Conference Content

Learn more about who, what, when, and where.
2 of 4

In & Out of the Hub

This is your portal to access more content in the Virtual Hub.
1 of 4

Live Content

On these streams you will find live content happening during the conference.
4 of 4

Connect with others

These spaces are built specifically for you to connect with others at the conference.
3 of 4
← Back to Agenda
Add to Calendar 3/10/21 20:00 3/10/21 21:45 UTC Paper Session 26 Check out this session on the FAccT Hub.
Track Two

Paper Session 26

Session Chair:
Kristian Lum
No items found.


Ask a Question?
Join the Conversation

Censorship of Online Encyclopedias: Implications for NLP Models

Eddie Yang, Margaret E. Roberts
View Paper


While artificial intelligence provides the backbone for many tools people use around the world, recent work has brought to attention that the algorithms powering AI are not free of politics, stereotypes, and bias. While most work in this area has focused on the ways in which AI can exacerbate existing inequalities and discrimination, very little work has studied how governments actively shape training data. We describe how censorship has affected the development of Wikipedia corpuses, text data which are regularly used for pre-trained inputs into NLP algorithms. We show that word embeddings trained on Baidu Baike, an online Chinese encyclopedia, have very different associations between adjectives and a range of concepts about democracy, freedom, collective action, and people and historical events in China than its regularly blocked but uncensored counterpart – Chinese language Wikipedia. We examine the implications of these discrepancies by studying their use in downstream AI applications. Our paper shows how government repression, censorship, and self-censorship may impact training data and the applications that draw from them.

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, Shmargaret Shmitchell
View Paper


The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models.

BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation

Jwala Dhamala, Tony Sun, Varun Kumar, Satyapriya Krishna, Yada Pruksachatkun, Kai-Wei Chang, Rahul Gupta
View Paper


Recent advances in deep learning techniques have enabled machines to generate cohesive open-ended text when prompted with a sequence of words as context. While these models now empower many downstream applications from conversation bots to automatic storytelling, they have been shown to generate texts that exhibit social biases. To systematically study and benchmark social biases in open-ended language generation, we introduce the Bias in Open-Ended Language Generation Dataset (BOLD), a large-scale dataset that consists of 23,679 English text generation prompts for bias benchmarking across five domains: profession, gender, race, religion, and political ideology. We also propose new automated metrics for toxicity, psycholinguistic norms, and text gender polarity to measure social biases in open-ended text generation from multiple angles. An examination of text generated from three popular language models reveals that the majority of these models exhibit a larger social bias than human-written Wikipedia text across all domains. With these results we highlight the need to benchmark biases in open-ended language generation and caution users of language generation models on downstream tasks to be cognizant of these embedded prejudices.

Live Q&A Recording

This live session has not been uploaded yet. Check back soon or check out the live session.