Using Machine and Human Intelligence for COVID Response: Three lessons for future programming

For 68-year-old Surji Devi from Nawada, Bihar life has come to a standstill amidst the global pandemic COVID- 19 outbreak. She can no longer mow grass and sell it to the market and earn her livelihood. In her own words, “Before the virus gets me, I might die of starvation since I no longer have food left to feed my family and myself for more than a week”.

COVID-19 has brought an unprecedented crisis with devastating consequences for hundreds of millions of people like Surji, affecting their health, livelihoods and food security. Informal sector workers, small businesses, migrant workers are among the hardest hit and the employment impacts of COVID-19 are deep and widespread. An ILO report estimates that 400 million people are likely to fall back into the poverty trap due to this crisis. In response of this global crisis, Grameen Foundation has set a goal to provide unconditional cash transfer to 6,000 poorest families, enabling 30,000 beneficiaries to meet their emergency needs, consumption smoothening or re-starting their small business. You can learn more about this crowdfunding initiative here.

Sophisticated digital platforms and Direct Benefit Transfer systems in India make it possible to transfer cash to large number of affected people in such emergencies in real time, with high degree of efficiency. However, a key impediment is, how to quickly identify and prioritize the most vulnerable and deserving segments. It is to address this challenge that the team at Grameen Foundation came together to quickly develop a robust system with a perfect mix of machine and human intelligence. Grameen4Giving — an android based application is a one-stop solution for the Front Line Workers (FLWs) to identify, screen, onboard and disburse cash grants and monitor the beneficiaries. The App is powered by intelligent logic for identifying the most vulnerable using a mix of economic, social and demographic indicators as well as community recommendations. With the algorithm, the application runs eligibility test for a beneficiary instantaneously and sends triggers to the project selection committee for their immediate review and decision making. Once the selection committee approves, Direct Money Transfer to the beneficiary’s bank account is triggered.

Jayraj Nath, Head, Technology at Grameen Foundation India, modestly mentions, “This is a quick-fix solution we have developed. However, as we collect more data points and profiles in the system, we will upgrade this with latest technology such as Machine Learning and Artificial Intelligence”.

Assessing Vulnerability

Core architecture of this system is based on the understanding that there are multiple factors that make an individual or a household vulnerable. This includes economic indicators (like income sources and level, land-ownership), proxy indicators which are a good predictor of poverty, like type of dwelling unit, household assets, access to risk management tools like insurance or social security, demographic and social characteristics such as old age, single parent households, disability, incidence of chronic diseases etc. At the first stage, the Field Level Officers conducts a Participatory Wealth Ranking (PWR) exercise in the targeted village with the help of community leaders. Then, the beneficiary profile and all the data points and photographs as outlined above is uploaded in the Grameen4Giving mobile application. The application assesses the vulnerability of the selected households on these parameters and develops a Household Vulnerability Index, rating the vulnerability of the household on a 5-point scale and recommends eligible households. To ensure that we reach the poorest of the poor households, households which are rated more than 3 on the index are categorised as eligible by the application. However, the households characterized as ineligible go through another round of scrutiny on the recommendation of Field Level Workers and supervisors to avoid any exclusion errors creeping from the logic model. In case the selection committee finds the profiles appropriate, it overrides the machine recommendation. Thus, empathy and human intelligence has been retained at the core of this technology-enabled solution.

Grameen Foundation has been working in the fields of Financial Inclusion, Health and Agriculture for over two decades. However, the COVID-19 pandemic has thrown a new challenge about how to quickly and effectively serve the poorest of the poor. The agile thinking and untiring efforts of teams at Grameen Foundation has culminated in key learning for any similar future programming. These learnings can be summarised in the three key points mentioned below:

Driven by empathy, enabled by technology — Technology has a great role as an enabler in addressing the vulnerabilities arising out of a pandemic such as COVID 19, but empathy has to be the main driving force and at the core of any programme design.

Use a mix of machine and human intelligence — It is highly recommended to use machine intelligence to negate selection bias but one shouldn’t discount human intelligence in assessing phenomena as complex as vulnerability. Giving an appropriate weightage to community voices is crucial.

Co-Design by building prototypes — In scenarios, when the system development needs to be agile, it is always recommended to co-design by rapid prototyping of the applications. The user experience and workflows for systems like these are as important as any other and one should not lose sight of these aspects. At GFI, this co-design process integrated all departments such as strategy, programme, M&E and technology with design inputs from the Front Line Workers.

Grameen Foundation India

VIEW ALL POSTS

Leave a reply