Figure 1 - available via license: Creative Commons Attribution-NonCommercial 4.0 International
Content may be subject to copyright.
Source publication
We present a methodological workflow using two open science tools that we developed. The first, FBAdLibrian, collects images from the Facebook Ad Library. The second, Pykognition, simplifies facial and emotion detection in images using computer vision. We provide a methodological workflow for using these tools and apply them to a case study of the...
Context in source publication
Context 1
... purpose of our workflow is to deduplicate images, classify emotions in core images, and then hydrate those annotations back into the broader dataset of duplicates returned by the Ad Library API. Figure 1 graphically depicts our workflow, and we detail each step in the following sections. ...
Similar publications
Open science practices have been widely discussed and have been implemented with varying success in different disciplines. We argue that computational-x disciplines such as computational social science, are also susceptible to the symptoms of the crises, but in terms of reproducibility. We expand the binary definition of reproducibility into a tier...
Citations
... To address this limitation of Radlibrary, [6] developed FBAdLibrarian, a Command-Line tool that helps researchers retrieve images from the Meta Ad Library. With a valid Meta access token, FBAdLibrarian takes the output from the Meta Ad Library API, reconstructs the ad hyperlinks, opens them, and stores the images in an output folder. ...
... While FBAdLibrarian can download images, it has significant compatibility and maintenance issues, including frequent errors and version mismatches, which prevented reliable performance test and comparison with AdDownloader. Table 1 shows FBAdLibrarian lacks 'Analysis' functionalities, but it is worth mentioning that the authors also built Pykognition for facial and emotion detection in images of political candidates [6]. Since Pykognition was designed for the specific research questions posed by the authors in their research, and is not part of FBAdLibrarian, it is not included as an 'Analytical' feature of the tool. ...
AdDownloader is a Python package for downloading advertisements and their media content from the Meta Online Ad Library. With a valid Meta developer access token, AdDownloader automates the process of downloading relevant ads data and storing it in a user-friendly format. Additionally, AdDownloader uses individual ad links from the downloaded data to access each ad's media content (i.e. images and videos) and stores it locally. The package also offers various analytical functionalities, such as topic modelling of ad text and image captioning using AI, embedded in a Dashboard. AdDownloader can be run as a Command-Line Interface or imported as a Python package, providing a flexible and intuitive user experience. Applications range from understanding the effectiveness and transparency of online political campaigns to monitoring the exposure of different population groups to the marketing of harmful substances.
... To address this limitation of Radlibrary, [6] developed FBAdLibrarian, a Command-Line tool that 48 helps researchers retrieve images from the Meta Ad Library. With a valid Meta access token, 49 ...
... for manual download. Table 1 52 shows FBAdLibrarian lacks 'Analysis' functionalities, but it is worth mentioning that the authors 53 also built Pykognition for facial and emotion detection in images of political candidates [6]. Since 54 ...
AdDownloader is a Python package for downloading advertisements and their media content from the Meta Online Ad Library. With a valid Meta developer access token, AdDownloader automates the process of downloading relevant ads data and storing it in a user-friendly format. Additionally, AdDownloader uses individual ad links from the downloaded data to access each ad's media content (i.e. images and videos) and stores it locally. The package also offers various analytical functionalities, such as topic modelling of ad text and image captioning using AI, embedded in a Dashboard. AdDownloader can be run as a Command-Line Interface or imported as a Python package, providing a flexible and intuitive user experience. Applications range from understanding the effectiveness and transparency of online political campaigns to monitoring the exposure of different population groups to the marketing of harmful substances.
... We do not study candidate pictures in social media advertising, since currently there are questions regarding the completeness of data delivered by the Facebook Ad Library API. However, existing research suggests that Facebook ads from candidate accounts aim to positively promote candidates (Fowler et al., 2021) and most often depict them as happy (Schmøkel & Bossetta, 2021). Still, in analyzing only candidate accounts, we may be overlooking other aspects of digital campaigning that issue divisive messages with different emotions. ...
... Schmokel and Bossetta [18] introduced a methodological workflow that utilises two open science tools that they developed: FBAdLibrian, which collects images from the Facebook Ad Library, and Pykognition, which simplifies facial and emotion detection in images using computer vision. Using this workflow, they analysed the 2020 US primary elections and found that only a small fraction of overall ads (<.1%) contained unique images of campaigning candidates. ...
