Loading...

Media is loading
 

Paper Type

ERF

Abstract

Virtually every organization uses writing in one form or another to communicate with its internal and external stakeholders and with the public at large. Although ideally all of an organization’s written communications would be evaluated for political bias and credibility prior to public release, a notable constraint currently limits the feasibility of such activities. Specifically, the volume of written communications produced by many organizations is so large that accurately screening all of those communications for credibility and political bias would be a prohibitively time-consuming, expensive, and labor-intensive enterprise. In light of this constraint, the current project seeks to evaluate whether and to what extent a deep neural network can learn to automatically identify and accurately quantify political bias and credibility in written documents. The preliminary results reported in this emergent research forum paper are highly promising, and indicate that a deep neural network can achieve human-level performance in rating written documents for their degrees of credibility and political bias. The implications of these findings for companies and researchers are discussed, and next steps for the project are presented.

Share

COinS
 
Aug 10th, 12:00 AM

Detecting and Quantifying Political Bias and Credibility in Written Communications using an LSTM-Based Deep Neural Network

Virtually every organization uses writing in one form or another to communicate with its internal and external stakeholders and with the public at large. Although ideally all of an organization’s written communications would be evaluated for political bias and credibility prior to public release, a notable constraint currently limits the feasibility of such activities. Specifically, the volume of written communications produced by many organizations is so large that accurately screening all of those communications for credibility and political bias would be a prohibitively time-consuming, expensive, and labor-intensive enterprise. In light of this constraint, the current project seeks to evaluate whether and to what extent a deep neural network can learn to automatically identify and accurately quantify political bias and credibility in written documents. The preliminary results reported in this emergent research forum paper are highly promising, and indicate that a deep neural network can achieve human-level performance in rating written documents for their degrees of credibility and political bias. The implications of these findings for companies and researchers are discussed, and next steps for the project are presented.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.