SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures

By Megan Ung, Jing Xu, Y-Lan Boureau. 2021

55 citations recorded by the source corpus.

Current open-domain conversational models can easily be made to talk in inadequate ways. Online learning from conversational feedback given by the conversation partner is a promising avenue for a model to improve and adapt, so as to generate fewer of these safety failures. However, current state-of-the-art models tend to react to feedback with defensive or oblivious responses. This makes for an unpleasant experience and may discourage conversation partners from giving feedback in the future. This work proposes SaFeRDialogues, a task and dataset of graceful responses to conversational feedback about safety failures.We collect a dataset of 8k dialogues demonstrating safety failures, feedback signaling them, and a response acknowledging the feedback. We show how fine-tuning on this dataset results in conversations that human raters deem considerably more likely to lead to a civil conversation, without sacrificing engagingness or general conversational ability.

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

OPT: Open Pre-trained Transformer Language Models (2022, background): “The first, SaferDialogues (Ung et al., 2021), measures the ability to recover from explicit safety failures, usually in the form of apologizing or recognizing its mistake.”
Guiding the Release of Safer E2E Conversational AI through Value Sensitive Design (2022, background): “…situations, by directly generating toxic or otherwise harmful content, by agreeing with offensive statements uttered by the conversation partner (Dinan et al., 2022), or by responding defensively or dismissively when provided with corrective feedback by the conversation partner (Ung et al., 2021).”
BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage (2022, methodology): “, 2021b) X Recovery & Feedback SaFeRDialogues (Ung et al., 2022) X FITS (Xu et al.”
Using In-Context Learning to Improve Dialogue Safety (2023, methodology): “7B which has been fine-tuned on safe responses from Proso-cialDialog and SaFeRDialogues.”
Towards Identifying Social Bias in Dialog Systems: Framework, Dataset, and Benchmark (2022, background): “The understanding of biases may inevitably have some differences among populations and cultures (Schmidt and Wiegand, 2017; Ung et al., 2022).”
ProsocialDialog: A Prosocial Backbone for Conversational Agents (2022, background): “Tackling a related but different task, Ung et al. (2021) compile a dataset of feedback to dialogue agent failures along with recovery utterances (e.g., apologies).”
DialGuide: Aligning Dialogue Model Behavior with Developer Guidelines (2022, background): “Other approaches include strategies for responding to problematic contexts, such as steering away from toxicity (Baheti et al., 2021; Arora et al., 2022), using apologies (Ung et al., 2022), and non-sequiturs (Xu et al., 2021b).”
Why So Toxic?: Measuring and Triggering Toxic Behavior in Open-Domain Chatbots (2022, methodology): “We evaluate our attack on the Recovery (fine-tuned) BBl, which is fixed with SD and released by [62].”