A look at what happens when algorithms used to automate health assessments in Arkansas start cutting essential state-sponsored home care for disabled patients
Colin Lecher / The Verge : Tweets: @mslopatto , @flyosity , and @hccandrea Tweets: @mslopatto : This is what happens when you put an algorithm in charge of healthcare: people's benefits are cut and no one knows why http://www.theverge.com/... http://twitter.com/... Mike Rundle / @flyosity : Algorithms are written by people. They reflect the intent and priorities and prejudices of the people who write them. The misconception that they're bias-free and created by computers is very dangerous. http://twitter.com/... A E Johnson / @hccandrea : When algorithms intended for good, do harm...there's really a lot of design thinking we need to do before we unleash these powerful tools. http://www.theverge.com/...
Context & Ripple Effects
When The Verge reported on Arkansas's assessment algorithm in 2018, it documented an early case of a state handing benefit decisions to a scoring system whose cuts to disabled patients' home care nobody could explain. The reporting captured the core problem: the model encoded its designers' priorities, yet operated without transparency or appeal. What followed across the coverage arc shows the pattern hardening rather than correcting — investigators later found similar care-targeting software infusing racial bias into decisions about who gets stepped-up treatment, while [[a:960741|civil lawyers began building litigation strategies specifically against automated systems that deny the poor basic services]].
By 2023 the same mechanism had migrated from public benefits into private markets, with [[a:837733|insurers using unregulated predictive algorithms to pinpoint when they can plausibly cut off payment for older patients' treatment]]. Arkansas was the template case: cheap, scalable, and legally untested.
First-order effects
- Disabled Arkansans on state-sponsored home care lose hours of support with no explanation of how the score was computed, leaving them unable to contest a decision neither caseworkers nor beneficiaries can audit.
- State administrators gain a cost-control tool that shifts blame from policy choices to 'the algorithm,' insulating officials from the political consequences of cutting benefits.
Second-order effects
- Civil-rights and legal-aid lawyers respond by developing targeted litigation strategies against automated denials of public services, turning opacity itself into a legal vulnerability.
- Other states evaluating the Arkansas model face a choice between adopting the same scoring approach for their own Medicaid-style budgets and waiting for courts to establish whether unexplained algorithmic cuts survive challenge.
Third-order effects
- If the pattern holds, algorithmic rationing becomes default infrastructure across both public benefits and private insurance — with the racial-bias findings in care-targeting software suggesting the harm concentrates on populations least equipped to appeal.
- The gap between deployment speed and oversight capacity pushes accountability out of legislatures and into courtrooms, making judges the de facto regulators of benefit-setting algorithms until formal rules exist.
The trend: Algorithms are moving from advisory tools to de facto decision-makers in health and welfare rationing faster than transparency requirements, appeal mechanisms, or regulation can be built around them.