Co-authored with Yasodara Córdova.
This blog was originally posted at Governance for Development, World Bank Group.
Read the Part 1.
Rosie the robot was built to analyze the public expenses of Brazil’s congress members and empower citizen demands for social accountability. As part of a flagship project, Operação Serenata de Amor, the focus was to use data science to better use and interpret government data that is publicly available.
Despite the current public anxiety over bots and artificial intelligence in social media, Rosie has built quite a fan base. She currently has over 40,000 followers on Twitter and communicates with citizens in an accessible and easy to comprehend way. While the project was a technical challenge, a crucial component of its success depended on how Brazilian citizens responded to Rosie and how they used her findings to hold their elected officials accountable.
We were pretty sure from the start that a group of geeks wouldn’t make Rosie work by themselves. Journalists, marketing people, designers and communicators—about half of the people involved in the project—were part of the effort from the first day. Communications professionals worked alongside the computer and data scientists to give Rosie a consistent voice and help her find her audience. Average citizens don’t have the time or technical expertise to sift through and interpret vast volumes of data.
She was programmed to do these tedious tasks and then present her findings in a neutral tone. Misuse of public funds isn’t limited to one party, so the team made a deliberate decision to give her a politically non-polarizing personality. Rosie’s corruption suspicions now became tangible to data scientists, journalists, and citizens alike for them to further learn whether government officials were spending public funds ethically.
Citizens could be more politically engaged every day by focusing their attention on the results suggested by these robots. By combining Rosie and Jarbas, and making open data more meaningful and accessible, the initiative made room for journalists to browse data and find their own stories. This increased the relevance of the project and the value of the open data shared by the government. This helped journalists feed continued public interest around the issue.
This project, Serenata de Amor, is the top Brazilian project on GitHub (an international platform for open-source projects). It is the top political project on GitHub worldwide, and it has also become one of the top political pages in Brazil on Facebook. The entire project was open to the public, so whoever was interested in contributing, auditing or merely following Rosie’s strategies – whether they are a civil servant or a citizen—could have access to all information about it, from the discussions to the resulting computer code and data. Also, the core project managers put together a comprehensive toolkit in English and Portuguese, Brazil’s official language, for people to learn about the project and gave talks and workshops to the data science community. Openness, transparency and building an engaged community nationwide were big reasons why the initiative was so successful.
The team also maintains a Telegram group, a message app similar to WhatsApp, only for data and computer scientists interested in the project and in civic tech in general. It is aimed at tutoring and mentoring those interested in data science, so that the initiative can be sustainable with a new generation of data scientists who will continue creating more apps, new functionalities, and better versions of the current prototype with other types of data and regulation information. Our project sparked discussions on technical topics around the web on platforms such as Y Combinator, or Reddit, forum sites popular with youth. Sustained engagement and interest around our platform continued, because we communicated Rosie’s findings in a way that was relevant for citizen’s lives and enabled them to act on them.
As artificial intelligence is a field that is constantly developing, the documentation of our process for the public to use and consistent community engagement focus are vital to increasing the capacity of the population to engage and replicate the initiative, ideas, and mindset into other areas and further expand social accountability. This active and transparent community build around Rosie has fostered trust in the algorithms sustaining Rosie and willingness to go to politicians and demand they justify themselves to her. This engagement also guarantees sustainability, since all open source projects are maintained partially by the community, which generate alternative resources to keep the project alive.
Where Communication Meets Community for Accountability
In Brazil, the traditional way of addressing problems with accountability in public official’s reimbursements has been through the courts, an action that is the role of the attorney general’s office. However, the pace of the resolving the cases is slow, and the costs of taking a politician to the courts tops $15,000. It is not an efficient way to address the issue.
Taking the problem to Twitter in a transparent way – by using an open-source robot with neutral language and factual data – was a solution that has proven to be more efficient and effective. Rosie forced members of Congress to explain themselves and correct their behavior. Public pressure can be more influential than an arrest, especially if the alternative is a lengthy and expensive court case that uses public funds. Second, using a social media platform to gather people around the open data itself, through an automated platform, has the potential to transfer power to the citizens and for them to take advantage of a pre-built space to hold their representatives accountable.