Summary
teLLMe is a system by Qiwei Li and Jorge Ortiz (arXiv 2607.15254) for exploratory causal analysis of urban driving datasets. Since traffic data is observational and collected without interventions, causal questions like how rain changes traffic density are hard to answer. teLLMe builds a structured event table from dashcam annotations, applies PC-algorithm causal structure learning with bootstrap stability checks, and performs query-specific effect estimation using linear regression and DoWhy. A schema-aware LLM maps natural-language questions into structured causal queries where users specify treatments, outcomes, and subpopulations. Results are returned as a Causal Card summarizing effect estimates, adjustment sets, DAG support, and assumptions, plus a short natural-language explanation. Case studies on BDD-derived traffic events surface plausible relationships among weather, peak hours, and traffic density while making uncertainty explicit. The tool targets hypothesis generation and expert reasoning, not definitive causal claims.
Paper Overview
- Research area: cs.AI, cs.HC
- Authors: Qiwei Li, Jorge Ortiz
- Published: 2026-07-16
- arXiv: 2607.15254
Introduction
Traffic agencies now have access to large volumes of video-derived data for studying safety and congestion. Most of these data are observational and collected without interventions, which makes causal questions such as "How would rain change traffic density?" difficult to answer.The teLLMe System
teLLMe is a system for exploratory causal analysis of urban driving datasets. Key components:
- Structured event table built from dashcam annotations
- Causal structure learning with the PC algorithm
- Bootstrap-based stability checks for robustness
- Query-specific effect estimation using linear regression and DoWhy
Natural-language questions are mapped to structured causal queries through a schema-aware LLM, enabling users to specify treatments, outcomes, and subpopulations.
Output: Causal Cards
teLLMe returns a "Causal Card" that summarizes:
- Effect estimates
- Adjustment sets
- DAG support
- Assumptions
This is followed by a short natural-language explanation.
Evaluation
Case studies on BDD-derived traffic events show that the system can surface plausible relationships involving weather, peak hours, and traffic density, while making uncertainty and modeling choices explicit.
Positioning
The system is designed as a tool for hypothesis generation and expert reasoning rather than a source of definitive causal claims.
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178446935