AoIR 2022

Platform-Based Political Advertising: New Approaches for Enhancing Platform Observability

Daniel Angus, Mark Andrejevic, Bronwyn Carlson, Axel Bruns, Abdul Obeid, Philip Mai, and Anatoliy Gruzd

• 3 Nov. 2022 – Paper presented at the AoIR 2022 conference, Dublin

Presentation Slides

Abstract

Introduction

Platform-based political advertising is emerging as a key focal point of modern elections. It has also fast become a new regulatory battleground, as governments and citizenries around the globe grapple with the consequences of this new campaign practice (Kreiss & Mcgregor, 2019). Key concerns relate to the ability to algorithmically target hyper-partisan and false information, and with the lack of transparency from political actors and the palatforms themselves, regarding the advertisements placed, money spent, and ultimately who is consuming these advertisements (Dommett & Power, 2019).

The Cambridge Analytica scandal revealed the susceptibility of platform-based political advertising to exploitation, particularly in relation to the UK’s Brexit referendum and Donald Trump’s 2016 election (Cadwalladr, 2018). In the wake of this scandal – and due in no small part to significant sustained pressure from academics, journalists, civil society groups, and eventually governments – the major platforms have been forced to curtail certain advertising practices, and implement a range of political advertising transparency initiatives. A key feature of these initiatives is the provision of new digital tools, often in the form of transparency ‘dashboards’. These platform-provided dashboards offer basic information as to the ads that are or have recently been run, who is sponsoring these ads, and basic aggregated statistics as to the reach and audience of the ads.

While welcome, the dashboards provided by large platforms like Meta and Google stop short of full transparency (Edelson et al. 2019). As one example, Facebook and Instagram ads are readily targeted to highly specific geolocations (down to individual suburb/postcode level) by ad buyers, however Meta’s dashboards only reveal geolocation data at an entire state level. Interest categories for ads are also heavily abstracted making it difficult to interrogate where political advertising may be engaging in racial, gendered, economic, religious, or other harmful forms of discrimination. More so, we are left to trust that information provided by the platforms is accurate, given there is no independent oversight or verification of this data.

Due to the fundamental limitations of platform-provided transparency tools, researchers are turning to other techniques to provide much needed platform observability (Rieder & Hofmann, 2020). In some cases, this involves the augmentation of transparency dashboards to turn them into more easily searchable ad archives, with additional statistics and information visualisation. In other cases researchers are side stepping platform-provided approaches completely using data donation plugins that enrol platform users to contribute any ads encountered while browsing these platforms into platform-independent archives (ProPublica, 2020).

New Approaches for Advertising Accountability

In this paper we outline a suite of digital methods developed to study platform-based political advertising, and ultimately enhance platform observability and accountability. We focus on a study of political advertising that will be conducted throughout the upcoming Australian Federal Election campaign, due to take place no later than May 2022. The data collection methods that have already been developed and deployed in the lead up to this national election range from data gathering techniques that utilise existing platform-provided ad transparency APIs (PoliDashboard: https://global.polidashboard.com/), through to citizen-science data donation approaches (Australian Ad Observatory: https://www.admscentre.org.au/adobservatory/). Additionally, we have also developed a range of computational methods to support both quantitative and qualitative data analysis.

The first tool, PoliDashboard, was originally developed by the Social Media Lab, Ryerson University, as part of an international election transparency initiative. PoliDashboard has been extended in partnership with members of our research team for the Australian context. The tool interfaces with the existing Meta and Google ad transparency libraries, but offers additional archival, ranking, search, data aggregation, and visualisation abilities (see Figure 1).

