Uncovering Gender Bias within Journalist-Politician Interaction: Analysis Methodology

17 May 2024

This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license.


(1) Brisha Jain, Independent researcher India and brishajain02@gmail.com;

(2) Mainack Mondal, IIT Kharagpur India and mainack@cse.iitkgp.ac.in.


In this work, we primarily performed quantitative analysis to uncover general bias in Indian journalist-politician interaction. Specifically, we performed frequency analysis using statistical testing, emotion analysis, and topic analysis of the collected tweets.

Statistical analysis of interaction frequency and popularity: To analyze gender bias in the frequency of interaction between Journalists and politicians, we examine the popularity of tweets between journalists and politicians across genders of politicians. If journalists do indeed promote gender bias, we expected to see more tweets by male journalists mentioning male politicians compared to mentioning female politicians. We expect female journalists to tweet less or as much about male politicians as they do about female politicians. To check for gender bias in the traction that tweets from politicians receive from journalists, we also examine the popularity of tweets (via retweets, replies and likes for these tweets). We use the Kruskal-Wallis H-Test to determine if significant differences exist in different categories of tweets. Furthermore, we employ pairwise Mann-Whitney U-test to analyze the differences in popularity across our four categories in more granular detail.

Emotion analysis: To examine gender bias in the content to the tweets addressed to politicians by journalists, we employ emotion analysis of tweets (using TweetNLP, a cutting-edge large language model-based Twitter-specific multilingual emotion detection tool [6]) from “MJ-MP”, “MJ-FP”, “FJ-MP” and “FJ-FP” categories. Specifically, we seek to test for differences in these tweets for expressions of “Anger”, “Joy”, “Optimism” and “Sadness”. We use a KruskalWallis H test to determine if significant differences exist in the emotion scores of tweets (along the four dimensions) from our categories.