An official website of the United States government
Three studies developed and validated a linguistic dictionary to measure negative affective polarization in English and Spanish political texts. It captures three dimensions: negative affect, delegitimization, and political context. In the first study, two independent judges evaluated the candidate words, and reliability indicators were calculated, showing acceptable values for short texts (.572 in English, .541 in Spanish) and higher values for larger corpora (.964 in English, .957 in Spanish). The second study tested discriminant validity by comparing negative affective polarization scores in social media comments on politics and entertainment. Results showed significantly higher polarization scores in political content, confirming the dictionary's validity. The third study compared the dictionary to an existing online polarization measure, finding greater coverage and alignment with the construct. Additionally, it was observed that polarization scores were higher in texts containing hate speech compared to those where it was absent. The findings suggest that the dictionary in both languages have strong psychometric properties, making it a valuable tool for analyzing online content, particularly social media comments. It can be used as an independent measure or as input for machine and deep learning models.
more »
« less
From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models
Language models (LMs) are pretrained on diverse data sources—news, discussion forums, books, online encyclopedias. A significant portion of this data includes facts and opinions which, on one hand, celebrate democracy and diversity of ideas, and on the other hand are inherently socially biased. Our work develops new methods to (1) measure media biases in LMs trained on such corpora, along social and economic axes, and (2) measure the fairness of downstream NLP models trained on top of politically biased LMs. We focus on hate speech and misinformation detection, aiming to empirically quantify the effects of political (social, economic) biases in pretraining data on the fairness of high-stakes social-oriented tasks. Our findings reveal that pretrained LMs do have political leanings which reinforce the polarization present in pretraining corpora, propagating social biases into hate speech predictions and media biases into misinformation detectors. We discuss the implications of our findings for NLP research and propose future directions to mitigate unfairness.
more »
« less
- PAR ID:
- 10433148
- Date Published:
- Journal Name:
- ACL: Annual Meeting of the Association for Computational Linguistics
- Page Range / eLocation ID:
- 11737–11762
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Hate speech and offensive language are rampant on social media. Machine learning has provided a way to moderate foul language at scale. However, much of the current research focuses on overall performance. Models may perform poorly on text written in a minority dialectal language. For instance, a hate speech classifier may produce more false positives on tweets written in African-American Vernacular English (AAVE). To measure these problems, we need text written in both AAVE and Standard American English (SAE). Unfortunately, it is challenging to curate data for all linguistic styles in a timely manner—especially when we are constrained to specific problems, social media platforms, or by limited resources. In this paper, we answer the question, “How can we evaluate the performance of classifiers across minority dialectal languages when they are not present within a particular dataset?” Specifically, we propose an automated fairness fuzzing tool called FuzzE to quantify the fairness of text classifiers applied to AAVE text using a dataset that only contains text written in SAE. Overall, we find that the fairness estimates returned by our technique moderately correlates with the use of real ground-truth AAVE text. Warning: Offensive language is displayed in this manuscript.more » « less
-
Abstract Humans use language toward hateful ends, inciting violence and genocide, intimidating and denigrating others based on their identity. Despite efforts to better address the language of hate in the public sphere, the psychological processes involved in hateful language remain unclear. In this work, we hypothesize that morality and hate are concomitant in language. In a series of studies, we find evidence in support of this hypothesis using language from a diverse array of contexts, including the use of hateful language in propaganda to inspire genocide (Study 1), hateful slurs as they occur in large text corpora across a multitude of languages (Study 2), and hate speech on social-media platforms (Study 3). In post hoc analyses focusing on particular moral concerns, we found that the type of moral content invoked through hate speech varied by context, with Purity language prominent in hateful propaganda and online hate speech and Loyalty language invoked in hateful slurs across languages. Our findings provide a new psychological lens for understanding hateful language and points to further research into the intersection of morality and hate, with practical implications for mitigating hateful rhetoric online.more » « less
-
Although many fairness criteria have been proposed to ensure that machine learning algorithms do not exhibit or amplify our existing social biases, these algorithms are trained on datasets that can themselves be statistically biased. In this paper, we investigate the robustness of existing (demographic) fairness criteria when the algorithm is trained on biased data. We consider two forms of dataset bias: errors by prior decision makers in the labeling process, and errors in the measurement of the features of disadvantaged individuals. We analytically show that some constraints (such as Demographic Parity) can remain robust when facing certain statistical biases, while others (such as Equalized Odds) are significantly violated if trained on biased data. We provide numerical experiments based on three real-world datasets (the FICO, Adult, and German credit score datasets) supporting our analytical findings. While fairness criteria are primarily chosen under normative considerations in practice, our results show that naively applying a fairness constraint can lead to not only a loss in utility for the decision maker, but more severe unfairness when data bias exists. Thus, understanding how fairness criteria react to different forms of data bias presents a critical guideline for choosing among existing fairness criteria, or for proposing new criteria, when available datasets may be biased.more » « less
-
Language models (LMs) are pretrained to imitate text from large and diverse datasets that contain content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, among others. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the alignment and capabilities of pretrained LMs. We find a Pareto-optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversarially-chosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
