The team from the Slovak Kempelen Institute on Intelligence Technologies received an award for the best contribution to the annual ACM Conference on Recommender Systems (RecSys), held in Amsterdam.
In their study, the researchers focused on how recommendation systems work on YouTube and what impact they have on spreading disinformation. They also tried to find out whether misinformation filter bubbles are created on the channel, and whether people can get out if they watch videos that disprove hoaxes, the Denník N daily reported.
What researchers found out
The researchers launched bots that watched 3,951 videos, and recorded 78,763 recommendations for watching other videos. They carried out 10,075 searches, gathering 201,404 results. They subsequently processed 17,405 unique videos from 6,342 channels.
They found out that YouTube had not significantly improved its fight against disinformation, as the situation has been complicated by the pandemic and hoaxes on vaccination.
Moreover, misinformation bubbles exist in the lists of recommended videos, but not in the list of attained results.
At the same time, watching videos that debunk other videos reduces the amount of disinformation appearing in the search results along with recommendations on the channel.
The institute now wants to create tools to measure the impact of algorithms on the spread of disinformation, Denník N reported.
7. Oct 2021 at 11:51 | Compiled by Spectator staff