Detecting Egregious Conversations between Customers and Virtual Agents

TL;DR

This paper presents a method for detecting egregious customer-agent conversations in virtual customer service systems by analyzing behavioral cues and interaction patterns, significantly improving detection accuracy over text-based methods.

Contribution

It introduces a novel approach combining behavioral and interaction features for egregious conversation detection, demonstrating its effectiveness across different domains.

Findings

20% improvement in F1-score over textual features

Features are effective across diverse domains

Behavioral cues are key indicators of egregious conversations

Abstract

Virtual agents are becoming a prominent channel of interaction in customer service. Not all customer interactions are smooth, however, and some can become almost comically bad. In such instances, a human agent might need to step in and salvage the conversation. Detecting bad conversations is important since disappointing customer service may threaten customer loyalty and impact revenue. In this paper, we outline an approach to detecting such egregious conversations, using behavioral cues from the user, patterns in agent responses, and user-agent interaction. Using logs of two commercial systems, we show that using these features improves the detection F1-score by around 20% over using textual features alone. In addition, we show that those features are common across two quite different domains and, arguably, universal.