The Slippery Slope of Algorithmic Campaigns: The 2019-20 Israeli Electoral Digital Campaigns

Dr. Sharon Haleva-Amir

Between 2019 and 2020, Israel held three election cycles within a single year, an unprecedented period that created a unique research opportunity to examine the evolution of online campaigns in real time. The electoral campaign for the 20th Knesset (2015) was the last to belong to the fourth generation of data-driven political campaigns that characterized the first decade and a half of the current millennium.

In an article published in 2022 in the journal Israel Affairs, I examined the online campaigns of these three election cycles through their most prominent negative trends. In the article, I argue that these electoral campaigns mark a transition to a new phase in political communication - the fifth generation of political campaigns: Algorithmic Campaigns. These campaigns are characterized by extensive use of algorithms and artificial intelligence techniques that enable sophisticated manipulation of public opinion while remaining hidden from public view and evading oversight and enforcement.

The central question around which the article revolves concerns the threats that algorithmic campaigns pose to democratic processes, and the way in which the gap between legal tools and technological advancement is exploited by political actors. The research, based on case study methodology, systematically monitored the online campaigns through social networks, media coverage, and text messages over approximately three months of each election campaign.

Major Negative Phenomena

The research identified several key negative phenomena. The first negative phenomenon is disinformation (fake news, fake accounts, and fictitious identities). The term "fake news" refers to a variety of actors who create and disseminate disinformation. The term, when applied to politics, refers to manufactured narratives and the confrontational and defensive techniques utilized by politicians. Fake news can influence public perceptions, distort election results, and damage the legitimacy of institutional media. During the 2019-20 election campaigns, the term became deeply embedded in political discourse. Political actors, who serve as central nodes in social networks due to their substantial number of followers, can greatly increase the range and scope of information dissemination.

As an example of deliberate disinformation aimed at preventing an entire public from reaching the polls, I note that in various Arabic WhatsApp groups, a file was circulated posing as an official announcement from the Ministry of Health, using the official format and the ministry's website logo, stating that Israeli citizens were instructed to wear face masks when leaving home. The plan was designed to reduce the voter turnout among Arab Israelis during the initial phase of the COVID-19 pandemic. The most dramatic example illustrates the helplessness of oversight mechanisms in the face of real-time interference in the electoral process: on the actual date of the third election round (March 2, 2020), PM Netanyahu distributed a doctored video of Benny Gantz. While in the original video Gantz called on his supporters to vote for Kachol-Lavan, the tampered video created the opposite impression whereby Gantz was calling not to vote for his party. Despite urgent petitions to the Central Elections Committee, seven hours passed, during which the video was virally distributed, and the electoral damage was done, before the committee ordered its removal.

Another negative phenomenon is political bots, influence and interruption networks and organic support simulation (astroturfing). Political bots are algorithms operating on social media that impersonate real people with the goal of manipulating public opinion while disseminating information in massive quantities. In recent years, political bots have become an integral part of political discourse worldwide, and their activity accounts for up to 42% of all online traffic volume.

During the 2019-20 campaigns, several politicians operated chatbots through Facebook Messenger, but Netanyahu's "Bibi-Bot" surpassed them all in sophistication. Through playful interaction, the Bibi-Bot collected data on users' political positions, thereby enabling mapping according to the probability of their voting for the Likud. Subsequently, it turned supporters into a call centre for voter recruitment and gave the Likud the means to transmit inciting and illegal messages, and to publish polls while violating election laws and Facebook regulations below the media radar.

Private groups in which people and bots promote content with the aim of deceiving the algorithm create hidden influence networks. Some news media stories exposed influence networks in all three elections rounds whereby non-politically related groups and pages (such as The Big Brother VIP lovers’ Forum and Bold Love) were transformed into designated pages, which, in a blunt and unambiguous manner, promoted Likud’s platform. The Big Bots Project identified in March 2019 a network of 154 fake accounts that distributed more than 130,000 tweets in favor of Netanyahu. According to data published in the press, in the last election campaign in November 2022, bot activity and fake profiles in Israel increased by 328%, and 54% of "swing voters" who were exposed to influence network activity on behalf of a certain party even voted for that party.

