This post was written by Sandy Rafaela Krambeck, a Master’s student in New Media and Digital Culture at the University of Amsterdam.
Brazil established its Access to Information Law (LAI) in 2012, which opened up possibilities to develop databases that can be subjected to new experiences with Artificial Intelligence (AI). Operação Serenata de Amor (OSA) is a ‘Data Publics experience’ that reuses public data to fight corruption. It is an AI project conceived in 2016 and operationalized in 2017. OSA aims to help control public spending by taking advantage of the opportunity created by the LAI to audit parliamentary costs through Rosie, an AI robot that uses algorithms to automatically read receipts presented in expense claims. Rosie finds among the receipts presented values that are outliers, that is, their values differ significantly from other receipts for similar activities. The suspicious spending that Rosie encounters are posted on Twitter (@RosieDaSerenata), featuring an emoji of a siren followed by the words “suspicious expenses found by – the name of the deputy. Can you help me verify?”. In addition, the post contains a link to a numbered document that exposes the data of the receipt presented by the deputy, such as company name, date, and value, including why the receipt is considered suspicious. This numbered document is made available by a platform, also created by OSA, called Jarbas whose objective is precisely to facilitate the visualization of the outliners encountered by Rosie. OSA is an open-source experience both to contribute to its development and to use. It is available on the GitHub platform, a source code hosting tool for private or open-source systems, which allows anyone in the world to collaborate.
Musskopf (2018), one of the founding members of the project, points out that there were changes in the deputies’ behavior both because supporters publicly called them out due to Rosie’s publications and the perception that there are now people monitoring. Coutinho and Freitas demonstrated the importance of OSA in the realization, facilitation, and stimulation of social control of public expenses, and Savaget, Chiarini and Evans pointed out the importance of technologies such as OSA to make citizens more politically active than simply choosing representatives. Nevertheless, perhaps the project’s most important contribution is the recovery of a considerable amount of money and exposing politicians who would have improperly pocketed them, especially considering that Brazil has high levels of corruption. According to the Transparency International website, the country scored 38 on the Corruption Perception Index in 2021, a scale that goes from 0 to 100, the higher the less the perception of corruption. At the time of writing, OSA has registered on its website 8,276 suspicious refunds found, totaling BRL 3.6 million (Brazilian currency: Reais) among 735 deputies.
The Transparency Agenda (TA) is the release of government information and data to the citizens. They materialize mainly through the legislation that makes access to public information possible, in Brazil, the LAI. TA is seen as a model to be followed, as it is believed that they represent a more democratic way of auditing public data, however, the format in which this information is made available makes most people unable to assess them. As AI gains new approaches related to TA, awareness of who is able to use this type of data and what consequences this entails is necessary. Ruppert (2015) uses the term “Data Public” to both describe the people who reuse public data generated by TA’s and characterize how the handling of public data affects civic engagement. According to Ruppert “transparency devices are staged ‘as if’ everyone can easily participate and witness the inner workings of the state”(7), but often these devices are complex reports or huge databases that require knowledge that few people have. In this regard, Kennedy and Moss (2015) argue that it is necessary to shift the focus from “known publics” to “knowing publics“, that is, from data mining practiced by powerful groups to an active and reflective form of dealing with data. From their perspective, data mining should be used “for the public to be more knowing of itself and to participate in the active production of itself, the public” (9).
Due to the complexity of both TA and AI technologies, OSA became the mediator between a part of the Brazilian government’s public data and the citizens. Harvey, Reeve, and Ruppert call mediators people who understand and know how to work with the huge amount of data available. Consequently, people who have the ability to understand data have become the new gatekeepers of information, a position that was held by journalists before the World Wide Web and remains confined to a very small community. The presence of the intermediary in the consultation and handling of transparency portals highlights the lack of public autonomy in the matter. Thus, a crucial aspect of TA needs to be the citizen’s ability to understand the information. The existence of TA does not in itself guarantee transparency, it is necessary for the public to know how to use the tools and data.
According to Associação das Empresas de Tecnologia da Informação e Comunicação – (Brasscom) 2019 report, the Information Technology area in Brazil will require around 420,000 professionals by 2024, a number that is practically impossible to reach, as the country currently has around 50,000 trained professionals. Another issue that must be taken into account when it comes to access to technology in Brazil is inequality. As exposed by Oxfam International, someone earning the minimum wage in the country would have to work 19 years to earn the same as a Brazilian from the richest 0.1% of the population earns in a month. The characteristics presented above aggravate the digital divide, that is, the concentration of knowledge in a particular social class, which would reinforce the inequality that already exists in various forms in the country.
Teams that develop algorithms need to take into account the different realities of the society in which they will be impacted. OSA is an AI step in that direction as it develops open-source algorithms to fight political fraud and also releases all the data it collects, nevertheless, its role as a mediator between data generated by the government and the population needs to be seen as a temporary measure. Discussions around Data Publics cannot be summed up in government transparency, they must also strengthen the transparency of power relations in order to move from known to fully knowing publics. This means, in addition to more agency in data processing, ensuring access to disadvantaged and minority groups. The digital divide reflects and reinforces pre-existing social inequalities; the future of TA, as well as AI, needs to take these issues into account.
“2021 Corruption Perceptions Index – Explore the Results.” Transparency.org, https://www.transparency.org/en/cpi/2021?gclid=CjwKCAiA9tyQBhAIEiwA6tdCrMUcGJvYsCRtIuSO3u3vcCttVMkyrnBZHjBsP55cS9LFR698J1thVxoCoosQAvD_BwE.
“Brazil: Extreme Inequality in Numbers.” Oxfam International, 20 Oct. 2019, https://www.oxfam.org/en/brazil-extreme-inequality-numbers#:~:text=Brazil%20is%20decades%20away%20from,as%20the%20remaining%2095%20percent.
Constituição Federal. Presidência da República. Lei 12.527. Brasil, 18 nov 2011. http://www.planalto.gov.br/ccivil_03/_ato2011-2014/2011/lei/l12527.htm
Coutinho, Edurado D., and Angioberto S. Freitas. Public value through technologies developed with open government data: The Love Serenade Operation case. Revista de Administração Mackenzie, 22(6), 1–26, 2021, doi:10.1590/1678-6971/eRAMD210079
“Estudo BRASSCOM – Formação Educacional E Empregabilidade Em Tic.” Brasscom, 28 Oct. 2021, https://brasscom.org.br/pdfs/estudo-brasscom-formacao-educacional-e-empregabilidade-em-tic/.
Gillespie, Tarleton. “The relevance of algorithms.” Media technologies: Essays on communication, materiality, and society 167.2014 (2014): 167.
Kennedy, Helen, and Giles Moss. “Known or knowing publics? Social media data mining and the question of public agency.” Big Data & Society 2.2 (2015): 2053951715611145.
Harvey, Penny, Madeleine Reeves, and Evelyn Ruppert. “Anticipating failure: Transparency devices and their effects.” Journal of Cultural Economy 6.3 (2013): 294-312.
Musskopf, Irio. “O Impacto Do Controle Social Na Câmara Dos Deputados.” Medium, 6 June 2018, medium.com/serenata/o-impacto-do-controle-social-na-c%C3%A2mara-dos-deputados-c2b2a34db09e.
Ruppert, Evelyn. “Doing the transparent state: Open government data as performance indicators.” (2015): 127-150.
Savaget, Paulo, Tulio Chiarini, and Steve Evans. “Empowering political participation through artificial intelligence.” Science and Public Policy 46.3 (2019): 369-380, https://doi.org/10.1093/scipol/scy064