Research roundup

The Politics of Synonyms

Word choice is critical in politics and punditry, but can we identify a political voice solely based on their word choice? This was the question examined in a study titled "Can we detect conditioned variation in political speech? Two kinds of discussion and types of conversation" conducted by Professors Daniel Oppenheimer and Simon DeDeo, along with Ph.D. student Sabina Sloman, all from the Department of Social and Decision Sciences. The study identified words commonly used in Republican and Democratic congressional records and presidential debate documents. Participants were then presented with these words out of context and asked to associate them with a political party. They found that participants were "more likely to associate 'Republican language' with Republicans."

Read more here:

AI May Mistake Chess Discussions as Racist Talk

After a chess YouTube account was taken down after being flagged as "harmful and dangerous," researchers Ashiqur R. KhudaBukhsh and Rupak Sarkar from Carnegie Mellon's Language Technologies Institute wondered if YouTube's hate speech classifiers may have flagged discussions that mention chess piece colors, which come in black and white, as hate speech. To test their hypothesis, they collected 680,000 comments from chess YouTube channels and ran two speech classifiers on them. They then randomly selected 1,000 comments flagged as having hate speech, finding that 82 percent of the sample did not have hate speech. The researchers speculated that the data used to train the hate speech algorithms most likely did not contain chess-related discussions.

Read more here:

Developing a Lander Vision System for Mars Rover

When NASA's Perseverance rover landed on Mars this past Thursday, it was in part due to Andrew Johnson, an alumnus who earned his Ph.D. in Carnegie Mellon University's Robotics Institute. As part of NASA's Jet Propulsion Laboratory, Johnson worked on Perseverance's lander vision system, a crucial system that ensured a safe landing for Perseverance. The system he helped develop takes pictures of the Mars terrain and processes them to determine whether Perseverance will encounter any landing hazards; that information can then be used to steer the rover out of harm's way.

Read more here: