 Hello, I'm Francesco Bailo, and I'm going to present this very preliminary version of our presentation with some initial findings. We hope to have more as we will actually participate live in the next two weeks. So we are investigating the ideology of the collective construction of the knowledge of unfolding events on Wikipedia. So this is basically our setting and what we are aspiring to do. So ideology broadly intended as somebody's system of beliefs is critical in understanding how we interpret facts. As we are presented with new facts, we naturally fall back on our ideology to understand and make sense of them. So political ideology in particular help us interpret or correct social arrangement through the lenses of our normative ideas or what this arrangement should be. So we what we will plan to do is to use a small sample of editors that have Wikipedia editors that have self-disclosed ideological labels to estimate the ideological alignments of all active editors on Wikipedia, active on articles that have basically appeared during 2020 on the current events portal on the English version of the site. So this is the data that we are using. We're using the Wikipedia dump, the XML dump of all user pages, so namespace 2, and we collected the user boxes. So we went through all the dumps of all the pages, sorry, and we collected all the user boxes that the user indicated in their page. So user boxes present information about interest, general interest of the user, but then of course can also be about political ideology in a subsection of users have a political ideology aligned user box, so identifying their ideology. So of the 400,000 editors that include at least about 400,000 editors include at least one user box on their personal page, so user page, and in total there are lots of user boxes, different user boxes, because of course everybody can create one, so there are more than a family in user boxes in total where identified. Of these of course, very limited numbers are political user boxes. So we have identified by scraping from the portal page 124 events, and of this 96 we have found that there are revisions for 96 of those. In total again, we have 10 96 event pages and 91,000 events, sorry revisions for each, for in total for all these pages. We can immediately see that these revisions are distributed, but also highly concentrated, because of course the COVID-19 event was one of those, so we have a lot of revisions that are concentrated around that page. We started mapping events and user interaction, so the correlation that the relationship between users and editors and the pages that we that we have collected. But again, let's dig into the methods that we have been using. So we have extracted 606 political ideology boxes about, user boxes about political ideologies, and we use this, and we know that they are about political ideologies because they all belong to the category of boxes, so the project page, user boxes, politics, and ideologies. Then we associated 59 of these user boxes to 15 ideological labels that are as proposed by Hermann and Doring in a political analysis paper, and this is of course important because this association allows us to assign to each one of these 59 user boxes a score on a political ideology score using the information provided in the paper. So this is the, again, the network linking user boxes. So the 59 user boxes to the 15 ideologies, the Green Politics in Blue, Green Politics, Liberalism, Marxism, Social Liberalism, Conservatism, and the others, and of course we have many more boxes, user boxes that are linked to this. And this is the network that links 2017 editors to one of these 15, at least one of these 15 ideologies, okay? So again, users are linked, editors are linked to user boxes and user boxes are associated to an ideology, so we can construct the relationship linking users to ideology, and by doing this, of course, we are able to associate users with a score. So this is how the 15 route ideologies actually score on a left-to-right ideological score, so the most left-wing being communism and the most right-wing being Social Conservatism, and most of the ideology being concentrated around zero. That again is the threshold that divides right and left. And when we associate these 2014 users to an ideology, so the numbers actually slightly varies. Again, we're still in the preliminary phase because sometimes when we use the username or the user ID, we get different matchings because we assume that a few users have been changing user ID, user names over time, so probably we should fall back on the using user IDs. So this is again how the 2014 users actually are scored on the left-to-right ideological score. So it's quite balanced, although again the left-wing is more extreme according to the ideological scoring proposed by the paper that we are using, although most of the users, of course, are around zeroes, and this is also how we proceed with the classification, so that we plan to do for the remaining of the users, so also users that don't have a direct connection with an ideology. So we do it in a binary way, we classify binaries, so either left or right, and again the threshold being zero. So every user that is as a score below zero is intended as left-wing leaning, and any user that is scored above zero is intended as right-wing leaning. So what we, the classification in practice works using the network of interest, so the fact that, although a small minority of users actually used, actually are indicating a user box that pertains to a specific ideology, a much larger number of users are using user boxes to indicate their interest. So we are using this much broader network that involves a much wider sample of users to try to classify them. So we are using, we are training our classifier on the subset of users that are linked to political ideology directly through the user boxes, but they also express other interests through user boxes, and using this information to see whether we can classify the remaining of the users, so their users did not express directly a user box or an ideology, and see whether this is possible. And so far, preliminary results have indicated that, indeed, it's complicated to do this, but it's still effective. So, so far, our classification has been able to, on the 400,000 users that are not connected to a specific ideology, but again, this is a preliminary result that we hope to polish and to improve. We have been able to identify of this about 4,000 users as right-wing. So, again, we're using the binary distinction. And we are working again on trying to see whether it's possible to improve the classification here to have a much more nuanced classification, probably that, again, instead of a specifically binary classification.