Machine Learning
Plagiarism Detection Tools Offer a False Sense of Accuracy
The tools that likely brought down Harvard president Claudine Gay are improperly used on students all the time
Soo Oh
Investigative Editor
Soo Oh is an editor and data journalist. Before The Markup, she was the data editor at Reveal from the Center for Investigative Reporting. She has reported stories, analyzed data, coded interactive visuals, and built internal tools at the Center for Investigative Reporting, The Wall Street Journal, Vox.com, the Los Angeles Times, and The Chronicle of Higher Education. In 2018, she was a John S. Knight Fellow at Stanford University, where she researched how to better manage and support journalists with technical skills. She currently teaches students how to code as a lecturer at UC Berkeley’s School of Journalism.
Soo uses she/her pronouns.
LevelUp
Without a Trace: How to Keep Your Phone Off the Grid
We answer the questions readers asked in response to our guide to anonymizing your phone
Privacy
Your Smart TV Knows What You’re Watching
Here’s how to turn off “automated content recognition,” the Shazam-like software on smart TVs that tracks what you’re watching
How Certain Algorithms to Improve the Human Condition Have Failed
18th-century German forest management explains why government websites make you want to pull your hair out
Still Loading
What It’s Like Living With Limited Access to Internet in the Black Rural South
Across the rural South, about 38% of Black households don’t have home internet, a higher percentage than White people in the same region and the national average
Still Loading
FCC Approves Rules to Prevent ‘Digital Discrimination’
Congress tasked the FCC with closing the digital divide. Here’s what the agency’s doing
News
The Problems Biden’s AI Order Must Address
The Markup gives a section-by-section breakdown of the summarized executive order on artificial intelligence
Hello World
How to Use Reported Crime Data to Actually Prevent Crime
Ask why crimes happen where they do
Show Your WorkPrediction: Bias
How We Assessed the Accuracy of Predictive Policing Software
We found that Geolitica’s crime prediction algorithm had a success rate of less than 1% in Plainfield, New Jersey
Prediction: Bias
Predictive Policing Software Terrible At Predicting Crimes
A software company sold a New Jersey police department an algorithm that was right less than 1% of the time
Page 2 of 3