatorial and, in general, NP-hard. Although package queries can efficiently retrieve bundles given a well-formed query, two key user-centric challenges remain: (1) expressing and tuning multi-dimensional bundle intent through a user-friendly interface, and (2) ensuring feasibility when the query yields empty results. We introduce Ex2Bundle, an Example-driven Bundle retrieval framework that enables users to specify their intent through example bundles and automatically synthesizes package queries that capture the intent implicit in those example bundles via aggregate constraints. Ex2Bundle also addresses a challenge unique to bundle retrieval: when inferred aggregate constraints are infeasible over the target data, our data-aware constraint relaxation minimally adjusts the constraint bounds while preserving alignment with user intent. We instantiate a specific application of focused text snippet extraction by example to demonstrate the efficacy of the Ex2Bundle framework. Extensive experiments over real-world datasets and a user study demonstrate that Ex2Bundle improves usability and consistently returns intent-aligned bundles even under distributional shifts of the target database.

Metamodels for discrete-event simulations approximate the behavior of simulation models without running expensive simulations. Prior work introduced modular dynamic Bayesian networks (MDBNs) – a class of metamodels that can estimate a range of probabilistic and causal queries (PCQs) using a single, trained model – but the method was limited to Markovian systems. In this paper, we initiate an extension of MDBNs to non-Markovian queues by approximating non-exponential distributions using phase-type distributions. This approach raises novel challenges, including balancing metamodeling accuracy and tractability when choosing the number of phases, efficiently learning metamodel parameters, and choosing the sampling interval that is used to approximate a continuous-time simulation by a discrete-time MDBN. We provide preliminary solutions to these challenges, yielding the first causal metamodeling technique for non-Markovian systems. Experiments on a G/M/1 queue demonstrate that the MDBN can produce accurate answers to PCQs with orders-of-magnitude speedup of inference times relative to direct simulation.

What-if analysis is widely used to explore hypothetical scenarios and evaluate alternative pathways to desired results. However, current approaches are fragmented: systems implement what-if capabilities under diverse terminologies with different analytic techniques. Such fragmentation limits expressiveness, impedes flexible composition and reuse of workflows, and hinders tighter integration with AI. We present PRAXA, a compositional grammar of what-if analysis derived from recurring patterns across 141 publications in visual analytics and HCI venues. PRAXA formulates three primitives: (1) data, defining variables under analysis, (2) model, specifying predictive mechanisms, and (3) interaction operations-pairs of user actions and system responses that execute analyses. We encode PRAXA into a declarative specification language, PSL. To evaluate PRAXA, we first show expressiveness by reconstructing representative workflows from prior work as structured compositions, exposing the predominant focus on single-step rather than multi-step reasoning. Second, we demonstrate composability by revealing that capabilities described under distinct terminologies share the same grammatical structure with different parameterizations, and that new multi-step workflows emerge through composition. Third, we illustrate PSL as an intermediate representation for translating natural-language what-if queries into executable interactive interfaces, enabling inspection, validation, and more transparent AI integration. By unifying diverse what-if approaches as a grammar, PRAXA provides a foundation for analyzing, composing, and supporting workflows in next-generation what-if systems.

Decision making under uncertainty often requires choosing packages, or bags of tuples, that collectively optimize expected outcomes while limiting risks. Processing Stochastic Package Queries (SPQs) involves solving very large optimization problems on uncertain data. Monte Carlo methods create numerous scenarios, or sample realizations of the stochastic attributes of all the tuples, and generate packages with optimal objective values across these scenarios. The number of scenarios needed for accurate approximation-and hence the size of the optimization problem when using prior methods- increases with variance in the data, and the search space of the optimization problem increases exponentially with the number of tuples in the relation. Existing solvers take hours to process SPQs on large relations containing stochastic attributes with high variance. Besides enriching the SPaQL language to capture a broader class of risk specifications, we make two fundamental contributions toward scalable SPQ processing. First, we propose risk-constraint linearization (RCL), which converts SPQs into Integer Linear Programs (ILPs) whose size is independent of the number of scenarios used. Solving these ILPs gives us feasible and near-optimal packages. Second, we propose STOCHASTIC SKETCHREFINE, a divide and conquer framework that breaks down a large stochastic optimization problem into subproblems involving smaller subsets of tuples. Our experiments show that, together, RCL and STOCHASTIC SKETCHREFINE produce high-quality packages in orders of magnitude lower runtime than the state of the art.

