Technological Approaches to Detecting Online Disinformation and Manipulation

TL;DR

This paper reviews AI and machine learning techniques for automatically detecting online disinformation and manipulation, focusing on methods supporting fact-checking, topic identification, and message filtering on social media.

Contribution

It provides a comprehensive overview of technical approaches, datasets, and evaluation methods for automatic disinformation detection using AI and machine learning.

Findings

Most techniques employ AI and machine learning with feature extraction.

Various datasets are used for benchmarking disinformation detection methods.

The paper categorizes tasks related to disinformation and manipulation detection.

Abstract

The move of propaganda and disinformation to the online environment is possible thanks to the fact that within the last decade, digital information channels radically increased in popularity as a news source. The main advantage of such media lies in the speed of information creation and dissemination. This, on the other hand, inevitably adds pressure, accelerating editorial work, fact-checking, and the scrutiny of source credibility. In this chapter, an overview of computer-supported approaches to detecting disinformation and manipulative techniques based on several criteria is presented. We concentrate on the technical aspects of automatic methods which support fact-checking, topic identification, text style analysis, or message filtering on social media channels. Most of the techniques employ artificial intelligence and machine learning with feature extraction combining available information resources. The following text firstly specifies the tasks related to computer detection of manipulation and disinformation spreading. The second section presents concrete methods of solving the tasks of the analysis, and the third sections enlists current verification and benchmarking datasets published and used in this area for evaluation and comparison.