The objective of this study is to utilise machine learning for predicting user engagement with Facebook post images from universities, aiming to provide strategic guidance for their social media content. A comprehensive analysis was performed on a dataset comprising 1,200 images from 24 prominent universities in the United States, the United Kingdom, and Australia. The open-source machine learning software WEKA 3.8.5 was utilised, with three of its classifiers - J48, Sequential Minimal Optimisation (SMO), and Random Forest (RF) - chosen for their proven efficiency with high-dimensional data. These classifiers were trained using a unique mix of features from the Google Cloud Vision API and attributes derived from relevant literature. Cross-validation techniques were employed to assess the accuracy of the models. A robust performance was exhibited by all models: J48 yielded an accuracy of 90.92%, SMO 94.85%, and RF 97.08%. Even under 66/34 and 90/10 data splits, the high accuracy was sustained. These findings underscore the efficacy of machine learning models in predicting user engagement with Facebook post images for leading universities. Such techniques can be leveraged by universities to optimise their Facebook user engagement strategies, thereby informing the creation of content that aligns with their target audience preferences. This study illuminates promising directions for updating social media strategies in response to evolving communication trends.
... candidates most often display happiness and calm in their facial expressions, and they rarely attack opponents in imagebased ads from their official Facebook pages. When candidates do attack, opponents are portrayed with emotions such as anger, sadness, and fear [11]. ...
The mental states of feelings are represented by a person's emotions. At the moment, emotion recognition is an explosive topic. Various tools that are offered by some languages, algorithms, and systems can be used to detect emotion. For emotion detection, almost all detection systems employ baby faces. The accuracy of any emotion detection system at this moment is about 90%, which is due to the fact that emotion identification techniques are still far from being flawless. This study analyses emotion detection by image from the late 2000s to the present. The history, an overview, and a few stages of emotion detection by the image are presented in this study. The system could currently only identify six emotions: joy, sadness, fear, surprise, disgust, and anger.
... Overall confidence in the presented results would surely be higher if we based our analysis on a computer-vision approach which was shown to have higher accuracy, at least on FER2013 data. One of the possible alternatives would be to follow recent work (Bossetta & Schmøkel, 2023;Schmøkel & Bossetta, 2022) and use proprietary API such as Amazon's Rekognition. Nonetheless, the aim of this paper was to introduce an open-source approach focusing on free computing resources and the deployment of machine learning algorithms on systems that are accessible to individual researchers in social sciences, including graduate students without significant institutional support for computational work. ...
Online media has revolutionized the way political information is disseminated and consumed on a global scale, and this shift has compelled political figures to adopt new strategies of capturing and retaining voter attention. These strategies often rely on emotional persuasion and appeal, and as visual content becomes increasingly prevalent in virtual space, much of political communication too has come to be marked by evocative video content and imagery. The present paper offers a novel approach to analyzing material of this kind. We apply a deep-learning-based computer-vision algorithm to a sample of 220 YouTube videos depicting political leaders from 15 different countries, which is based on an existing trained convolutional neural network architecture provided by the Python library fer. The algorithm returns emotion scores representing the relative presence of 6 emotional states (anger, disgust, fear, happiness, sadness, and surprise) and a neutral expression for each frame of the processed YouTube video. We observe statistically significant differences in the average score of expressed negative emotions between groups of leaders with varying degrees of populist rhetoric as defined by the Global Party Survey (GPS), indicating that populist leaders tend to express negative emotions to a greater extent during their public performance than their non-populist counterparts. Overall, our contribution provides insight into the characteristics of visual self-representation among political leaders, as well as an open-source workflow for further computational studies of their non-verbal communication.
... With the advent of computational analysis techniques, researchers have new opportunities to collect more visual data and code the data more efficiently. Recently, researchers have employed automated computer vision techniques when examining politicians' nonverbal behavior in televised debates (Joo et al., 2019) or in social media ads (Schmøkel & Bossetta, 2021) and when examining their visual self-presentation on social media platforms (Peng, 2021). These case studies are valuable and future research should build upon their work to gain deeper insights into the visual representation of politicians. ...
... We do not study candidate pictures in social media advertising, since currently there are questions regarding the completeness of data delivered by the Facebook Ad Library API. However, existing research suggests that Facebook ads from candidate accounts aim to positively promote candidates (Fowler et al., 2021) and most often depict them as happy (Schmøkel & Bossetta, 2021). Still, in analyzing only candidate accounts, we may be overlooking other aspects of digital campaigning that issue divisive messages with different emotions. ...
This study provides a cross-platform, longitudinal investigation of pictures depicting political candidates posted to Facebook and Instagram over a 15-month period during the 2020 US election (n = 4,977). After motivating an exploratory research design, we set out to expound: the extent of cross-platform image posting across Facebook and Instagram; the emotion expression of politicians across the two platforms; and the relationship between these emotions and post performance. Our analysis of eight political campaigns (seven Democratic challengers and the Republican incumbent) finds relatively high and stable levels of cross-posting candidate pictures across the two platforms. The exception is the incumbent campaign, where cross-posting activity rose in proximity to the primary elections. Regarding emotions, we utilize both computer vision and crowd coding to identify happiness as the dominant emotion on Facebook and Instagram. Overall, we detect little variation in candidate emotion expressions – across campaigns and across platforms. However, we do find differences in how platform audiences respond to emotions, proxied here through post performance. Results from binomial logistic regressions show that in comparison with Calm, posts exhibiting Anger are less likely to overperform on both Facebook and Instagram. Most interestingly, we find diverging patterns for Happiness, which performs better than Calm on Instagram but not on Facebook. We interpret these findings to suggest first, that Instagram users reward emotionality from politicians. Second and more importantly, we argue that differing audience responses to emotions – captured through social media metrics – may reveal a generation polarization in what different segments of the electorate prefer their political leaders to be.
... His conclusions suggested a slight difference in terms of topic diversity between microtargeted and non-microtargeted ads, but not a significant difference in terms of negative content. Schmøkel and Bossetta (2022) carried Baviera, T., Sánchez-Junqueras, J. & Rosso, P. ...
Facebook's advertising platform provides political parties with an electoral tool that enables them to reach an extremely detailed audience. Unlike television, the sponsored content on Facebook is seen only by the targeted users. This opacity was an obstacle to political communications research until Facebook released advertiser-sponsored content in 2018. The company's new transparent policy included sharing metadata related to the cost and number of impressions the ads received. This research studies the content sponsored on Facebook by the five main national political parties in Spain during the two General Elections held in 2019. The research corpus consists of 14,684 Facebook ads. An extraction algorithm detected the key terms in the text-based messages conveyed in the ad. The prominence of these topics was estimated from the aggregate number of impressions accumulated by each term. Different content patterns were assessed following three categories: user mobilization, candidate presence, and ideological issues. PSOE and PP positioned themselves more toward calls to action. Podemos had the greater number of issues related to policy among the most salient topics in its advertising. Ciudadanos' strategy focused more on its candidate and mobilization. Vox sponsored few Facebook ads, and they barely included policy issues. Spain was a highly prominent term in all parties' campaigns. Ciudadanos shared the middle ground on the ideological axis: they promoted social issues more aligned with left-wing parties as well as economic topics usually advocated by the right-wing. Overall, our results point to a greater emphasis on candidates rather than issues.
... Edelson et al. [4] use the Facebook Ad Library to find suspicious and malicious advertising practices. Schmøkel and Bossetta [22] provide tools to analyse the images in Facebook ads. ...
Social media platforms have been trying to be more transparent about the political ads they run on their platforms, because the Cambridge Analytica scandal revealed that political campaigns are using social media on a large scale. One such transparency effort is the Facebook Ad Library, a public repository of all political ads run on Facebook and Instagram. This library provides journalist and researchers with data to get a better understanding of political advertising and microtargeting on Facebook's platforms. Unfortunately, the Facebook Ad Library only provides estimates and basic information. This paper analyses political ads in more depth, by examining the themes that ads are about. We provide a method to match themes to political Facebook ads and we apply this method to analyse Facebook ad campaigns ran by Dutch political parties during the 2021 Dutch general election.
