Ziyi Wang, Yuxuan Lu, Yimeng Zhang, Pei Chen, Ziwei Dong, Jing Huang, Jiri Gesi, Xianfeng Tang, Chen Luo, Qun Liu, Yisi Sang, Hanqing Lu, Manling Li, Jin Lai, Dakuo Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The adoption of responsible data science (RDS) practices in AI development remains inadequate despite growing awareness of algorithmic harms. One measure of success is by observing practitioners’ behaviors – namely, their adoption of responsible sequences of behaviors in their model building practice. This paper evaluates two interventions for changing problematic behaviors: (i) a motivational priming intervention that introduces short, relevant stories, and (ii) a fairness toolkit (Aequitas)—to bridge the gap between ethical principles and practitioner behavior. Through a mixed-methods study with data scientists (N=12), we assess how these interventions influence fairness practices, model outcomes, and cognitive load across credit risk and income classification tasks. Results indicate that both interventions were efficient in promoting responsible data science behaviors and improving the delivered models’ fairness, while maintaining baseline accuracy. We argue that effective behavior change interventions must balance technical tooling with motivational scaffolding to provide actionable insights for fostering sustainable RDS practices.

Deploying a conversational shopping agent at production scale remains challenging despite rapid advances in large language models. Unlike research prototypes, production systems must satisfy strict requirements on accuracy, latency, reliability, and cost while serving millions of customers. We share our year-long journey of building a production shopping agent at scale and show that combining agentic reasoning with production-aware retrieval, tool orchestration, and system optimization enables a single shopping agent to support product discovery, shopping question answering, and agentic actions under real-world traffic. Our experience shows that LLM-based agents can be reliably operated at Amazon scale and provides practical design principles for industrial search and recommendation systems.

Data science pipelines inform and influence many daily decisions, from what we buy to who we work for and even where we live. When designed incorrectly, these pipelines can easily propagate social inequity and harm. Traditional solutions are technical in nature; e.g., mitigating biased algorithms. In this vision paper, we introduce a novel lens for promoting responsible data science using theories of behavior change that emphasize not only technical solutions but also the behavioral responsibility of practitioners. By integrating behavior change theories from cognitive psychology with data science workflow knowledge and ethics guidelines, we present a new perspective on responsible data science. We present example data science interventions in machine learning and visual data analysis, contextualized in behavior change theories that could be implemented to interrupt and redirect potentially suboptimal or negligent practices while reinforcing ethically conscious behaviors. We conclude with a call to action to our community to explore this new research area of behavior change interventions for responsible data science.

This manuscript considers distributed optimization problems in systems with cooperative-collaborative relationships, involving two groups of nodes, each with its own optimization problem, but with a coupled communication topology. For the signed graph representing the cooperation and collaboration between agents, this manuscript introduces DIG-JOR, a discrete-time distributed algorithm that consists of three key modules: an inexact consensus and gradient descent module, a group gradient-tracking module, and a dynamic Jacobi over-relaxation (JOR) inverse-tracking module. To support the convergence analysis of the distributed optimization algorithm, this manuscript proposes the Multi-Loop Small Gain Theorem. Under the assumption of strong convexity and with appropriately chosen step sizes, it is proved that the DIG-JOR algorithm converges to the optimal solutions of both groups at an R-linear rate. The theoretical results are validated through a simulation example.

Detecting straight lines in video plays a fundamental role in camera-based industrial automation. With the increasing demands on production efficiency, detection speed has become one of the bottlenecks for highly efficient industrial automation. Because of data dependence and hardware limitations, existing vision systems based on central processing unit/graphics processing unit are unable to detect straight lines at an ultrahigh speed. This article addresses this problem and proposes a hardware-friendly Hough transform that can be implemented in fully parallel for the ultrahigh-speed detection, because of the following two key features: it processes multiple pixels in parallel and directly calculates line parameters while capturing the current frame; and it simultaneously initializes the Hough parameter space and votes in the Hough parameter space without any delay. Based on the proposed hardware-friendly Hough transform, its chip-level implementation and system-level hardware design are presented. Experimental results show that the main benefits of the proposed architecture are in real-time performances at a high frame rate (784 frames/s) and an ultralow delay (0.7749 ms/frame).

Many problems in industrial smart manufacturing, such as process operational optimization and decision-making, can be regarded as distributed nonconvex optimization problems, whose goal is to utilize distributed nodes to cooperatively search for the minimal value of the global objective function. With the consideration of data transmission mode, transmission condition, and communication waste in industrial applications, it is meaningful to study the distributed nonconvex optimization problem with an event-triggered strategy over time-varying directed networks. To solve such a problem, a distributed nonconvex event-triggered algorithm is proposed in this article. Under some assumptions on local objective functions, gradients, and step sizes, the convergence of the proposed event-triggered algorithm to the local minimum is established theoretically. Moreover, it is obtained that the proposed distributed event-triggered algorithm has a convergence rate of$O(1/\ln (t))$. Finally, two examples of industrial systems are provided to validate the effectiveness of the proposed algorithm.

The economic dispatch problem (EDP) plays a fundamental and significant role in smart grids. Its purpose is to decide the output power of every generator in smart grids for achieving the minimal generation cost. With advantages in flexibility, robustness, and scalability, it is desirable to apply distributed optimization methods to solve EDPs. In most existing distributed optimization approaches, all generators explicitly exchange their states with neighbors to obtain the optimal solution, which may result in disclosing the privacy information of generators. This problem becomes worse if there are some adversaries aimed at inferring privacy information from the communication network for nefarious purposes. For privacy preservation, a privacy preserving distributed optimization algorithm over time-varying directed communication networks is proposed in this article by adding conditional noises to the exchanged states. It is proved that this proposed algorithm is able to solve the EDP. Moreover, the convergence rate and privacy analysis of the proposed algorithm are also shown in this article. An example is provided to confirm the effectiveness of this proposed algorithm.

The economic dispatch problem (EDP) is one of the fundamental and important problems in power systems. The objective of EDP is to determine the output generation of generators to minimize the total generation cost under various constraints. In this article, a finite-time consensus-based distributed optimization algorithm is proposed to solve EDP. It is only required that each device in the communication network has access to its own local generation cost function, designed virtual local demand and its neighbors' local optimization variables. The proposed finite-time algorithm can solve EDP, if the gain parameters in the algorithm satisfy some conditions under undirected and connected time-varying graphs. Moreover, the bounded or linear increasing assumption on the gradient and subgradient of objecive functions is relaxed in this algorithm. Examples under several cases are provided to verify the effectiveness of the proposed distributed optimization algorithm.