We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling

Abstract: Over the last several years, end-to-end neural conversational agents have vastly improved in their ability to carry a chit-chat conversation with humans. However, these models are often trained on large datasets from the internet, and as a result, may learn undesirable behaviors from this data, such as toxic or otherwise harmful language. Researchers must thus wrestle with the issue of how and when to release these models. In this paper, we survey the problem landscape for safety for end-to-end conversational AI and discuss recent and related work. We highlight tensions between values, potential positive impact and potential harms, and provide a framework for making decisions about whether and how to release these models, following the tenets of value-sensitive design. We additionally provide a suite of tools to enable researchers to make better-informed decisions about training and releasing end-to-end conversational AI models.
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2107.03451 [cs.CL]
  (or arXiv:2107.03451v1 [cs.CL] for this version)

Submission history

From: Emily Dinan [view email]
[v1] Wed, 7 Jul 2021 19:25:57 GMT (1180kb,D)
[v2] Mon, 12 Jul 2021 15:37:52 GMT (632kb,D)
[v3] Fri, 23 Jul 2021 15:05:22 GMT (631kb,D)

Link back to: arXiv, form interface, contact.