From Chat Logs to Collective Insights: Aggregative Question Answering

TL;DR

This paper introduces a new task called Aggregative Question Answering that involves reasoning over large-scale chatbot conversations to extract collective insights, supported by a benchmark dataset and experimental analysis.

Contribution

It proposes the novel task of Aggregative Question Answering, creates the WildChat-AQA benchmark, and highlights the challenges faced by current methods in reasoning and computational efficiency.

Findings

Existing methods struggle with reasoning accuracy.

Current approaches are computationally expensive.

The benchmark enables future research in collective conversational insights.

Abstract

Conversational agents powered by large language models (LLMs) are rapidly becoming integral to our daily interactions, generating unprecedented amounts of conversational data. Such datasets offer a powerful lens into societal interests, trending topics, and collective concerns. Yet, existing approaches typically treat these interactions as independent and miss critical insights that could emerge from aggregating and reasoning across large-scale conversation logs. In this paper, we introduce Aggregative Question Answering, a novel task requiring models to reason explicitly over thousands of user-chatbot interactions to answer aggregative queries, such as identifying emerging concerns among specific demographics. To enable research in this direction, we construct a benchmark, WildChat-AQA, comprising 6,027 aggregative questions derived from 182,330 real-world chatbot conversations. Experiments show that existing methods either struggle to reason effectively or incur prohibitive computational costs, underscoring the need for new approaches capable of extracting collective insights from large-scale conversational data.