The study was explicitly about social media, as distinguished from in person meetings. The results are interesting, but need to be understood based on the qualifications.
http://www.pnas.org/content/early/2018/08/27/1804840115.full
Structure, Republicans agree to follow a liberal bot, which would retweet from a pool of liberal thought leaders/elites.
Democrats agrees to follow a conservative bot.
Result: After a month of exposure, Republicans exposed to liberals became much more conservative. Liberal exposed to conservatives were slightly more liberal, but the effect was much smaller.
http://www.pnas.org/content/early/2018/08/27/1804840115.full
Structure, Republicans agree to follow a liberal bot, which would retweet from a pool of liberal thought leaders/elites.
Democrats agrees to follow a conservative bot.
Result: After a month of exposure, Republicans exposed to liberals became much more conservative. Liberal exposed to conservatives were slightly more liberal, but the effect was much smaller.