Political Astroturfing is a term based on the name of the artificial grass company AstroTurf and means simulating organic support - as opposed to grassroots (authentic organic support). This manipulation is carried out through coordinated dissemination of disinformation on social networks by participants posing as authentic users. Network members spread expressions of support, distract from expressions of opposition, direct attention to polarizing issues, and attack opponents, thereby potentially influencing election results. The State Comptroller's report on party expenditures during the election period from November 2021 reveals that during the elections to the 21st and 22nd Knesset, several parties operated fake profiles on social media with the aim of tilting election results. The main difficulty is the lack of ability to identify the puppet-masters without inside information. One solution is recruiting the user public through civil society organizations like FakeReporter to fight malicious activity.

Another phenomenon concerns illegal data accumulation. After the Cambridge Analytica scandal in the 2016 US presidential elections, Facebook claimed it established a policy limiting data collection, but this remained vague. On March 18, 2019, Facebook announced that the Likud violated the policy by using a chatbot to collect users' personal data; this announcement did not actually change the continued violation of the policy. It was later discovered that the Likud maintained a database of one million citizens tagged by political affiliation - contrary to privacy protection laws (the Elector application).

The last negative phenomenon concerns comprehensive non-compliance of political actors with the law, and particularly anonymous text messages. The Election Law (Propaganda Methods), 1959, left the online sphere unregulated despite repeated criticisms by successive chairs of the Central Elections Committees since 2001 regarding the legislative lacuna requiring correction. A public committee headed by former Supreme Court President Dorit Beinisch recommended in November 2017 amending the law and adapting it to the internet era, but the political echelon did not implement the recommendations.

Central Elections Committee Chairman Judge Melcer issued a precedential decision in February 2019 that applied The Election Law (Propaganda Methods) to the internet by prohibiting anonymous campaign publications on social media and search engines (CEC 8/21). In March 2019, the dissemination of anonymous text messages was explicitly prohibited (CEC 27/21). The new directives reduced the scope of violation but did not eliminate it. Anonymous text messages played a vital role in spreading disinformation, fake polls, defaming candidates, and reducing the election chances of rival parties. This practice indicates a growing unwillingness of political actors to comply with the law, as well as a practical inability of the committee to control the online electoral jungle. In June 2022, Amendment No. 40 to the Elections Law (Propaganda Methods), 1959, came into effect, applying the transparency obligation to the online sphere as well. 

The Elections Law forbids the release of polls within 96 hours of an election, yet Netanyahu's chatbot disseminated poll data two days before the 2019 elections. Following complaints, Facebook briefly suspended the bot for three hours. Both Facebook and the Elections Committee stated that enforcing laws in real-time was an "impossible mission." Additionally, an incendiary message targeting Arabs, sent by the Bibi Bot, was only removed under significant pressure. Furthermore, concerning the manipulated Gantz video, Judge Hendel declared that the committee would not initiate its removal due to impossibility.

Conclusions

The rare opportunity to study three electoral rounds held within one year provides a broad picture of the state of affairs. From this perspective, it can be assumed that we are not dealing with circumstantial or temporary trends, but rather a deep and long-term change. Campaigns in the third decade of the 21st century are personal, segmented, micro-targeted, sophisticated, and far more manipulative than in the past. The combination of widespread distribution of information items on social networks and advanced computational techniques creates highly toxic environments, reduces the possibility of holding civilized discussions, and deepens political polarization.

Given the accelerated development of artificial intelligence, it is clear that we are still at the beginning of a steep slope. In the future, methods of manipulating public opinion will be more sophisticated than we can imagine. Since technology precedes law and justice, and political actors exploit this, the public and democracy are harmed.

The Israeli case study provides vital lessons for democracies worldwide regarding the balance between freedom of political expression and the right to information, and the protection of the integrity of democratic processes. In an algorithmic world, where it is challenging, if not entirely unfeasible to prove a connection between the puppet-masters and those commissioning false advertising campaigns, a legislative solution alone will not suffice. Consequently, it is imperative to both increase public understanding of this matter and create multifaceted solutions, educational, legislative, and regulatory, that are suitable for the unique challenges presented by the age of algorithmics and artificial intelligence.

Dr. Sharon Haleva-Amir is a political communication scholar. This post is based on the following article:

Haleva-Amir, S. (2022). The 2019-20 Israeli electoral digital campaigns: Algorithmic campaigns' slippery slope. Israel Affairs, 28(6), 940-955. https://doi.org/10.1080/13537121.2022.2134397