Periodic Reporting for period 4 - EXPO (Citizens exposed to dissimilar views in the media: investigating backfire effects)
Okres sprawozdawczy: 2023-03-01 do 2023-08-31
Exposure to cross-cutting content is hoped to foster tolerance and minimize polarization. However, evidence also suggests that such exposure could increase extremity and political conflict. ERC EXPO shed light on pressing societal questions relevant across disciplines, such as do citizens encounter dissimilar political perspectives in the current media environment; does exposure to cross-cutting views minimize extremity and polarization, or does it backfire; how can polarization be minimized, among others. In the first par, EXPO relied on latest techniques in computational social science, pairing overtime panel surveys with online behavioral data from large samples of citizens in Poland, the Netherlands, and USA (N = 7,266, over 106M visits), and state-of-the art automated classifications of online content seen by those citizens. We also fielded several experiments. Among other results, reported across 15+ papers, we show that (a) actual over time exposure to like-minded and cross-cutting news websites has no effects on attitude and affective polarization, (b) encouraging partisans to consume cross-cutting media for twelve days does not influence polarization, (c) individuals are not open to cross-cutting information no matter its source (AI, a human or an AI-assisted human), and (d) making people feel happy does not minimize attitude or affective polarization, among other political attitudes. In a key publication, we explain these effects and offer more nuance into people’s online information diets. We find that news consumption is very infrequent: just 3.4% of participants’ online browsing comprised visits to news sites and only between 14% (NL) and 36% (US) of these visits were to news about politics. In other words, news and politics is a small part of individual online information consumption and --- as such --- cross-cutting political information represents an even smaller drop in this ocean of content. Therefore, EXPO also fielded three innovative studies (a field experiment on Twitter, an over-time experiment using a browser extension for YouTube, and an experiment using a mock social media platform) to incentivize citizen exposure to quality and diverse news. We find that computational nudges to social media users (in the Twitter experiment) and the algorithm (in the YouTube experiment) can encourage both greater news consumption and also more diverse and cross-cutting exposure to political content. In sum, many scholars attribute the growing populism, polarization, and wavering support for democratic norms to “echo chambers,” “filter bubbles,” misinformation, and rabbit holes of radicalization. ERC EXPO suggests that the problem is less that people consume “bad” political content (radical, unverified, or otherwise problematic), but that most do not consume any at all.
ERC EXPO project entailed the collection of original data using cutting-edge methodologies and tools, all of which were made open source. We fielded 3 panel surveys, one every 3 months, in 3 countries among over 7,000 individuals. These over time survey data on participants’ attitudes, perceptions, cognitions, and behaviors were combined with their online web browsing data and with automated content-level classifications. This entailed developing questionnaires, translations, training coders toward developing open-source automated classifications, compiling country-specific lists of news media sources, managing large databases, among other tasks. In addition, we fielded online experiments using quality samples and innovative methodological approaches to examine the factors that minimize polarization and increase people’s openness to cross-cutting views, the effects of exposure to partisan media from the other side, the effects of increased or decreased consumption of news media, among others. Furthermore, we also tested innovative solutions to incentivizing exposure to quality, diverse, and verified news. In particular, we fielded an over-time field experiment on Twitter, relying on NLP-trained chat bots that interacted with 28,000 Twitter users to encourage them to follow news accounts. We also fielded a month-long experiment on regular YouTube users who installed a browser extension that nudged the algorithm versus the users toward incentivizing greater recommendations and exposure to quality news. We also developed an interactive mock platform, where pop-up messages encouraged people to follow political and news accounts. This work was widely exploited and disseminated. Scientifically, we have over 15 well-cited publications and 10+ papers under review, in the revise and resubmit process, or finalized for submission. These activities have led to important technological achievements and have advanced state of the art when it comes to online experimentation, computational approaches for data collection and analysis, publicly available categorizations of the ideological leaning of news sources and of automated classifications for online content. These activities also contributed to open and replicable social science (pre-registrations, making all tools, data, and code public offering trainings on open science). Lastly, ERC EXPO team has given presentations at conferences, workshops, and as invited keynotes internationally.
ERC EXPO project has several achievements that advance the state of the art and exceed the proposed objectives. Scientifically, the finding of very limited news exposure online advances the state of the art and social sciences. The finding that online exposure to partisan news and untrustworthy websites has no effects on polarization is an important advancement. In contrast to the fears that partisan media and the online environment lead to extremity and intergroup hostility because people consume politically like-minded information and inhabit “echo chambers”, we find that most citizens are not consuming politics online. These findings put the fears about polarization and echo chambers in perspective, and offered new directions for research on how to encourage citizens to consume quality, verified, and diverse news sources. Methodologically, our reliance on transparent tool to collect browsing histories is an important advancement in research. Most tools are proprietary black boxes. In contrast, Web Historian, which we used, adapted, translated, and popularized, is an open-source software that allows for transparent data sharing, facilitates replication and validation. Furthermore, the open source browser extension for YouTube goes beyond the state of the art scientifically and methodologically: we show that nudging the algorithm can increase recommendations and exposure to quality and diverse sources without decreasing user engagement on the platform.