Political Discourse on Social Media: Echo Chambers, Gatekeepers, and the Price of Bipartisanship

Political Discourse on Social Media: Echo Chambers, Gatekeepers, and the Price of Bipartisanship Kiran Garimella Gianmarco De Francisci Morales Aristides Gionis Michael Mathioudakis

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Political Discourse on Social Media: Echo Chambers, Gatekeepers, and the Price of Bipartisanship 2018 The Web Conference (WWW) 1801.01665 garimella2018-political-discourse garimella2018-political-discourse.txt, the project's staged copy 17d685c94d8864aba216889be5e80f6087d06c21 arxiv-open political-science

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them existence Quotation withheld because this record is not cleared for redistribution. political echo chambers on social media Quotation withheld because this record is not cleared for redistribution. observational-study observational-study Quotation withheld because this record is not cleared for redistribution. mechanisms content and network are measured JOINTLY at the level of the individual user, rather than content consumption alone, and the authors state that as the study's distinguishing move Quotation withheld because this record is not cleared for redistribution. mechanisms POLITICAL LEANING IS NOT INFERRED FROM THE TEXT AT ALL: scoring short snippets is called very challenging, so the leaning is inherited from an externally supplied per-domain ground truth of news organisations about this quote: 2026-08-16: THIS QUOTE IS TRUNCATED MID-SENTENCE BECAUSE THE FILE INTERRUPTS IT. The staged text continues "...news organizations with 6https://en.wikipedia.org/... 7http://www.bbc.com/... 8https://www.nytimes.com/..." - three footnote URLs are spliced between 'with' and the rest of the clause, which resumes as 'a presence on social media obtained from Bakshy et al.' The span stops at the last word before the splice. The sentence is not quotable in full from this file without quoting three URLs in the middle of it. Quotation withheld because this record is not cleared for redistribution. mechanisms and that borrowed score has a stated meaning of its own - the FRACTION OF FACEBOOK USERS VISITING A DOMAIN WHO IDENTIFY THEMSELVES AS CONSERVATIVE ON THEIR PROFILE - which the authors read as the domain having a conservative or liberal bent in its coverage Quotation withheld because this record is not cleared for redistribution. mechanisms the measurement is deliberately restricted to what can be scored: only tweets carrying a link to a news organisation with a known and independently derived leaning are counted at all Quotation withheld because this record is not cleared for redistribution. mechanisms CONSUMPTION IS DEFINED THROUGH THE NETWORK: a user's produced content is the tweets they post, and their consumed content is the tweets posted by the accounts they follow Quotation withheld because this record is not cleared for redistribution. mechanisms partisanship is given a PARAMETERISED definition rather than a threshold - a user is delta-partisan when their production polarity lies within delta of either extreme - and the authors state the nesting property that follows, so a claim at one delta implies the claim at every larger one Quotation withheld because this record is not cleared for redistribution. mechanisms VARIANCE IS MEASURED ALONGSIDE THE MEAN over the same set of tweets, with the stated objective of quantifying the RANGE of opinions a user's produced content covers Quotation withheld because this record is not cleared for redistribution. mechanisms FIVE NON-POLITICAL DATASETS ARE COLLECTED AS A BASELINE for the political ones - tbt, ff, gameofthrones, love, foodporn - each built from one hashtag by parsing an archive, filtering to users with at least five tweets, and then taking all of those users' tweets Quotation withheld because this record is not cleared for redistribution. mechanisms automated accounts are excluded by minimum AND maximum thresholds on tweets per day, followers and friends, together with a requirement that the account be at least one year old at collection time Quotation withheld because this record is not cleared for redistribution. mechanisms the paper STIPULATES 'gatekeeper' against the established sense and says so: in communication studies the term names news media sources acting as filters, and here it names an individual user who receives content from both leanings but produces only one Quotation withheld because this record is not cleared for redistribution. political leaning is measured only from tweets that link to news sources of known, independently-derived leaning, not the message content itself what this limits: every political-leaning measurement in the study: WE CONSIDER ONLY MESSAGES THAT CONTAIN A LINK TO AN ONLINE NEWS ORGANIZATION WITH A KNOWN AND INDEPENDENTLY DERIVED POLITICAL [leaning] - users without such links are unmeasured