Mitigating Toxic Degeneration with Empathetic Data: Exploring the Relationship Between Toxicity and Empathy

Lahnala, Allison; Welch, Charles; Neuendorf, Béla; Flek, Lucie

doi:10.18653/v1/2022.naacl-main.363

Cited by 7 publications

(5 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, we see no improvement from using the EPITOME data. Similarly, recent work found separate empathy types were found to have different effects on toxicity reduction (Lahnala et al, 2022).…”

Section: Resultsmentioning

confidence: 83%

CAISA at WASSA 2022: Adapter-Tuning for Empathy Prediction

Lahnala¹,

Welch²,

Flek³

2022

Proceedings of the 12th Workshop on Computational Approaches to Subjectivity, Sentiment &Amp; Social Media Analysis

Self Cite

View full text Add to dashboard Cite

We build a system that leverages adapters, a light weight and efficient method for leveraging large language models to perform the task Empathy and Distress prediction tasks for WASSA 2022. In our experiments, we find that stacking our empathy and distress adapters on a pretrained emotion classification adapter performs best compared to full fine-tuning approaches and emotion feature concatenation. We make our experimental code publicly available. 1

show abstract

Section: Resultsmentioning

confidence: 83%

CAISA at WASSA 2022: Adapter-Tuning for Empathy Prediction

Lahnala¹,

Welch²,

Flek³

2022

Proceedings of the 12th Workshop on Computational Approaches to Subjectivity, Sentiment &Amp; Social Media Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…Specifically, we only fine-tuned on one medical dataset and one empathy dataset. As argued by Lahnala et al (2022), there are limitations in the way that empathy datasets are crafted, particularly concerning applications such as ours that aim for assessing cognitive empathic skills rather than surface-level emotional response.…”

Section: Discussionmentioning

confidence: 99%

Challenges of GPT-3-based Conversational Agents for Healthcare

Lechner,

Lahnala,

Welch

et al. 2023

Proceedings of the Conference Recent Advances in Natural Language Processing - Large Language Models for Natural Language Proce

View full text Add to dashboard Cite

The potential to provide patients with faster information access while allowing medical specialists to concentrate on critical tasks makes medical domain dialog agents appealing. However, the integration of large-language models (LLMs) into these agents presents certain limitations that may result in serious consequences. This paper investigates the challenges and risks of using GPT-3-based models for medical question-answering (MedQA). We perform several evaluations contextualized in terms of standard medical principles. We provide a procedure for manually designing patient queries to stress-test high-risk limitations of LLMs in MedQA systems. Our analysis reveals that LLMs fail to respond adequately to these queries, generating erroneous medical information, unsafe recommendations, and content that may be considered offensive.

show abstract

“…One of the decoding approaches used to mitigate toxicity in model output was presented in Lahnala et al [48], where they modified the model's probabilities based on an expert-desired attribute, namely empathy. By favouring tokens that express a more empathic response, they observed a correlation between empathy and toxicity, albeit dependent on the context.…”

Section: Decoding Techniquesmentioning

confidence: 99%

A Systematic Review of Toxicity in Large Language Models: Definitions, Datasets, Detectors, Detoxification Methods and Challenges

Villate-Castillo,

Ser,

Urquijo

2024

Preprint

View full text Add to dashboard Cite

The emergence of the transformer architecture has ushered in a new era of possibilities, showcasing remarkable capabilities in generative tasks exemplified by models like GPT4o, Claude 3, and Llama 3. However, these advancements come with a caveat: predominantly trained on data gleaned from social media platforms, these systems inadvertently perpetuate societal biases and toxicity. Recognizing the paramount importance of AI Safety and Alignment, our study embarks on a thorough exploration through a comprehensive literature review focused on toxic language. Delving into various definitions, detection methodologies, and mitigation strategies, we aim to shed light on the complexities of this issue. While our focus primarily centres on transformer-based architectures, we also acknowledge and incorporate existing research within the realm of deep learning. Through our investigation, we uncover a multitude of challenges inherent in toxicity mitigation and detection models. These challenges range from inherent biases and generalization issues to the necessity for standardized definitions of toxic language and the quality assurance of dataset annotations. Furthermore, we emphasize the significance of transparent annotation processes, resolution of annotation disagreements, and the enhancement of Large Language Models (LLMs) robustness. Additionally, we advocate for the creation of standardized benchmarks to gauge the effectiveness of toxicity mitigation and detection methods. Addressing these challenges is not just imperative, but pivotal in advancing the development of safer and more ethically aligned AI systems.

show abstract

Mitigating Toxic Degeneration with Empathetic Data: Exploring the Relationship Between Toxicity and Empathy

Cited by 7 publications

References 23 publications

CAISA at WASSA 2022: Adapter-Tuning for Empathy Prediction

CAISA at WASSA 2022: Adapter-Tuning for Empathy Prediction

Challenges of GPT-3-based Conversational Agents for Healthcare

A Systematic Review of Toxicity in Large Language Models: Definitions, Datasets, Detectors, Detoxification Methods and Challenges

Contact Info

Product

Resources

About