Jury Learning: Integrating Dissenting Voices into Machine Learning Models

Gordon, Mitchell; Lam, Michelle; Park, Joon Sung; Patel, Kayur; Hancock, Jeff; Hashimoto, Tatsunori; Bernstein, Michael S.

doi:10.1145/3491102.3502004

Cited by 69 publications

(74 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…social norm prediction (Jiang et al, 2021) and toxicity detection (Halevy et al, 2021;. We encourage conscious efforts to recruit diverse pools of annotators so multiple perspectives are considered, and future work on modeling reaction frames can consider learning algorithms that mitigate harmful effects of biases, depending on use case (Khalifa et al, 2021;Gordon et al, 2022).…”

Section: Future Directions and Limitations Of Reaction Framesmentioning

confidence: 99%

Misinfo Reaction Frames: Reasoning about Readers’ Reactions to News Headlines

Gabriel¹,

Hallinan²,

Sap³

et al. 2022

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

Even to a simple and short news headline, readers react in a multitude of ways: cognitively (e.g. inferring the writer's intent), emotionally (e.g. feeling distrust), and behaviorally (e.g. sharing the news with their friends). Such reactions are instantaneous and yet complex, as they rely on factors that go beyond interpreting factual content of news. Headline Reader"they are lying to us." Writer IntentReader Perception "feel distrust" Reader Action"not trust the government"Likelihood of Spread "likely to share or read" (4/5)

show abstract

Section: Future Directions and Limitations Of Reaction Framesmentioning

confidence: 99%

Misinfo Reaction Frames: Reasoning about Readers’ Reactions to News Headlines

Gabriel¹,

Hallinan²,

Sap³

et al. 2022

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

show abstract

“…For Simplicity and Misunderstanding, one could maintain a list of everyday concepts and a list of concepts with multiple interpretations for ease of automatic check. Recent work on jury learning (Gordon et al 2022) proposed a method to conduct automatic pseudo-human value judgement with machine learning models, which can be an alternative to expertbased quality evaluation, while accounting for the subjectivity of each dimension.…”

Section: Discussionmentioning

confidence: 99%

It Is like Finding a Polar Bear in the Savannah! Concept-Level AI Explanations with Analogical Inference from Commonsense Knowledge

He¹,

Balayn²,

Buijsman³

et al. 2022

HCOMP

View full text Add to dashboard Cite

With recent advances in explainable artificial intelligence (XAI), researchers have started to pay attention to concept-level explanations, which explain model predictions with a high level of abstraction. However, such explanations may be difficult to digest for laypeople due to the potential knowledge gap and the concomitant cognitive load. Inspired by recent work, we argue that analogy-based explanations composed of commonsense knowledge may be a potential solution to tackle this issue. In this paper, we propose analogical inference as a bridge to help end-users leverage their commonsense knowledge to better understand the concept-level explanations. Specifically, we design an effective analogy-based explanation generation method and collect 600 analogy-based explanations from 100 crowd workers. Furthermore, we propose a set of structured dimensions for the qualitative assessment of analogy-based explanations and conduct an empirical evaluation of the generated analogies with experts. Our findings reveal significant positive correlations between the qualitative dimensions of analogies and the perceived helpfulness of analogy-based explanations. These insights can inform the design of future methods for the generation of effective analogy-based explanations. We also find that the understanding of commonsense explanations varies with the experience of the recipient user, which points out the need for further work on personalization when leveraging commonsense explanations.

show abstract

“…Training on soft instead of hard labels can improve robustness and generalization (Pereyra et al 2017;Müller, Kornblith, and Hinton 2019). Soft labels have been constructed using smoothing mechanisms (Szegedy et al 2016), auxiliary teacher networks as in knowledge distillation (Hinton et al 2015;Gou et al 2021), and aggregate human annotations (Sharmanska et al 2016;Peterson et al 2019;Recht et al 2019;Uma et al 2020;Gordon et al 2021Gordon et al , 2022Uma, Almanea, and Poesio 2022;Koller, Kauermann, and Zhu 2022). While the first two methods have led to significant advances in model performance, hand-crafted or learned soft labels often rely on hard labels, which tend to be impoverished representations of human precepts over datapoints.…”

Section: Related Workmentioning

confidence: 99%

Eliciting and Learning with Soft Labels from Every Annotator

Collins

Bhatt

Weller

2022

HCOMP

View full text Add to dashboard Cite

The labels used to train machine learning (ML) models are of paramount importance. Typically for ML classification tasks, datasets contain hard labels, yet learning using soft labels has been shown to yield benefits for model generalization, robustness, and calibration. Earlier work found success in forming soft labels from multiple annotators' hard labels; however, this approach may not converge to the best labels and necessitates many annotators, which can be expensive and inefficient. We focus on efficiently eliciting soft labels from individual annotators. We collect and release a dataset of soft labels (which we call CIFAR-10S) over the CIFAR-10 test set via a crowdsourcing study (N=248). We demonstrate that learning with our labels achieves comparable model performance to prior approaches while requiring far fewer annotators -- albeit with significant temporal costs per elicitation. Our elicitation methodology therefore shows nuanced promise in enabling practitioners to enjoy the benefits of improved model performance and reliability with fewer annotators, and serves as a guide for future dataset curators on the benefits of leveraging richer information, such as categorical uncertainty, from individual annotators.

show abstract

Jury Learning: Integrating Dissenting Voices into Machine Learning Models

Cited by 69 publications

References 50 publications

Misinfo Reaction Frames: Reasoning about Readers’ Reactions to News Headlines

Misinfo Reaction Frames: Reasoning about Readers’ Reactions to News Headlines

It Is like Finding a Polar Bear in the Savannah! Concept-Level AI Explanations with Analogical Inference from Commonsense Knowledge

Eliciting and Learning with Soft Labels from Every Annotator

Contact Info

Product

Resources

About