January 1st, 2030

São Paulo, January 1st, 2030.

Another year has gone by. Balance of feelings: frustration, sadness and an everlasting ray of hope. A esperança é a última que morre we say. Today a new president should take office in Brasília. This would have supposed fair elections last year. And four years ago. And four years before that. Instead, Brazil has fallen into the authoritarian populist trap, and we remain stuck with General M. It has now been twelve years since dictatorship-lover Capitan B. democratically took over Palacio do Planalto. Nine years since his opportunistic even-further-right military vice-president arranged his stay in power following the President’s death. How did we get here? It started by approving an act against terrorism, targeting primarily leftist social movements. It was soon used to justify closing the borders during the Venezuelan crisis to stop any “communist militant” from infiltrating soon-to-be-glorious-again Brazil. Then they dismantled the main press conglomerates, because…well, fake news. Then militaries and neo-Pentecostal were “elected” heads of our 27 states, and became the large majority in Congress. Extreme statements and policies progressively discouraged most of foreign investors. Today, our routines involve low paying jobs for the (very) lucky, arbitrary arrests and a few disappearances. Rampant violence. And corruption. Unfortunately, chaos is far from being limited to Brazil. I wonder if Pinker would still call this time as the most peaceful of human history. This past decade singles out for its genocides… Myanmar, Yemen, Venezuela, Honduras, Salvador, DRC, CAR…to name a few. With Europe and North-America completely withdrawing from the 1951 Refugee Convention, emerging economies became the primary destination of cohorts of migrants and asylum-seekers fleeing either violence or starvation. Or both. Statistics still seem unbelievable: the overall number of migrants and refugees has quadrupled over the past 15 years and now represents 40% of the population worldwide.

When we observed this trend years ago, it did seem like an area of opportunity for social good: the number of asylum applications was increasing exponentially and countries were not able to keep up with the demand. Asylum requests’ backlog was growing every month, in Latin America, Africa, Europe, US, Canada and elsewhere. We nearly automatized the assessment process by standardizing Country of Information databases to run a high performance fact-checking algorithm. In addition, our fast track system conceived for ‘easy cases’ allowed applicants from countries with an acknowledged in crisis to have an earlier and shorter interview. Our tool reduced the overall length of the asylum procedure by four. In Brazil and Mexico, our pilot countries, we virtually wiped out the backlog. Over 50% of requests were approved (an increase compared to previous years) and 30% were redirected towards immigration services to apply for another status.

Our pilot standardized Country of Information database was first implemented and used in Brazil, and over the course of a few years, and through a partnership with key European NGOs, it ended up being adopted by nearly 90% of countries. It was hosted in Europe, it was GDPR compliant. How could we have imagined such a high-level behind the scenes agreement between the most powerful to turn this official database into a center of misinformation used to reject nearly all asylum procedures? Any person daring to raise that hypothesis would have been laughed at and accused of childish conspiratorial theories. Combined with the fast-tracking system, which ended up being used to expedite the rejection of the ‘easy cases’, our technology was used as a highly efficient tool to sustain anti-immigration / asylum policies.

Of course, we had initially talked with ethics and legal scholars, migration and refugee specialists and other well-intentioned advisors. Concerns about refugees’ privacy, information (and power) imbalance between asylum seekers and governments, questions about whether it was fair or not to benefit certain nationalities and the consequences for the non-priority cases were raised. We held extensive brainstorm sessions about the potential misuses of our technology. But nobody saw it coming. The problem wasn’t really the technology itself. The truth is that nobody believed that the world would take such a drastic turn to the extreme-right. Nobody believed that international treaties would be so disregarded. Not in the 2030 Agenda era: Sustainable development for all countries, human rights, gender equality, climate change, protection of vulnerable populations… International migration was supposedly a critical concern for the implementation of the SDG Agenda. This year, all countries should have eradicated poverty and inequalities, our cities would be prosperous and breathable, our forests green, and refugees protected. I guess the right to asylum was just the first of the many of our principles to be neglected in governments’ journeys to neo-authoritarianism.


Assignment Week 9: IDEO cards


I used the “Flow analysis” card to represent the flow of information throughout the asylum procedure in Brazil (see image above). It allowed to map the bottlenecks (and corresponding lengths) in the procedure, the main information products associated with each stage (questionnaires, etc.) and actors involved. The flow diagram was also useful during the interviews, both to explain to interviewees the solutions we are imagining but also to structure the conversation. The next step would be creating the “ideal flow”, incorporating proposed solutions.

Asylum services

Asylum services are extremely lengthy, frustrating and adding to the hardship of asylum-seekers that are trying to re-build their life. Part of the problem is that asylum services are overwhelmed with more demands than they are equipped to handle. We are looking to shorten processing times using data analysis and inference.

Brainstorm 5 core values:

  • (more) Justice through Transparency (i.e. in the decision-making chain)
  • Consider negative impact (i.e. consider access/marginalization when thinking about digitalization of public services)
  • Rethink representations (all the more important as representations are heavily present in immigration/asylum topics)
  • Co-design and dialogue with every individual / stakeholder involved in the asylum process, from asylum seekers to judges making asylum judgements.
  • Shape alternative future


A 1-day consultation with every stakeholder involved in the asylum process (in Brazil)

Goal: based a on draft [ongoing], consolidate a holistic map of the asylum process in the country. The objective is to identify bottlenecks and assess feasibility of solutions.

Who: 1-2 asylum seeker (waiting for decision), 1-2 refugees (accepted as refugees), 1-2 officials involved in registration of asylum request, 1-2 ‘Eligibility officials’ (interviews, analysis), 1-2 officials in the decision committee, 1-2 representatives from UNHCR

Where: São Paulo, in offices of asylum services (needs to be perceived as an ‘official’ effort, but still be accessible to all – will need to provide commute support for refugees).

Asylum / social programs / crime

Asylum requests in Brazil

Asylum services are extremely lengthy, frustrating and adding to the hardship of asylum-seekers that are eager and impatient to rebuild their life. Part of the problem is that asylum services are overwhelmed with more demands than they are equipped to handle. According to a recent article, in Brazil, 86,000 asylum seekers are waiting to hear about their asylum request while there are only 14 officials able to undertake the assessments.

CODE: pre-assessment, i.e. notifying straightforward cases (e.g. applications from countries at war), fact-checking (e.g. consistency between event dates) using ML, NLP, could decrease the length of the procedure significantly.

LAW: regulation enforcing higher (public) budget allocation to asylum services would allow to increase Human Resources dedicated to the different stages of the process (I.e. conducting the interviews, reviewing assessments and evaluating requests) and therefore decrease overall length of the process.

NORMS: Part of the reason why the Brazilian (and overall Latin-American) asylum services are so under-equipped to handle the increasing demand is because the topic is not considered a priority, by the government nor the civil society. A push from civil society could help bring the topic to the public agenda, increase the number of volunteers to support assistance and even legal processes.

MARKET: in Brazil, asylum-seekers are allowed to work once they have initiated their asylum process. Companies could partner with civil society, even government, to offer jobs for asylum-seekers thus reducing anxiety over papers.

Targeting of social programs

Targeting is intrinsically one of the core challenges posed by Conditional Cash Transfers (social programs aiming to reduce and break the intergenerational cycle of poverty). Currently, there are more than 100 nation-wide CCT programs.

CODE: improve inclusion/exclusion error rates using algorithmic fairness methods, i.e. reducing bias in selection against certain population subgroups (for e.g. rural vs urban).

NORMS: In certain countries, social programs such as CCTs face heavy criticism, for several reasons including ideology, accusations of patronage, corruption and inclusion errors (“people that are not poor receive it”). Improves in targeting could themselves contribute to changes in public opinion.

LAW: CCTs should be as institutionalized as possible, i.e. legally defined as initiatives independent from elected offices.

MARKET: private companies can partner with the government to offer employment for professionals who are themselves head of a beneficiary family or graduates / young professionals whose family was a beneficiary.

