Using Generative Agents to Create Tip Sheets for Investigative Data Reporting

TL;DR

The paper investigates using a triad of generative AI agents—analyst, reporter, and editor—to generate tip sheets that surface potentially newsworthy insights from datasets for investigative reporting. It implements a four-step pipeline (question generation, analytical planning, execution with feedback, and compilation) via GPT-4 through the OpenAI Assistants API. Compared to a non-agent baseline, the agent-based approach yields higher overall newsworthiness and generally strong validity, though results vary by dataset and some projects present challenges. The work demonstrates the potential of AI-driven leads to augment computational journalism while underscoring the ongoing need for editorial validation and human judgment in reporting.

Abstract

This paper introduces a system using generative AI agents to create tip sheets for investigative data reporting. Our system employs three specialized agents--an analyst, a reporter, and an editor--to collaboratively generate and refine tips from datasets. We validate this approach using real-world investigative stories, demonstrating that our agent-based system generally generates more newsworthy and accurate insights compared to a baseline model without agents, although some variability was noted between different stories. Our findings highlight the potential of generative AI to provide leads for investigative data reporting.

Figures (2)

• Figure 1: Given a dataset and a description of the dataset, our generative agents pipeline outputs a tip sheet.
• Figure 2: A complete overview of the pipeline. Each box represents a single prompt assigned to one of the three agents.