A Crowd-based Evaluation of Abuse Response Strategies in Conversational Agents

By Amanda Cercas Curry, Verena Rieser. 2019

36 citations recorded by the source corpus.

How should conversational agents respond to verbal abuse through the user? To answer this question, we conduct a large-scale crowd-sourced evaluation of abuse response strategies employed by current state-of-the-art systems. Our results show that some strategies, such as “polite refusal”, score highly across the board, while for other strategies demographic factors, such as age, as well as the severity of the preceding abuse influence the user’s perception of which response is appropriate. In addition, we find that most data-driven models lag behind rule-based or commercial systems in terms of their perceived appropriateness.

Subfields: Artificial Intelligence, Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling (2021, background): “Cer-cas Curry & Rieser (2019) assess human over-hearers’ evaluations of these strategies, ﬁnding varying preferences among different demographic groups.”
Recipes for Safety in Open-domain Chatbots (2020, methodology): “What is offensive to one may not be offensive to another (Curry and Rieser, 2019).”
Bot-Adversarial Dialogue for Safe Conversational Agents (2021, background): “While several different “safe” response strategies can be considered (Curry and Rieser, 2019; Paranjape et al., 2020), in this work we respond with a non-sequitur: we select a topic at random from 1,087 topics judged as safe from the Wizard of Wikipedia conversational topic list (Dinan et al.,…”
Neural Generation Meets Real People: Towards Emotionally Engaging Mixed-Initiative Conversations (2020, background): “Users sometimes give offensive or critical utterances, and it is important for our bot to handle these appropriately (Curry and Rieser, 2018, 2019).”
SafetyKit: First Aid for Measuring Safety in Open-domain Conversational Systems (2022, background): “And in a follow-up study, Cer-4115 cas Curry and Rieser (2019) assess human over-hearers’ evaluations of these strategies, finding varying preferences among different demographic groups.”
Alexa, Google, Siri: What are Your Pronouns? Gender and Anthropomorphism in the Design and Perception of Conversational Assistants (2021, background): “…non-gendered voices (Sutton, 2020), and voice assistants that are designed as or perceived to be female attract abusive behaviour (Cercas Curry and Rieser, 2019, 2018), designers may consider attempting to reddress the gender imbalance by designing assistants with servile roles to be…”
Assessing Political Prudence of Open-domain Chatbots (2021, methodology): “Besides measurement and mitigation methods, Curry and Rieser (2019); Chin and Yi (2019); Chin et al. (2020) have studied different response methods (empathy, counter-attack, question etc.) to adversarial attacks from users.”
Conversational AI: Social and Ethical Considerations (2019, background): “In this scenario the domain and goals of the agent are important, but the user demographic is the most in-ﬂuential factor when designing the agent’s response strategy [10].”
ConvAbuse: Data, Analysis, and Benchmarks for Nuanced Abuse Detection in Conversational AI (2021, background): “Following this report, several recent works have investigated possible abuse mitigation strategies (Cercas Curry and Rieser, 2018, 2019; Chin and Yi, 2019; Ma et al., 2019).”
Conversational Assistants and Gender Stereotypes: Public Perceptions and Desiderata for Voice Personas (2020, unknown): “(Fowler, 2011).”