Check-in and access this session from the IGF Schedule.

IGF 2019 WS #112
Assessing the role of algorithms in electoral processes

    Subtheme

    Organizer 1: Hartmut Glaser, Brazilian Internet Steering Committee (CGI.br)
    Organizer 2: Flávia Lefèvre Guimarães,
    Organizer 3: Cappi Juliano, NIC.br
    Organizer 4: Beatriz Rossi Corrales, NIC.br

    Speaker 1: Sergio Amadeu da Silveira Silveira, Technical Community, Latin American and Caribbean Group (GRULAC)
    Speaker 2: Chris Marsden, Civil Society, Western European and Others Group (WEOG)
    Speaker 3: Scott Cunningham, Technical Community, Western European and Others Group (WEOG)
    Speaker 4: Lorena Jaume-Palasi, Civil Society, Western European and Others Group (WEOG)

    Online Moderator

    Cappi Juliano, Technical Community, Latin American and Caribbean Group (GRULAC)

    Rapporteur

    Beatriz Rossi Corrales, Technical Community, Latin American and Caribbean Group (GRULAC)

    Format

    Round Table - U-shape - 90 Min

    Policy Question(s)

    What are the main algorithm dimensions to consider in the challenge of regulating its use in political campaigns considering different countries electoral processes experience?

    SDGs

    GOAL 16: Peace, Justice and Strong Institutions

    Description: The workshop will be divided into three blocks. In the first block we will have three presentations of 7 to 8 minutes each. Moderator will open the workshop with a 3 minutes explanation of the session and will give the floor to the speakers. First presentation will be about the Brexit Referendum. The second one will be about the USA 2016 presidential election, and the third one about the 2018 Brazilian presidential election. Speakers will be oriented to prepare their presentations considering the following topics: 1) General context; 2) Main actors involved (this includes not only political actors but also digital platforms, big data marketing firms, etc.); 3) The role of algorithms in the main communication strategies used by political actors. The second block will be an open debate where participants and remote participants will be invited to present their ideas and make questions to the speakers, considering one questions: Are there similarities and possible analogies about the use of algorithm in political campaigns in different countries? Participants will be given 3 minutes to present their ideas. In the third block, participants will be invited to propose up to 3 dimensions to be considered regarding the challenge of regulating the use of algorithms in political campaigns.

    Expected Outcomes: The workshop will produce a list of dimensions and propositions to be considered to produce legal, political and governance instruments to deal with the challenge of regulating the use of algorithms in political campaings.

    The discussion will be facilitated by the onsite moderator who will guide the debate in each of the proposed segments for the workshop. Moreover, the online moderator will make sure the remote participants are represented in the debate.

    Relevance to Theme: Algorithms play a central role in controlling the flow of information in the digital era, becoming a crucial element in social life. The networked society witnessed the enormous dissemination of algorithms into a broad variety of social activities. Search engines help us to find a way through the web; recommendation algorithms map our preferences defining what is relevant to see and what is not; pairing on social networks are supported by algorithms; and so on. As data collection advances, the models behind algorithms become more robust, predictive, precise and better oriented, making it possible to anticipate and even influence individual choices and behaviors. However, this increasing power that arises from the use of algorithms is not equally distributed. Knowledge inequalities and data concentration have favored a specific group of actors who gained prominence in a global scale in economic, cultural and political fields. During electoral processes this prominence becomes evident. Digital platforms, search engines and Big Data Marketing Firms mediate much of the political debate and employ advanced data analysis methods to accomplish their objectives. But the potential impact of algorithms in electoral processes has raised concerns about the reach of the influence that these actors may produce and the risks of undermining democratic processes. The surprising victory of the Leave campaign during the Brexit referendum in 2016 is an example. The victory made the headlines of newspapers and magazines all around the world. It is worth to remember that the United Kingdom's presence in the European Block had several advantages. David Cameron, re-elected in 2015, a year before the referendum, was one of the important public figures to oppose Brexit. The fact that Cameron supported the Remain campaign made both communication apparatus of the conservative party - his party - and that of the Labor party - that was reportedly in favor of the Remain campaign - to be used against Brexit. Even so, the Leave campaign, which was led by a tiny far-right party, the UK Independence Party - UKIP, was victorious. The role of the use of algorithms in the referendum campaign is still an issue for debate. The spread of fake news in the presidential campaigns of USA in 2016 and in Brazil in 2018 is another example. The increase of disinformation in electoral process have put more pressure on the debate about the creation of legal, political and governance instruments to deal with the challenge of regulating (not only) algorithms. The workshop proposes to gather researchers, platform representatives, public agents and other civil society actors from different countries to bring experiences from different political contexts in order to discuss algorithm dimensions and propositions to the regulation of its use in political campaigns. From algorithm dimensions we understand the ways in which the introduction of algorithms into human practices may have political ramifications, as explored by Gillespie Tarleton, in his paper The Relevance of Algorithms, 2016. Some important algorithm dimensions are Patterns of inclusion, Evaluation of relevance, Algorithmic objectivity, Production of calculated publics, which will be explored during the workshop. Others may be proposed.