Quotation withheld because this record is not cleared for redistribution. All data come from Twitter and are only a sample of social-media discussion. what this limits: the generalisation to political discourse at large: all data come from Twitter and are only a sample of social-media discussion Quotation withheld because this record is not cleared for redistribution. Results are drawn from politically-savvy users and should not be assumed to generalise to the general public. what this limits: the network measured: the analysis rests on the retweet and follow graph available through the API, not on the full communication structure Quotation withheld because this record is not cleared for redistribution. The polarity measures depend on externally supplied political labels for news sources. what this limits: every polarity score: OUR PRODUCTION AND CONSUMPTION SCORES RELY ON EXTERNAL LABELING OF NEWS SOURCES ALONG A POLITICAL AXIS Quotation withheld because this record is not cleared for redistribution. Following is used as a proxy for actual consumption; users may not read everything their neighbours post. what this limits: every consumption measure: OUR ANALYSIS ASSUMES THAT EACH USER CONSUMES ALL CONTENT PRODUCED BY ALL THEIR NEIGHBORS - following stands in for reading Quotation withheld because this record is not cleared for redistribution. Some articles from the scored news domains may not be political. what this limits: the political-content assumption: NOT ALL NEWS ARTICLES FROM THE NEWS SOURCES WE BASE THE POLARITY MEASURES [on] ARE POLITICAL Quotation withheld because this record is not cleared for redistribution. The causal relations between selective exposure, biased assimilation, algorithmic filtering and echo chambers are not addressed. what this limits: any causal reading of echo chambers: UNDERSTANDING HOW ALL THESE PHENOMENA INTERACT WITH EACH OTHER AND THE PRECISE CAUSALITY RELATIONS IS BEYOND THE SCOPE OF THIS PAPER Quotation withheld because this record is not cleared for redistribution. The literature has no consensus definition of the paper's central term, so the authors chose one. what this limits: the finding that echo chambers are prevalent: the authors state there is no consensus on a formal definition of an echo chamber in the literature and opt for their own, under which an echo chamber exists when the leaning of content a user receives agrees with the leaning of what they share Quotation withheld because this record is not cleared for redistribution. A term is borrowed from another field with its unit of analysis changed, which the authors state. what this limits: any comparison with the gatekeeping literature: in media and communication studies gatekeepers are media sources acting as filters, whereas here they are individual users identified by their joint consumption and production leaning Quotation withheld because this record is not cleared for redistribution. A second definition of the same category was tested and its results not presented. what this limits: the gatekeeper results reported here: the authors also tested a different definition - high consumption variance with low production variance - found the results almost identical, and did not present them explicitly Quotation withheld because this record is not cleared for redistribution. What the identified users are actually doing is left open by the authors. what this limits: any interpretation of the gatekeeper class: the authors say it is not clear from their analysis what such users are doing, so the label marks a position on two measures and not a motive Quotation withheld because this record is not cleared for redistribution. ten-fold cross-validation of the partisan and gatekeeper classification model eliminates: an accuracy figure earned on the same data the classifier was fitted to · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. echo chamber echo chamber Quotation withheld because this record is not cleared for redistribution. gatekeeper gatekeeper Quotation withheld because this record is not cleared for redistribution. media sources media sources Quotation withheld because this record is not cleared for redistribution. leaning leaning Quotation withheld because this record is not cleared for redistribution. consumption variance consumption variance Quotation withheld because this record is not cleared for redistribution. price of bipartisanship price of bipartisanship Quotation withheld because this record is not cleared for redistribution. follow → content consumption THE INDEXING / PROXY RELATION - 'from' STANDS IN FOR 'to'. We have no type for it: 'causes' asserts production, and 'operationalised-as' says the authors CHOSE a measure, whereas a proxy claim says THIS STANDS IN FOR WHAT WE ACTUALLY CARE ABOUT AND WE KNOW IT IS NOT THE SAME THING. Measured at 8% of papers. A measurement substitution stated plainly: 'we use the “follow” relationship as a proxy