Figure 1: Screenshots of the PoliDashboard (credit: Social Media Lab, Ryerson University)

The second tool, the Australian Ad Observatory (Burgess et al., 2021) extends the work of the Ad Observatory developed by researchers at NYU, which itself extends the ProPublica Political Ad Collector (ProPublica, 2020). The Observatory relies on the use of a browser plugin, installed by volunteer members of the Australian public on their personal computers. The browser plugin detects any sponsored posts that participants encounter in their Facebook news feeds during regular use of Facebook through their browser. Once detected by the plugin, ads are anonymously sent to a central server along with additional demographic data from the participants (if they have chosen to provide such information). The browser plugin works without requiring any manual intervention from the participants, who can also review their own personal catalogue of ads encountered at any time via the tool.

In addition to the collection of political ads via the above data collection tools, we have implemented a range of new critical data analytic approaches to assist in analysis of campaign materials (Burgess et al., 2021). These approaches encompass optical character recognition (OCR), logo and object detection, and visual and text-based content analysis support, including topic modelling and machine vision. The approaches serve to assist in the discovery of specific political messaging, analyse campaign spending and reach, and examine visual presentation and communication within ads (see Fig 2).

Figure 2: Screenshot of the Australian Ad Observatory researcher dashboard.

While the approaches are developed as a general and highly versatile toolset for the study of platform-based political advertising materials, a specific focus for our team is on the detection and analysis of false or misleading advertising materials, and to also understand the experiences of Indigenous users of these platforms given existing concerns regarding discriminatory algorithmic advertising practices (Andrejevic et al., 2022). We will highlight specific examples encountered, and other general findings as part of our analysis, in addition to guidance on how these techniques can be adopted in future studies of political and non-political platform-based advertising.

References

Andrejevic, M., Fordyce, R., Li, N., & Trott. V. (2021) Unregulated and segmented dark ads on social media: consumer education and regulatory options. Retrieved from: https://apo.org.au/sites/default/files/resource-files/2021-06/apo-nid313053.pdf

Andrejevic, M., Fordyce, R., Li., N, & Trott, V. in collaboration with Angus, D. & Tan, J. (forthcoming) Ad accountability online: a methodological approach, in Pink, S., M. Berg, D. Lupton, & M. Ruckenstein (forthcoming) Everyday Automation: Experiencing and Anticipating Emerging Technologies. London, NY: Routledge, pp. 213-224

Burgess, Jean, Daniel Angus, Nicholas Carah, Mark Andrejevic, Kiah Hawker, Kelly Lewis, Abdul Karim Obeid et al. "Critical simulation as hybrid digital method for exploring the data operations and vernacular cultures of visual social media platforms." (2021): https://osf.io/preprints/socarxiv/2cwsu/download.

Cadwalladr C. (2018, March 18), ‘I made Steve Bannon’s psychological warfare tool’: meet the data war whistleblower, The Guardian, Retrieved from https://www.theguardian.com/news/2018/mar/17/data-war-whistleblower-christopher-wylie-faceook-nix-bannon-trump.

Dommett, K., & Power, S. (2019). The political economy of Facebook advertising: Election spending, regulation and targeting online. The Political Quarterly, 90(2), 257-265.

Edelson, L., Sakhuja, S., Dey, R., & McCoy, D. (2019). An analysis of United States online political advertising transparency. arXiv preprint arXiv:1902.04385.

Krafft, T. D., Gamer, M., & Zweig, K. A. (2019). What Did You See? A Study to Measure Personalization in Google’s Search Engine. EPJ Data Science, 8 (1), 1–23. https://doi.org/10.1140/epjds/s13688-019-0217-5

Kreiss, D., & Mcgregor, S.C. (2019) The “Arbiters of What Our Voters See”: Facebook and Google’s Struggle with Policy, Process, and Enforcement around Political Advertising, Political Communication, 36:4, 499-522, DOI: 10.1080/10584609.2019.1619639

Rieder, B., & Hofmann, J. (2020). Towards platform observability. Internet Policy Review, 9(4), 1-28.

ProPublica. (2020). Facebook Political Ad Collector. Retrieved from: https://projects.propublica.org/facebook-ads/