Crime prevention

In Latin America, homicide rates are four times higher than the world average. Violence creates fear and uncertainty, affecting not only the primary victim but also his/her family and community, also known as secondary victims. Although abundant research exists on the nature and sociology of crime, only a few studies worldwide have explored the effects of crime and violence on daily routines / activities of both primary and secondary victims.

CODE: use behavioral data such as bank card transactions’ metadata to assess the effect of crime shocks on daily lives, with a focus on the differential impacts between subgroups, starting by women and men – and provide insights for evidence-based policy to increase community resilience in the face of shocks.

LAW: Crime is a multifaceted issue. In certain Latin-American countries, one of (the many) drivers of criminality is impunity, especially for crimes such as murders and rape. Lobbying for stricter law enforcement and persecution of high-level crimes instead of low-level crimes.

NORMS: perception of insecurity, not just insecurity, also has an impact on individuals’ behavior. In areas where perception of insecurity is higher than insecurity, information-based campaigns can help citizens feel safe in their neighborhoods.

MARKET: local economy and businesses are also affected by crime and should be part of public initiatives to increase resilience of communities in the face of crime shocks.

Post 1 and 2 / Intro and project reflection

Hi everyone! I happily joined the class last week, so I will combine post 1 & 2 here.

I’m Julie and I’m currently based at the Harvard Humanitarian Initiative, as part of the think tank Data-Pop Alliance co-created by Sandy Pentland and Patrick Vinck (HHI, HSPH), where I lead applied research projects around the use of big data/AI for public policy and development – with a strong focus on the LATAM region.

I studied in France, where I graduated from two masters in (1) Gender Studies (Université Paris Diderot – VII) and (2) International Development (Sciences Po Paris). My program in Gender Studies (officially in “Gender and development” at the time), gave me (much needed) materialist-intersectional-feminist views about the “development field” that had a strong impact on both my studies and my personal socio-political perspectives.

So it is through a “development” approach that I got into the “tech for social change” field. The projects I am currently involved in look into leveraging CDRs and bank data to study crime dynamics in Colombia (focus on drivers of criminality) and Mexico (focused on the impact of crime waves on citizens’ routines).

Project reflection:

The following is not (yet) an implemented project – but one at the stage of conception being actively discussed with a Media Lab PhD student.

When I moved back to São Paulo in 2016, I volunteered at the National Committee for Refugees (the governmental body in charge of deciding on all asylum claims in Brazil). At the time, Brazil was receiving an increasing number of asylum applications, notably from Syria, as the country had adopted a “laissez passer” policy for Syrian nationals. There I worked as an interpreter for interviews but also answered (or tried to answer) the questions from asylum seekers and refugees that would come asking about procedures or oftenly to check on the status of their applications – which were taking months (sometimes up to 1-2 years). In 2017, I was part of a small team that conducted surveys in a Syrian refugee camp in Greece. Most of the people in the camp, had done their asylum request and had been waiting for several months already to hear back about their status – and their new papers that should allow for the start of a new phase in their lives. I vividly remember a young refugee telling us “Why are you studying our use of mobile phones? What we really need is to get out of here, study, work, continue with our lives”.

To summarize the scope of the problem: in both countries, and anecdotally in others, asylum services are extremely lengthy, frustrating and adding to the hardship of people that are eager and impatient to start building a better life. Part of the problem is surely that asylum services are overwhelmed with more demands than they were/are equipped to handle (in Brazil, we were a team of 20 volunteers, also doing contextual background checks to contribute to the asylum request processes). Bearing that in mind, we are exploring the idea of using an algorithm, built to improve the targeting of social programs, in a stage of the asylum procedure, with the goal to expedite the process.

The ideal person to address this issue would likely be a specialist of the legal and procedural asylum process (in each country of implementation) along with a data scientist well versed on algorithmic fairness and bias.

The first potential consequence harmful consequence I imagine could come from systematically unfair predictions – as certain cases will be easier to ascertain (for e.g. citizens from a war zone, such as Syria) than others (for e.g. an individual being politically persecuted in a “peaceful” country).