for the actual content consumption'. The word 'actual' is doing the same work as amodei2016's 'really care about' - the authors marking the distance between what they measured and what they meant. Quotation withheld because this record is not cleared for redistribution. an echo chamber - AS AGREEMENT BETWEEN TWO PER-USER SCORES: it exists, for this paper, when the leaning of the content a user receives agrees with the leaning of what they share; not a group, not a network structure, not an isolation measure, but a correlation computed per user - and the authors adopt it because the literature offers no consensus formal definition to inherit Quotation withheld because this record is not cleared for redistribution. a gatekeeper - AN INDIVIDUAL USER, NOT A MEDIA SOURCE: communication studies gives the word to institutions that filter what passes into a channel, and this paper reassigns it to a person at a particular position on two scores, stating the earlier sense and supplying a bridging argument - that such users block information from one side, and hence can be considered gatekeepers Quotation withheld because this record is not cleared for redistribution. leaning - A STIPULATED SCORE, WITH ITS OWN SYNONYMS NAMED IN THE FOOTNOTE THAT STIPULATES IT: the authors define it as a score quantifying alignment with one political side and list the terms the literature uses for the same thing, so four words - leaning, ideology, polarity, ideological stance - are declared interchangeable by the paper itself Quotation withheld because this record is not cleared for redistribution. extends → the study of political polarization and echo chambers Quotation withheld because this record is not cleared for redistribution. builds-on → the latent space model of Barberá et al. [7] Quotation withheld because this record is not cleared for redistribution. builds-on → the news-source leaning scores and hard/soft news classifier of Bakshy et al. [4] Quotation withheld because this record is not cleared for redistribution. replicates → Barberá [6] on higher polarization of retweet networks for political topics Quotation withheld because this record is not cleared for redistribution. extends → prior echo-chamber studies that measure either content or network alone Quotation withheld because this record is not cleared for redistribution. echo chamber —operationalised-as→ leaning echo chamber leaning strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. gatekeeper —contrasted-with→ media sources gatekeeper media sources strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span defines gatekeepers differently in media studies and in this study; contrasted-with is a loose reading. Quotation withheld because this record is not cleared for redistribution. Twitter users mostly see political opinions matching their own. Quotation withheld because this record is not cleared for redistribution. Engagement is measured as all content that a user clicks. Quotation withheld because this record is not cleared for redistribution. political leaning political leaning Quotation withheld because this record is not cleared for redistribution. echo chamber echo chamber about this quote: QUOTE REPAIRED: the extractor split a hyphen-space and the quote was recorded faithfully with the damage in it, which made it unanchorable. The repaired string is an EXACT SUBSTRING of the staged text - that is why it was accepted, not because a rule produced it. The text as originally recorded is kept in evidence as extracted. Quotation withheld because this record is not cleared for redistribution. production production Quotation withheld because this record is not cleared for redistribution. consumption consumption Quotation withheld because this record is not cleared for redistribution. engagement engagement Quotation withheld because this record is not cleared for redistribution. correlation correlation about this quote: QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. Quotation withheld because this record is not cleared for redistribution. effect-size effect-size Quotation withheld because this record is not cleared for redistribution. p-value p-value about this quote: QUOTE RECOVERED FROM THE PAPER'S OWN CHARACTERS. The recorded quote was LETTERS-ONLY IDENTICAL to this stretch of the staged text but not an exact substring of it - spaces, hyphens, ligature glyphs or mathematical operators differed, so it could never anchor. This is the file's own text between the same two letters: not produced by a rule about the damage, copied out of the file. Round-trip checked (the recovered span is letters-only equal to the original) and length-checked. The text as originally recorded is kept in evidence as extracted. Quotation withheld because this record is not cleared for redistribution. statistical-significance statistical-significance Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://arxiv.org/abs/1801.01665.