Towards Ethical NLP: On Class Disparities and Risks of Dual Use

Recent developments in natural language processing (NLP) technologies have afforded great advances in terms of performance, adaptability, and expanding NLP technologies to new tasks.
However, as these technologies become more readily integrated into our societies, it is necessary to ask how they can become harmful; to whom they are harmful; and how to address risks of well-intentioned work being reused to cause harm.
In this talk, I will present past and present research dedicated to identifying ethical concerns of NLP technologies and directions towards ethical research practices.
First, I will present documenting disparities in research on biases in NLP; then show how contemporary language technologies perform disparately depending on the socio-economic-based language variation of speakers.
Finally, I will present a survey on the risks of malicious reuse of NLP tools, and initial steps towards addressing such risks as a field.

Post Talk Link:  Click Here

Passcode: 0T+=ynU!