Despite the widespread use of computer systems, their societal impacts are often poorly understood, highlighting the need for the AIED community to acknowledge and contribute to incorporating Ethical and Responsible Computing (ERC) into Computer Science (CS) education. We introduce the Platform for Ethical and Responsible Computing Education (PEaRCE), an interactive tool created to integrate ERC into post-secondary education through realistic workplace simulations. PEaRCE scenarios guide students through ERC quandaries-awareness, decision-making, feedback-during the processes of developing advanced AIED and other technologies that may have beneficial and harmful impacts. Moreover, we integrate PEaRCE into CS courses via a sequence of structured learning modules and trained "ERC Teaching Assistants (TAs)" to support the integration process. We present insights from our deployment experiences, show PEaRCE's potential to enhance ERC awareness and reasoning, and discuss the possibility of embedding PEaRCE into AI in Education courses.

What-if analysis (WIA) is essential for data-driven decision-making, allowing users to assess how changes in variables impact outcomes and explore alternative scenarios. Existing WIA research primarily supports the workflows of data scientists and analysts, and largely overlooks business professionals who engage in WIA through non-technical means. To bridge this gap, we conduct a two-part user study with 22 business professionals across marketing, sales, product, and operations roles. The first study examines their existing WIA practices, tools, and challenges. Findings reveal that business professionals perform many WIA techniques independently using rudimentary tools due to various constraints. We then implement representative WIA techniques in a visual analytics prototype and use it as a probe to conduct a follow-up study evaluating business professionals’ practical use of the techniques. Results show that these techniques improve decision-making efficiency and confidence while underscoring the need for better support in data preparation, risk assessment, and domain knowledge integration. Finally, we offer design recommendations to enhance future business analytics systems.

This advanced tutorial explores some recent applications of artificial neural networks (ANNs) to stochastic discrete-event simulation (DES). We first review some basic concepts and then give examples of how ANNs are being used in the context of DES to facilitate simulation input modeling, random variate generation, simulation metamodeling, optimization via simulation, and more. Combining ANNs and DES allows exploitation of the deep domain knowledge embodied in simulation models while simultaneously leveraging the ability of modern ML techniques to capture complex patterns and relationships in data.

As social and economic conditions are key determinants of HIV, the United States 'National HIV/AIDS Strategy (NHAS)', in addition to care and treatment, aims to address mental health, unemployment, food insecurity, and housing instability, as part of its strategic plan for the 'Ending the HIV Epidemic' initiative. Although mechanistic models of HIV play a key role in evaluating intervention strategies, social conditions are typically not part of the modeling framework. Challenges include the unavailability of coherent statistical data for social conditions and behaviors. We developed a method, combining undirected graphical modeling with copula methods, to integrate disparate data sources, to estimate joint probability distributions for social conditions and behaviors. We incorporated these in a national-level network model, Progression and Transmission of HIV (PATH 4.0), to simulate behaviors as functions of social conditions and HIV transmissions as a function of behaviors. As a demonstration for the potential applications of such a model, we conducted two hypothetical what-if intervention analyses to estimate the impact of an ideal 100% efficacious intervention strategy. The first analysis modeled care behavior (using viral suppression as proxy) as a function of depression, neighborhood, housing, poverty, education, insurance, and employment status. The second modeled sexual behaviors (number of partners and condom-use) as functions of employment, housing, poverty, and education status, among persons who exchange sex. HIV transmissions and disease progression were then simulated as functions of behaviors to estimate incidence reductions. Social determinants are key drivers of many infectious and non-infectious diseases. Our work enables the development of decision support tools to holistically evaluate the syndemics of health and social inequity.

A package query returns a package - a multiset of tuples - that maximizes or minimizes a linear objective function subject to linear constraints, thereby enabling in-database decision support. Prior work has established the equivalence of package queries to Integer Linear Programs (ILPs) and developed the SketchRefine algorithm for package query processing. While this algorithm was an important first step toward supporting prescriptive analytics scalably inside a relational database, it struggles when the data size grows beyond a few hundred million tuples or when the constraints become very tight. In this paper, we present Progressive Shading, a novel algorithm for processing package queries that can scale efficiently to billions of tuples and gracefully handle tight constraints. Progressive Shading solves a sequence of optimization problems over a hierarchy of relations, each resulting from an ever-finer partitioning of the original tuples into homogeneous groups until the original relation is obtained. This strategy avoids the premature discarding of high-quality tuples that can occur with SketchRefine. Our novel partitioning scheme, Dynamic Low Variance, can handle very large relations with multiple attributes and can dynamically adapt to both concentrated and spread-out sets of attribute values, provably outperforming traditional partitioning schemes such as KD-tree. We further optimize our system by replacing our off-the-shelf optimization software with customized ILP and LP solvers, called Dual Reducer and Parallel Dual Simplex respectively, that are highly accurate and orders of magnitude faster.

