Could a Conversational AI Identify Offensive Language?

Daniela America Da Silva,Paulo Marcelo Tasinaffo,Luiz Alberto Vieira Dias,Gildarcio Sousa Goncalves,Johnny Cardoso Marques,Henrique Duarte Borges Louro,Adilson Marques Da Cunha

doi:10.3390/info12100418

Daniela America Da Silva, Paulo Marcelo Tasinaffo + Show 5 more

Open Access

https://doi.org/10.3390/info12100418

Copy DOI

Journal: Information	Publication Date: Oct 12, 2021
Citations: 7	License type: CC BY 4.0

Affiliation: Instituto Tecnológico de Aeronáutica

Abstract

In recent years, we have seen a wide use of Artificial Intelligence (AI) applications in the Internet and everywhere. Natural Language Processing and Machine Learning are important sub-fields of AI that have made Chatbots and Conversational AI applications possible. Those algorithms are built based on historical data in order to create language models, however historical data could be intrinsically discriminatory. This article investigates whether a Conversational AI could identify offensive language and it will show how large language models often produce quite a bit of unethical behavior because of bias in the historical data. Our low-level proof-of-concept will present the challenges to detect offensive language in social media and it will discuss some steps to propitiate strong results in the detection of offensive language and unethical behavior using a Conversational AI.

Highlights

Algorithms with Artificial Intelligence (AI) origins are currently used for many tasks in a variety of domains
In the Results section, we describe the application of the Dictionary to a text corpus to identify offensive language, applying the dictionary to Twitter posts
This paper focused on investigating if Conversational AI (NLP, Chatbots) could identify offensive language and/or unethical behaviors and the readiness of current foundation algorithms

Summary

Introduction

Algorithms with Artificial Intelligence (AI) origins are currently used for many tasks in a variety of domains. The general objective of this work is to raise questions about whether and how Chatbots and Conversational AI can handle unethical user behavior. It will present a low-level proof-of-concept for detecting offensive language in social media. The systematic mapping process comprises three steps: (1) the identification of relevant literature, (2) the composition of a classification scheme, and (3) the [5] literature mapping. In this way, the following mapping questions were defined: 1.

Objectives

Methods

Results

Discussion

Conclusion