    Relevance to Internet Governance: Considering algorithms as sets of rules that are applied to one or more initial inputs transforming them into a final output, even if we do not see their codes and do not know how they work, we do know their outputs and efficiency, and, more important, we can assess their capabilities and trust their results. This would not be an issue if algorithms were just a simple agglomeration of commands that blindly perform their tasks, but, as its uses impacts economy, culture and politics, we can surmise that algorithms may reproduce inequalities and serve purposes that may be of interest only to particular groups and not to the society at large. The pervasiveness of algorithms combined with their invisibility influences social participation in complex social processes - such as the three political events mentioned above. Thus, we can assume that algorithms play a sort of power that exercises a kind of discipline over others. In this sense, discussing the development of legal, political and governance instruments to regulate this crucial element of the networked society - the algorithms - is central to the global Internet governance debate, insofar as unintended influences of its uses can put at risk the Internet's development potential and jeopardize democratic processes in different countries.

    Online Participation

    Online participation and interaction will rely on the WebEx platform, also there will be a online moderator encouraging remote participation.

    Proposed Additional Tools: Social media (twitter and facebook) will also be employed by the online moderator who will be in charge of browsing social media using some hashtags (to be defined).

    Agenda

    0-3 min - Workhosp opening by moderator

    1st block - General context presentations:

    • 3-11min - Mr. Chris Marsden will talk about Brexit Referendum
    • 11-19min - Mr. Scott Cunninghan will talk about USA 2016 Elections
    • 19-27 - Mr. Sergio Amadeu will analyse 2018 Elections in Brazil

    2nd block - Open debate:

    • 27-52min - Open debate where participants and remote participants will be invited to present their ideas considering the following question: What are the main challenges to regulate the use of algorithms considering the following dimensions: transparency, secrecy, audit, others? Participants will be given 3 minutes to present their ideas. 

     

    3rd block - Final remarks:

    • 52-90 - In order to raise proposals that contribute to deal with the challenge of regulating the use of algorithms in political campaings, participants will be invited to discuss the following question: what regulatory and governance approaches can be implemented to regulate the use of algorithms in political campaigns? Each one will be given 3 minutes to present their ideas.
    1. Key Policy Questions and Expectations
    1. What is the role of algorithms in political campaings?
    2. What are the main challenges to regulate the use of algorithms considering the following dimensions: transparency, secrecy, audit, others?
    3. What are the main regulatory and governance aproaches to consider in the challenge of regulating the use of algorithm in political campaigns?

    We expect to produce a list of proposals that may contribute to the development of legal, political and governance mechanisms to deal with the challenge of regulating the use of algorithms in political campaings.