Fitting stochastic input-process models to data and then sampling from them are key steps in a simulation study but highly challenging to non-experts. We present Neural Input Modeling (NIM), a Generative Neural Network (GNN) framework that exploits modern data-rich environments to automatically capture simulation input processes and then generate samples from them. The basic GNN that we develop, called NIM-VL, comprises (i) a variational autoencoder architecture that learns the probability distribution of the input data while avoiding overfitting and (ii) long short-term memory components that concisely capture statistical dependencies across time. We show how the basic GNN architecture can be modified to exploit known distributional properties-such as independent and identically distributed structure, nonnegativity, and multimodality-to increase accuracy and speed, as well as to handle multivariate processes, categorical-valued processes, and extrapolation beyond the training data for certain nonstationary processes. We also introduce an extension to NIM called Conditional Neural Input Modeling (CNIM), which can learn from training data obtained under various realizations of a (possibly time series valued) stochastic "condition," such as temperature or inflation rate, and then generate sample paths given a value of the condition not seen in the training data. This enables users to simulate a system under a specific working condition by customizing a pre-trained model; CNIM also facilitates what-if analysis. Extensive experiments show the efficacy of our approach. NIM can thus help overcome one of the key barriers to simulation for non-experts.

This experience report details pilot deployments of the Platform for Ethics and Responsible Computing Education (PEaRCE) in undergraduate computer science (CS) courses. PEaRCE is an online learning environment designed to immerse undergraduate CS students in realistic interactive work scenarios. The simulated work scenarios illustrate how technical CS course material can be readily exploited towards the creation of appealing systems, which nevertheless can be instruments of harm to others. Students take the role of an employee within a project and are given details about the project. Students then decide whether to continue with the project immediately, or to first talk to various project stakeholders who could provide additional information about the project and its impacts. At the end of the simulation, students are provided information about the project scenario, including the potential impacts of the project and feedback on stakeholder conversation choices. We report on preliminary findings from pilot deployments of PEaRCE in undergraduate CS courses, including student feedback about PEaRCE and the simulation scenarios and discuss future directions for the design of the system and its use in future CS courses.

Simulation metamodeling is essential for speeding up optimization via simulation to support rapid decision making. During optimization, the metamodel, rather than expensive simulation, is used to compute objective values. We recently developed graphical neural metamodels (GMMs) that use graph neural networks to allow the graphical structure of a simulation model to be treated as a metamodel input parameter that can be varied along with scalar inputs. In this paper we provide novel methods for using GMMs to solve hybrid optimization problems where both real-valued input parameters and graphical structure are jointly optimized. The key ideas are to modify Monte Carlo tree search to incorporate both discrete and continuous optimization and to leverage the automatic differentiation infrastructure used for neural network training to quickly compute gradients of the objective function during stochastic gradient descent. Experiments on stochastic activity network and warehouse models demonstrate the potential of our method.

A traditional metamodel for a discrete-event simulation approximates a real-valued performance measure as a function of the input-parameter values. We introduce a novel class of metamodels based on modular dynamic Bayesian networks (MDBNs), a subclass of probabilistic graphical models which can be used to efficiently answer a rich class of probabilistic and causal queries (PCQs). Such queries represent the joint probability distribution of the system state at multiple time points, given observations of, and interventions on, other state variables and input parameters. This paper is a first demonstration of how the extensive theory and technology of causal graphical models can be used to enhance simulation metamodeling. We demonstrate this potential by showing how a single MDBN for an M/M/1 queue can be learned from simulation data and then be used to quickly and accurately answer a variety of PCQs, most of which are out-of-scope for existing metamodels.

