When Statisticians Can Also Be AI Engineers, What Could We Build?

Ruoqing Zhu

Home/Archive/arXiv

arXiv · 2610.32105

When Statisticians Can Also Be AI Engineers, What Could We Build?

Ruoqing Zhu

View original source ↗Read source document ↗Export metadata ↓

Abstract

Statistical software makes methods widely available. Good statistical work, however, often requires conversations that clarify the scientific question, how the data were collected, and what conclusions the evidence can support. Coding agents and reusable AI infrastructure create opportunities for statisticians to incorporate these processes into agentic tools for research, practice, and education. A review of agentic AI applications for data analysis and study design distinguishes general human participation from human interaction guided by statistical principles. We examine how statistical expertise can guide the design and evaluation of domain-specific interactive harnesses. Two practical entry points illustrate how statisticians can begin building: an implemented causal-consulting skill within an existing agent environment and Model Forge, a custom harness for interactive model and algorithm development. These design examples connect statistical principles to choices about context, memory, tools, and interfaces, and motivate evaluation of both computational correctness and the scientific consequences of interaction. We conclude with a broader interdisciplinary research agenda on adaptive human-agent interaction, including selective information acquisition, evolving user preferences, and evaluation under uncertainty.

Explore related subjects

stat.OT stat.AP

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Ruoqing Zhu. 2026-10-23. When Statisticians Can Also Be AI Engineers, What Could We Build?. https://arxiv.org/abs/2610.32105

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Heterogeneity and Subgrouping in Modern Statistics and AI

Heterogeneity is a defining feature of modern data: individuals, subpopulations, sites, modalities, tasks, and time points may differ in distributions, associations, treatment effects, or decision objectives. This review develops a unified perspective in which heterogeneous analysis is a problem of structured information borrowing: deciding what should be shared, what should remain distinct, and how uncertainty about that structure should enter inference and decisions. We connect classical stratification, interaction models, finite mixtures, mixed-effects models, and empirical Bayes with subgroup discovery, individualized learning, synthetic data generation, large language models, federated learning, privacy, security, and mobile health. Precision medicine provides a central example: subgroup rules offer interpretable approximations to individualized response surfaces, whereas partially pooled models stabilize patient-specific estimation. Across applications, the same tensions recur: population averages can conceal consequential variation, unrestricted personalization is unstable, adaptive subgroup discovery creates selection bias, and borrowing across incompatible sources causes negative transfer. These difficulties are amplified in contemporary AI, where sparse user histories, entangled representations, heterogeneous preferences, feedback loops, distribution shift, and privacy constraints make reliable personalization and evaluation especially challenging. We conclude with an agenda for calibrated personalization that integrates multilevel representations, selective borrowing, subgroup-aware evaluation, and valid uncertainty quantification.

stat.OT↗

arXiv2026

Movement synchronization in complex and dynamic team coordination

Mutual anticipation of teammates' actions enables efficient coordination that achieves a common goal and high performance. In team sports involving direct competition, such implicit and nonverbal interactions within short periods are required. If players begin moving after observing their teammates, gaps may emerge, allowing opponents to interfere. When mutual anticipation functions appropriately, players' interactions are smooth without gaps, and their movements are expected to become synchronized. Synchronization represents a temporally stable structure whose mechanisms have been examined. However, few studies have investigated synchronization in complex and dynamic real-world coordination involving heterogeneous roles and evolving interactions within short periods. The examinations have been essentially limited to determining whether there are statistical differences from a baseline. In our approach, we utilized team sports and introduced a statistical method to provide detailed probabilistic evaluations of these differences. The purpose of this study was to extract the temporal components of movement using 3-on-3 basketball. In a field experiment investigating offensive coordination, we calculated the relative phases in which players approached or moved away from their teammates during mini-games. The frequency distributions were estimated using Bayesian inference and compared before and after receiving advice from the coach. The results showed that the posterior probabilities that the frequencies in synchronization-related bins were higher after the advice than before the advice were 70\% or higher. As a pilot study, these findings may provide a first step toward understanding coordination processes related to mutual anticipation based on a common ground of team strategy.

stat.OT↗

arXiv2026

DagFlow AI: Agentic Synthetic Data Generation for Personalized Statistics Learning

To provide personalized learning experiences in courses that teach data analysis skills, instructors often search for datasets that appeal to a student's interests, align with specific learning objectives, and are appropriate for the student's skill level. Datasets that match all three criteria can be difficult to find. DagFlow-AI is an agentic AI data generation tool that creates realistic, synthetic datasets tailored to these criteria. Rather than mimicking existing datasets or generating observations directly, DagFlow-AI uses the broad knowledge and reasoning capabilities of large language models (LLMs) to specify the components of transparent, reproducible data-generating mechanisms from which observations are generated programmatically. We illustrate how the tool can be used to construct a realistic, synthetic dataset and describe how it aligns with both the Guidelines for Assessment and Instruction of Statistics Education (GAISE) and the Problem, Plan, Data, Analysis, and Conclusion (PPDAC) framework for teaching data analysis

stat.OT↗