Problematizing the 'family' in welfare and social services data systems

Governments have an obligation to provide social security welfare benefits in cash or kind, though the extent and methods through which this is implemented has varied between countries and over time. Modern welfare systems rely on identification systems and government record-keeping to process applications and determine eligibility.

Increasingly, governments are using data systems, algorithms, and AI tools as part of service provisions and collecting and linking much more personal information about individuals. One purpose for this linkage is to detect fraud: though there are numerous examples of inaccurate and harmful automated fraud detection, and storing and linking large amounts of data may not comply with the Fair Information Practice of data minimization.

Linking data about individuals together is also done when welfare benefits are designed to be received by groups of individuals: families, households, or other ‘benefits units:’ for example, Supplemental Nutrition Assistance Program (SNAP) in the US, or social services assistance through the ‘Supporting Families Programme’ in the UK. Eligibility for welfare services may be demonstrated through evidence provided by other family members, as in Pakistan’s eligibility determination for the Computerized National Identity Card.

How individuals are linked together, however, often relies on assumptions about the relationships between individuals within families and households: which may not reflect people’s lived experience. Family relationships may or may not be discernible from outside the family: families may be linked by genetics, legal relationships, or by choices made by individuals.  Policy choices that impose a definition of ‘family’ often rely on cultural assumptions: this is particularly the case for whether ‘extended’ family members should be included. Family relationships are not necessary stable over time, nor even necessarily reciprocal. Linking individuals together in welfare data sets, however, often ‘sorts’ individuals into static, mutually exclusive families, even where this does not reflect the reality of their lives.

Where eligibility for welfare benefits is determined based on a group of people--whether this is families, households, or other groups—then there may be multiple legitimate ways for people to apply as a group. An experienced benefits navigator can help ensure that applicants provide information enabling them to claim all the benefits to which they are entitled, in order to meet their needs. As more and more welfare benefits systems are datafied, automated, and outsourced to AI tools, however, the opacity of the system increases. This makes it harder for claimants—and navigators—to understand if benefits are being allocated correctly.