Large language models associate Muslims with violence

By Abubakar Abid, Maheen Farooqi, James Zou. 2021

186 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Safety, Robustness & Interpretability

Citations in the corpus

LaMDA: Language Models for Dialog Applications (2022, background): “, [58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72]), and inappropriately revealing personally identifying information (PII) from training data [73].”
The Capacity for Moral Self-Correction in Large Language Models (2023, background): “Large language models exhibit harmful social biases [1, 6, 8, 11, 15, 24, 29, 50, 61] that can sometimes get worse for larger models [2, 18, 20, 43].”
Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned (2022, background): “Large language models exhibit a wide range of harmful behaviors such as reinforcing social biases (e.g., [47, 28, 1, 33, 7]), generating offensive or toxic outputs [25], leaking personally identiﬁable information from the training data [13], aiding in disinformation campaigns [12], generating…”
Are aligned neural networks adversarially aligned? (2023, background): “Second, by virtue of faithfully modeling the distribution of Internet text, the base models tend to reflect and even exacerbate biases [Abid et al., 2021], toxicity, and profanity [Welbl et al., 2021, Dixon et al., 2018] present in the training data.”
Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions (2023, background): “There is already an emerging body of scholarly work (Abid et al., 2021; Weidinger et al., 2021; Bommasani et al., 2021; Deshpande et al., 2023; Wei et al., 2023; Zack et al., 2023; Dahl et al., 2024, inter alia ), as well as a plethora of individual accounts on social and traditional media outlets,…”
Undesirable biases in NLP: Averting a crisis of measurement (2022, background): “Choice of racial categories Racial categories do not generalize across cultural contexts (Ghosh et al., 2021; Malik et al., 2022; Abid et al., 2021).”
Dialog Inpainting: Turning Documents into Dialogs (2022, unknown): “We approached the problem by curating a set of 700 terms from the literature (Bolukbasi et al., 2016; Garg et al., 2018; May et al., 2019; Nadeem et al., 2020; Abid et al., 2021) related to sensitive characteristics—such as race, ethnicity, gender, and sexual orientation.”
Shifting machine learning for healthcare from development to deployment and from models to data (2022, background): “Training models on multi-institutional datasets can be most effective at combating model deterioration 15 , and directly combating existing biases in the training data can also mitigate their impact 171 .”
The ethics of ChatGPT in medicine and healthcare: a systematic review on Large Language Models (LLMs) (2024, background): “…the propensity of LLMs to disseminate inadequate information, input of sensitive health information or patient data, raising concerns regarding privacy, 22 and perpetuating harmful gender, cultural or racial biases, 25–28 well known from machine learning algorithms 29 especially in healthcare.”