    2. Summary of Issues Discussed
    The discussion started with presentations exploring the role that of communication strategies based on the use of algorithms to distribute content had in the dissemination of disinformation, mainly through social networks platforms, in electoral processes. Both presentations supported the connection between digital platforms and the emergence of an advertising industry that benefits from the spread of disinformation. However, one of the speakers proposed that the evidences on the influence that this strategies played during 2016 presidential election in USA is inconclusive. The open debate brought the social impact of algorithms beyond the issue of electoral processes. As a counterpoint, one participant exposed the influence of different aspects of the political contexts influencing democratic processes. This questioning led the discussion to an important point of disagreement where from one side participants where exposing illegal practices conducted during political campaigns - many of them based on the use of algorithms and the massive collection of personal data -, and from the other, researchers were trying to draw attention to other important elements of the electoral contexts, stating that an unnecessary effort is being made to regulate algorithms.
    
    3. Policy Recommendations or Suggestions for the Way Forward
    The workshop had an important contribution from other countries. As soon as the debate presented different positions on the understanding of the role of algorithms in electoral processes, international experience around specific cases where practices based on the use of algorithms have influenced democratic electoral processes were presented. The phenomenon proved so recent that proposals to deal with it are still being cooked . In the very beginning, one of the guest speakers quoted five recommendations to be considered regarding algorithms and disinformation: media literacy; strong human review and appeal processes where AI is used; independent appeal and audit of platforms; standardizing notice and appeal procedures (creating a multistakeholder body for appeals); and transparency in AI disinformation techniques.
    
    4. Other Initiatives Addressing the Session Issues

    This workshop provided an extremely rich meeting between different countries who shared their experiences. Beyond the contribution of electoral processes in United Kingdom, Unites States and Brazil, briefly described in the workshop resume, during the session representatives from Hong Kong, Italy, Trinidad and Tobago, Ecuador, Colombia e Honduras, brought examples of the influence of algorithms in their local elections and shared their experiences. Hong Kong representative spoke about the spread of disinformation in his country and its multifaceted form, such as the target of disinformation towards Hong Kong citizens and its source which, differently of the others cases, was based in China. Italy in turn mentioned the raise of hate speech during their elections and how this escalates outside the Internet boundaries as physical violence toward policy makers. Trinidad and Tobago pointed out that their couple last general elections were influenced by the actions of Cambridge Analytica and now, the Electoral Comission is investigating how so much personal data was available to political parties.

    5. Making Progress for Tackled Issues

    A few punctual initiatives towards framing algorithm impact in electoral processes were presented. Beyond the Co-regulation approach mentioned in the topic 3, Policy Recommendations or Suggestions for the Way Forward, some initiatives raised were (i) the development of fact check agencies (Honk Kong example); (ii) to focus on the understanding not only of paid advertisement, but also in spontaneous advertisement, spread by bots, for instance, fake accounts, etc.; (iii) to foster high quality journalism in order to combat disinformation. Aside this inputs, one important discussion about the function of disinformation took place. In this matter, Lorena Jaume reminded some historical and philosophical studies showing that disinformation does not appear to convince you to believe in something that you don't believe, but to build identities and gather together people who already believe in similar ideas; disinformation, in this studies, are a discourse of domination.

    6. Estimated Participation
    The estimated number of participants in the room was 74. Of these, about 36% (around 27 persons) were women, and 64% (47 persons) were men.
    7. Reflection to Gender Issues

    As the theme of this workshop was about algorithms in democratic processes, no gender issue was debated either directly or indirectly.

    8. Session Outputs

    Considering that our expected goal with this workshop was to produce a list of proposals in order to contribute to the development of legal, political and governance mechanisms to deal with the challenge of regulating the use of algorithms in political campaigns, we had two different approaches presented. The first was the co-regulation model, which aims at complement self and public regulation structures to define recommendations on specific principles, standards, and measures designed to establish principles and forms of collaboration to content moderation. The second was the recommendation of initiatives to be considered regarding algorithms and disinformation, namely: invest in media literacy; strong human review and appeal processes where AI is used; independent appeal and audit of platforms; standardizing notice and appeal procedures (creating a multistakeholder body for appeals); and transparency in AI techniques. It is worth highlighting that we carried out a deeper debate about the role of algorithmic technology in electoral processes in different countries.