Checkout our new preprint 'Neutral Bots Reveal Political Bias on Social Media'

Collaboration with Wen Chen, @diogofpacheco @OSoMe_IU .

Link: https://arxiv.org/abs/2005.08141 

1/8

We build bots, called drifters, that can mimic human behaviors to probe the biases in Twitter information ecosystem. We have them follow news sources that span the US political spectrum (left, center-left, center, center-right, right) and leave them in the wild.

2/n

Here are what we find:

The initial friends have a major impact on the trajectories of the drifters.

It appears that partisan drifters, especially conservative ones, tend to receive more followers.

3/n

We check the automated activities of the friends and followers of the drifters and find that partisan drifters, especially conservative ones, tend to follow more bots.

4/n

By analyzing the social networks of the drifters, we show that partisan drifters, especially conservative ones, tend to appear in echo chambers.

5/n

Our results also show that drifters initialized with right-leaning friends receive significantly more low-credibility content in their social feeds

6/n

We measure the political valence of the content the drifters receive in their home timeline (s_h), the content they generate (s_u), the content their friends generate (s_f). We use s_h - s_f to measure the algorithmic bias of the news feed ranking algorithm.

7/n

The liberal drifters receive moderate content shifting them toward the center; the interactions of conservative accounts are skewed right. We find weak evidence of liberal bias in the news feed ranking algorithm for conservative accounts.

Detail and discussion in the paper.

8/n