Probability proportional to size (PPS) sampling schemes with a target sample size aim to produce a sample comprising a specified number n of items while ensuring that each item in the population appears in the sample with a probability proportional to its specified “weight” (also called its “size”). These two objectives, however, cannot always be achieved simultaneously. Existing PPS schemes prioritize control of the sample size, violating the PPS property if necessary. We provide a new PPS scheme that allows a different trade-off: our method enforces the PPS property at all times while ensuring that the sample size never exceeds the target value n. The sample size is exactly equal to n if possible, and otherwise has maximal expected value and minimal variance. Thus we bound the sample size, thereby avoiding storage overflows and helping to control the time required for analytics over the sample, while allowing the user complete control over the sample contents. The method is both simple to implement and efficient, being a one-pass streaming algorithm with an amortized processing time of O(1) per item.

Decision makers in a broad range of domains, such as finance, transportation, manufacturing, and healthcare, often need to derive optimal decisions given a set of constraints and objectives . Traditional solutions to such constrained optimization problems are typically application-specific, complex, and do not generalize. Further, the usual workflow requires slow, cumbersome, and error-prone data movement between a database and predictive-modeling and optimization packages. All of these problems are exacerbated by the unprecedented size of modern data-intensive optimization problems. The emerging research area of in-database prescriptive analytics aims to provide seamless domain-independent, declarative, and scalable approaches powered by the system where the data typically resides: the database. Integrating optimization with database technology opens up prescriptive analytics to a much broader community, amplifying its benefits. In the context of our prior and ongoing work in this area, we discuss some strategies for addressing key challenges related to usability, scalability, data uncertainty, dynamic environments with changing data and models, and the need to support decision-making agents. We indicate how deep integration between the DBMS, predictive models, and optimization software creates opportunities for rich prescriptive-query functionality with good scalability and performance.

Business users perform data analysis to inform decisions for improving business processes and outcomes despite having limited formal technical training. While earlier work has focused on data analysts' and data scientists' practices and challenges, little is known about business users' decision-making practices and how they incorporate data and visual analytics into their workflows. To address this gap, we first conduct an interview study with 22 business users to understand the general practices and challenges in their data-driven decision-making processes. We contribute an end-to-end model of business users' data-driven decision-making processes elaborating the tasks, tools, and challenges at each stage. We find that business users analyze data without relying on data analysts due to various practical constraints and considerations. However, their existing tools are inadequate, particularly in helping understand the relationship between data variables and business goals and facilitating the exploration of what-if scenarios. These findings suggest a need for advanced predictive and prescriptive analytics (PPA) tools to support what-if analysis. Motivated by this need, we perform a follow-up, task-based study to understand PPA's role and potential in business users' decision-making processes. We find that PPA helps improve efficiency and confidence in decision-making. However, business users also believe that PPA-powered what-if analysis tools are currently in their nascent stages and report improvements before fully integrating them into their decision-making processes. Building upon these findings, we discuss the opportunities and challenges in incorporating PPA into data-driven decision-making and its implications for future data and visual analytics systems.

For large, complex simulation models, simulation metamodeling is crucial for enabling simulation-based-optimization under uncertainty in operational settings where results are needed quickly. We enhance simulation metamodeling in two important ways. First, we use graph neural networks (GrNN) to allow the graphical structure of a simulation model to be treated as a metamodel input parameter that can be varied along with real-valued and integer-ordered inputs. Second, we combine GrNNs with generative neural networks so that a metamodel can rapidly produce not only a summary statistic like E [Y], but also a sequence of i.i.d. samples of Y or even a stochastic process that mimics dynamic simulation outputs. Thus a single metamodel can be used to estimate multiple statistics for multiple performance measures. Our metamodels can potentially serve as surrogate models in digital-twin settings. Preliminary experiments indicate the promise of our approach.

