No, Twitter isn’t coded in favor of conservatives

Personal feeds are created by algorithms, differing for each person. So each person actually trains his or her own feed by interacting, or not, with the content that is proposed. At which point the finding of bias in algorithms should be obvious.

The media lean further to the left than the country as a whole. Again, this is not a surprise. Urban areas lean left, the college-educated lean left, the young lean left. The media are largely the product of college-educated young people in urban areas. They’re going to be further to the left than the median of the country. Quite a bit further to the left, that is.

The social media sites, and this will be as true of Facebook and all the others as it is of Twitter, trend more with the general population. To a large extent, they represent the general population in the raw. So the training of each recommendation algorithm by each user will trend further to the right than the average media position.

Thus, this prominence of right-wing ideas is nothing to do with any positive right-wing bias in the coding. It’s the system of measurement itself that causes the observation. In the Twitter paper, the comparison is between how left-wing and right-wing sources gain amplification. But our definitions of what is the “Right” and the “Left” are different. The media are a few steps to the left of the country, while the population on Twitter is more representative of a few steps to the right of the population. Given that different perception, the Right will gain more amplification from the personal tastes of the system users.

Remember, these algorithms try to serve up what the users desire.

This research will, undoubtedly, bring forth cries that the algorithms must be altered to wipe out this dreadful bias — which is entirely the wrong reaction. What is actually being said here is that the media should move to the right to match the interests and desires of the readership.