Providing transparency about the gender visibility in newspapers
summary
TX Group runs a daily Equality Alert that measures gender visibility in images and text by inferring gender from first names, registered users, or reading behavior, and flags percentages that typically fluctuate between 15–30%. A 10‑year view shows visibility rising from 17% to 24%. Female authors’ articles mention women 33% of the time versus 24% for male authors; foreign affairs is an outlier. Team trains a word2vec 300D on 300k+ internal articles and finds persistent gendered associations (homemaker, handbag vs. doctor, programmer), so the service surfaces concrete biases for newsroom action.
AI-written summary · about
chapters (14)
- 0:00Gender visibility in media
- 1:55Gender representation matters
- 3:20TX Group’s approach
- 6:11Equality Alert system
- 7:15Gender visibility defined
- 8:01Long-term trends
- 10:09Visibility by author’s gender
- 11:16Visibility across sections
- 13:44Impact on readership
- 14:27Detecting gender biases
- 16:32Gender stereotypes in embeddings
- 18:17Journalists vs data science
- 19:41Future directions
- 20:47Q&A
related talks
01h 40mAfternoon Session #1: Science, Pseudo-science, Canon and Belief
18mBuilding a Gendered Dictionary
32mData Science for Feminist Comebacks
20mApplying Computational Analysis to Textual Data from the Wild: a Feminist Perspective
05mEvery voice matters: An analysis of @WeAreRLadies
48mKeynote: Data Science Community and You
19mCreating Global Economic Opportunity with Responsible Data Science
15mThe unexamined bytes are not worth coding