While countries adopted various COVID-19 response approaches, the need for short- and long-term Cash and Voucher Assistance (CVA) quickly became clear at an unprecedented scale for previous cash and social protection beneficiaries and for newly vulnerable, unstable, or insecure populations.
At the same time, quarantines, lockdowns, travel restrictions, and bans on group assemblies forced a shift in standard humanitarian and social protection playbooks. The COVID-19 digital response drove a rapid move to remote and digital channels for targeting, registration, delivery, and monitoring of CVA.
Some predict that these shifts will lead to a ‘new normal’ that brings opportunities for improved scale and efficiency of CVA through digitization. However, there are ethical and data responsibility concerns that need to be considered and mitigated for immediate implementation and for future programme design.
The Data Responsibility and Digital Remote Targeting during COVID-19 Case Study explores digital remote targeting approaches that GiveDirectly is using for cash assistance and ways that the organization is addressing data responsibility.
Call Detail Records for Cash Assistance
GiveDirectly, together with Innovations for Poverty Action and the Center for Effective Global Action are exploring an emerging approach to remote targeting that uses non-traditional data sources to pinpoint geographic sub-areas with the poorest individuals and to subsequently identify those who might qualify for an emergency cash transfer due to COVID-19.
Early results indicate that this approach has been effective at quickly delivering COVID-19-related cash transfers to a large number of individuals living in extreme poverty in record time. The method has elicited debate, however, in that it uses mobile phone call detail records (CDRs) and other forms of unconventional data sources.
There are privacy concerns related to the use of this data as well as concerns about the potential that the most vulnerable who may not have access to mobile phones are excluded. At the same time, there is some evidence showing that big data methods can actually lead to greater inclusion, especially when the alternatives for in-person enrolment remain scant.
It remains to be seen whether this type of non-traditional approach to remote targeting could be attempted in a highly fragile context, or if it would be suitable and safe to use for targeting in contexts with a large number of transient migrant, internally displaced or refugee populations with limited identity verification.
Ethical Considerations
Although promising, implementers should thoroughly assess the ethical considerations that are associated with such new technologies. Ethics in humanitarian work is grounded in the principles of humanity, impartiality, neutrality and independence.
A fundamental principle of humanitarian action is the principle of Do No Harm. Data ethics considers moral problems related to the use of data and algorithms, among other areas. Common ethics issues that arise when working with advanced data analytics approaches such as predictive analytics and machine learning include:
- Validity: Is the data or model representative of what is being measured?
- Bias and Fairness: Is the data skewed? Is there any prejudice or favouritism in the data or model? Has there been an over- or underestimation of what is being measured? Are some members of the population more or less represented than others?
- Ossification: Is the model (or the underlying data) codifying existing, systemic biases and thereby making it harder to change?
- Transparency and Explainability: Is there documentation of the process? Can others easily comprehend and explain how the model or algorithm(s) function?
- Privacy and Anonymity: Is the data somehow revealing the identity of individuals or groups?
These aspects come into play alongside the more concrete procedural elements of data privacy and security that must be addressed along the full data lifecycle.
Sorry, the comment form is closed at this time.