Background: A model that jointly simulates infectious diseases with common modes of transmission can serve as a decision-analytic tool to identify optimal intervention combinations for overall disease prevention. In the United States, sexually transmitted infections (STIs) are a huge economic burden, with a large fraction of the burden attributed to HIV. Data also show interactions between HIV and other sexually transmitted infections (STIs), such as higher risk of acquisition and progression of co-infections among persons with HIV compared to persons without. However, given the wide range in prevalence and incidence burdens of STIs, current compartmental or agent-based network simulation methods alone are insufficient or computationally burdensome for joint disease modeling. Further, causal factors for higher risk of coinfection could be both behavioral (i.e., compounding effects of individual behaviors, network structures, and care behaviors) and biological (i.e., presence of one disease can biologically increase the risk of another). However, the data on the fraction attributed to each are limited. Methods: We present a new mixed agent-based compartmental (MAC) framework for jointly modeling STIs. It uses a combination of a new agent-based evolving network modeling (ABENM) technique for lower-prevalence diseases and compartmental modeling for higher-prevalence diseases. As a demonstration, we applied MAC to simulate lower-prevalence HIV in the United States and a higher-prevalence hypothetical Disease 2, using a range of transmission and progression rates to generate burdens replicative of the wide range of STIs. We simulated sexual transmissions among heterosexual males, heterosexual females, and men who have sex with men (men only and men and women). Setting the biological risk of co-infection to zero, we conducted numerical analyses to evaluate the influence of behavioral factors alone on disease dynamics. Results: The contribution of behavioral factors to risk of coinfection was sensitive to disease burden, care access, and population heterogeneity and mixing. The contribution of behavioral factors was generally lower than observed risk of coinfections for the range of hypothetical prevalence studied here, suggesting potential role of biological factors, that should be investigated further specific to an STI. Conclusions: The purpose of this study is to present a new simulation technique for jointly modeling infectious diseases that have common modes of transmission but varying epidemiological features. The numerical analysis serves as proof-of-concept for the application to STIs. Interactions between diseases are influenced by behavioral factors, are sensitive to care access and population features, and are likely exacerbated by biological factors. Social and economic conditions are among key drivers of behaviors that increase STI transmission, and thus, structural interventions are a key part of behavioral interventions. Joint modeling of diseases helps comprehensively simulate behavioral and biological factors of disease interactions to evaluate the true impact of common structural interventions on overall disease prevention. The new simulation framework is especially suited to simulate behavior as a function of social determinants, and further, to identify optimal combinations of common structural and disease-specific interventions.

Business users make data-informed decisions by understanding the relationships between their data and business outcomes, typically tracked by key performance indicators (KPIs). To this end, they evaluate different decision scenarios, often using what-if analyses, which can benefit from predictive and prescriptive analytics (PPA). However, incorporating PPA into data analysis systems in a way that is easy to use and understand by business users can be challenging. Numerous studies have examined how data analysts use simple predictive analytics in their work. However, there is little systematic understanding and empirical data on business users’ data analysis practices for decision-making and the effectiveness of various PPA affordances in facilitating their analysis. To address this gap and lay the groundwork for designing better interactive systems for business data analysis, we conduct a study with 22 business users to better understand their data analysis workflows and elicit feedback on the effectiveness of four example PPA functionalities in business data analysis. We find that business users need to perform their data analysis without relying on data analysts due to four related reasons: the limited availability of data analysts in enterprises, the inefficiency of communicating with data analysts, the business pressure to make quick decisions, and the importance of incorporating domain knowledge, which data analysts often lack. Furthermore, we discover that business users rely on interactive manual analysis using spreadsheet-like applications, which makes coping with large, live, and complex datasets difficult. While, in general, business users don’t use PPA directly, their data analysis needs reinforce that they can benefit from it. We also find that PPA functionalities can increase business users’ decision-making speed and confidence. Additionally, our results suggest that they need help in PPA-driven analysis for preparing datasets, understanding the risk factors of predictive and prescriptive recommendations, and incorporating their domain knowledge. Finally, we share the implications of our findings for developing future business data analysis systems.

Many applications require generation of summaries tailored to the user’s information needs, i.e., their intent. Methods that express intent via explicit user queries fall short when query interpretation is subjective. Several datasets exist for summarization with objective intents where, for each document and intent (e.g., “weather”), a single summary suffices for all users. No datasets exist, however, for subjective intents (e.g., “interesting places”) where different users will provide different summaries. We present SUBSUME, the first dataset for evaluation of SUBjective SUMmary Extraction systems. SUBSUME contains 2,200 (document, intent, summary) triplets over 48 Wikipedia pages, with ten intents of varying subjectivity, provided by 103 individuals over Mechanical Turk. We demonstrate statistically that the intents in SUBSUME vary systematically in subjectivity. To indicate SUBSUME’s usefulness, we explore a collection of baseline algorithms for subjective extractive summarization and show that (i) as expected, example-based approaches better capture subjective intents than query-based ones, and (ii) there is ample scope for improving upon the baseline algorithms, thereby motivating further research on this challenging problem.