The CPS Algorithm Nobody Knows About

What if I told you that whether your children are taken from you might depend on an algorithm you've never seen, reviewed by people you'll never meet, with criteria you can't challenge?

That's the reality for Texas families facing CPS involvement.

Sarah Chen, an investigative journalist specializing in family law systems, has been tracking a disturbing trend: the increasing use of predictive algorithms and automated decision-making systems in child protective services.

"CPS agencies are using risk assessment tools that are proprietary, opaque, and rarely subjected to external validation," Chen explains. "These tools heavily weight socioeconomic factors—poverty, parental unemployment, prior contact with police—and the net result is that poor families are far more likely to face investigation, removal, and loss of custody."

Her analysis revealed:

**The Disparities:** Families below the poverty line face investigation rates 5x higher than wealthier families. Families with any police contact (including minor traffic stops) are automatically scored as higher-risk, regardless of the incident.

**The Secrecy:** Most CPS agencies won't disclose what factors their algorithms consider or how they weight them. When families request this information, they're told it's "proprietary" or "part of our risk assessment process."

**The Appeals Problem:** Even when families challenge CPS's decision to investigate or remove, they can't examine the algorithmic factors because the algorithm itself is secret.

"This is the opposite of due process," Chen argues. "You can't challenge something you can't see."

Several states have passed laws requiring algorithm transparency in family services. Texas is not one of them.

The impact is profound: - Over 6,000 children enter foster care in Texas annually - Foster care outcomes for children are statistically worse than remaining with biological parents (even in imperfect situations) - Once in the system, it takes an average of 18-24 months to reunify families

"An algorithm can't account for a mother's love," Chen notes. "It can't understand cultural differences in parenting. It can't see the full humanity of a family. Yet it's making life-altering decisions every single day."

Her recommendations: 1. **Transparent Algorithms:** All risk assessment tools should be publicly available for examination 2. **External Validation:** Third-party audits of algorithmic accuracy and bias 3. **Appeal Rights:** Families should have the right to examine and challenge algorithmic assessments 4. **Bias Audit:** Regular review of whether algorithms perpetuate discrimination by race, income, or other protected classes 5. **Human Judgment:** Restore human decision-makers (appropriately trained) as primary determinants rather than algorithms

"Technology should serve families," Chen concludes. "Instead, we're using technology to automate family separation. That has to change."