Assignment 4: Topics on Social Computing 3

Fake News

What constitutes a Fake News?

Fake news is considered to be a news article purposely written to mislead readers into believing false information. Two important aspects of fake news are intent and authenticity. In short fake news is intentionally and verifiably false information designed to manipulate consumers.

What are the strategies and open problems mentioned in the paper, related with the detection of Fake News in Online Social Media?

First, detection of Fake News on Social Media is difficult due to a number of challenges. Content of News is diverse, and the truth can be distorted in numerous amounts of ways making it difficult to cover all areas of Fake News so to speak. Furthermore, often times news concern very recent events to which confirmation data is not yet available making it difficult to detect untruthful statements as facts. One strategy proposed is the extraction of features concerning news content and social context. Using content features like headlines, sources and linguistics and social context features such as looking at previous posts and the existing network around the news broadcaster as well as determining the credibility and reliability of users in said network all can help in detecting fake news. Further existing approaches are knowledge-based detection, which uses external sources to fact-check claims of news articles. And lastly Style-based detection analyzing the writing style of news to pick up on attempts of deception and news presented in an unobjective way.

Hate Speech

What constitutes Hate Speech?

Hate speech is defined as speech that targets disadvantaged social groups, or other minorities with the intention of harming them by promoting violence and hatred towards them. Though when it comes to Social Media it important to differentiate between hate speech and offensive language, as often times the intent of the latter is far more malevolent.

What strategy did the authors propose to detect Hate Speech?

The first strategy proposed to help detect hate speech online is to categorize language into three different categories. Hate speech, offensive language and neither one of them. For this it is useful to make use of things like a Hate speech lexicon, which is a database containing words and phrases classified as hate speech. Only using this is considered as a Bag-of-words approach. However, this often results in misclassification of offensive language as hate speech. Syntactic features like detecting nouns and verbs that together reveal harmful intent (verb) towards a target (noun) can help with classification as well as Non-linguistic data such as the gender and ethnicity of the author in question.

Algorithmic Bias

What constitutes an Algorithmic Bias?

Algorithmic Bias in the field of Hate Speech detection I understood as detection algorithms not taking differences in dialect into account, therefor falsely labeling language as offensive, or hate speech. Basically, part of the context of the message, such as the speaker´s identity and dialect is ignored making it difficult to classify the content correctly.

Mention some real-life consequences that can derive from the examples of Algorithmic Bias presented in the paper.

Algorithmic Bias in Hate Speech detection can have serious real-life consequences for minorities. This can go as far as actually increasing harm and suppression of disadvantaged social groups instead of doing the opposite. It hinders minority groups in reclaiming offensive words in order to make them less harmful by classifying such attempts as offensive language or hate speech. Further automatically removing content which is falsely classified helps in silencing already marginalized voices.

https://ifi7167socialcomputing.wordpress.com/2020/10/01/assignment-4-topics-on-social-computing-3/