To demonstrate how the integrated critical lens operates in practice, let us apply all four frameworks to a single AI system: a predictive risk scoring tool used by a local government to identify families at risk of child abuse and neglect.

Systems of this kind are deployed in multiple countries, including the Allegheny Family Screening Tool in the US documented by Virginia Eubanks, and similar tools in the UK, the Netherlands, and New Zealand.

Through a CRT lens:

The system’s training data draws heavily on records of families’ interactions with public services, welfare benefits, housing assistance, involvement with the criminal justice system. These records are more complete for families of colour, who are disproportionately represented in public service systems due to structural racism in housing, employment, and criminal justice. The algorithm therefore has more information about families of colour, and scores them as higher risk on this basis alone.

The system encodes the structural racism of existing child welfare practice, in which families of colour are disproportionately investigated, their children disproportionately removed, and their experiences disproportionately absent from the design of systems nominally intended to protect them. Interest convergence analysis asks: when child welfare AI reform occurs, does it primarily serve the interests of the families most affected, or does it serve administrative efficiency and liability management?

Through a feminist lens:

The families most affected by child welfare AI are overwhelmingly headed by single mothers, disproportionately women of colour in poverty. The system was designed without meaningful input from these families. Its definition of risk reflects assumptions about what constitutes a safe family environment that embed middle-class, two-parent household norms. Women’s poverty, itself partly a product of gendered economic structures, is encoded as evidence of risk.

The data collected on these families by welfare systems is more detailed than data on wealthier families, a manifestation of what Eubanks calls the “digital poorhouse.” The invisible labour of mothering, the care, the advocacy, the community maintenance performed by these women, is not counted in the risk assessment. Only contact with institutional systems is counted. Care ethics would ask: what responsibilities does the system have to the specific mothers and children it affects?

Through a Marxist lens:

The system is built and maintained by private companies contracted by local governments, a transfer of public welfare responsibility to private profit-seeking entities. The risk assessment outputs are used to justify interventions that affect the most economically vulnerable families, while the families of wealthy and upper-middle-class parents, whose child welfare issues are more likely to be handled privately, are largely invisible to the system.

The system serves the administrative interests of local government, identifying families for resource allocation, managing political and legal risk from child welfare failures, rather than primarily serving the interests of the families themselves. Surveillance capitalism analysis asks: what happens to the data generated by these systems? Who owns it? How might it be used beyond its stated purpose?

Through a postmodern lens:

The system constructs the category of “at-risk family”, a classification with real material consequences. The families so classified begin to modify their behaviour in anticipation of assessment: they may avoid seeking help they need in order not to generate data that could raise their risk score. The looping effect means that the system’s classifications shape the social reality it claims to describe.

The explanations the system provides for its risk scores are post-hoc approximations, they tell families and caseworkers something about the factors that contributed to the score, but not everything. The authority of the algorithmic score displaces the judgment of caseworkers who have direct knowledge of specific families. The system’s claim to objective risk assessment conceals the value judgments embedded in its design.

What the integrated analysis produces:

Applied together, the four lenses reveal that this system: encodes and amplifies structural racism; fails the women and families it nominally serves; transfers public welfare responsibility to private profit; and claims objectivity while constructing a reality that shapes the families it classifies. No single framework produces this complete picture. The integrated critical lens does.

Reflection question: Identify an AI system in your own professional context. Apply all four lenses to it, even briefly. What does the integrated analysis reveal that you had not previously noticed?