Case Study: Deontological Ethics in NLP

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3784–3798 June 6–11, 2021. ©2021 Association for Computational Linguistics 3784 Case Study: Deontological Ethics in NLP Shrimai Prabhumoye ∗, Brendon Boldt ∗, Ruslan Salakhutdinov, Alan W Black School of Computer Science Carnegie Mellon University Pittsburgh, PA, USA {sprabhum, bboldt, rsalakhu, awb}@cs.cmu.edu Abstract Recent work in natural language processing (NLP) has focused on ethical challenges such as understanding and mitigating bias in data and algorithms; identifying objectionable con- tent like hate speech, stereotypes and offen- sive language; and building frameworks for better system design and data handling prac- tices. However, there has been little discus- sion about the ethical foundations that underlie these efforts. In this work, we study one ethi- cal theory, namely deontological ethics, from the perspective of NLP. In particular, we focus on the generalization principle and the respect for autonomy through informed consent. We provide four case studies to demonstrate how these principles can be used with NLP systems. We also recommend directions to avoid the eth- ical issues in these systems. 1 Introduction The 21st century is witnessing a major shift in the way people interact with technology, and nat- ural language processing (NLP) is playing a cen- tral role. A plethora of NLP applications such as question-answering systems (Bouziane et al., 2015; Gillard et al., 2006; Yang et al., 2018) used in diverse fields like healthcare (Sarrouti and Ouatik El Alaoui, 2017; Zweigenbaum, 2009), ed- ucation (Godea and Nielsen, 2018; Raamadhurai et al., 2019), privacy (Ravichander et al., 2019; Shvartzshanider et al., 2018); machine translation systems (Cherry et al., 2019; Barrault et al., 2019; Nakazawa et al., 2019; Liu, 2018), conversational agents (Pietquin et al., 2020; Serban et al., 2018; Liu et al., 2016), recommendation systems (Al- harthi and Inkpen, 2019; Greenquist et al., 2019) etc. are deployed and used by millions of users. NLP systems have become pervasive in current hu- man lifestyle by performing mundane tasks like setting reminders and alarms to complex tasks like ∗ authors contributed equally to this work. replying to emails, booking tickets and recommend- ing movies/restaurants. This widespread use calls for an analysis of these systems from an ethical standpoint. Despite all the advances in efficiency and opera- tions of NLP systems, little literature exists which broadly addresses the ethical challenges of these technologies. Ethical theories have been studied for millennia and should be leveraged in a principled way to address the questions we are facing in NLP today. Instead, the topic of “ethics” within NLP has come to refer primarily to addressing bias in NLP systems; Blodgett et al. (2020) provides a crit- ical survey of how bias is studied in NLP literature. The survey finds that research on NLP systems con- ceptualize bias differently and that the techniques are not well tied with the relevant literature outside of NLP. This creates a gap between NLP research and the study of ethics in philosophy which leaves a …

Content truncated to 500 words