The Risk–Danger Fallacy: a diagnostic framework for predictive harm in public child welfare systems
Public child welfare systems increasingly use actuarial and algorithmic risk tools to inform screening, investigation, and service decisions. Although these tools estimate the probability of future harm, concerns arise when probabilistic risk is interpreted as evidence of present danger. This article introduces the Risk–Danger Fallacy as a diagnostic framework for examining how predictive risk may be operationalized as a substitute for individualized evidence in child welfare decision-making.A conceptual synthesis of interdisciplinary literature from child welfare, decision science, administrative law, and algorithmic governance was conducted to examine how model design, institutional reliance, and governance incentives interact within child welfare systems.Four interacting evidentiary domains were identified: predictive validity, bias and disparities, organizational incentives, and practitioner interpretation. Across these domains, evidentiary substitution emerged as a cross-cutting mechanism through which probabilistic outputs may acquire practical influence exceeding their evidentiary validity.Predictive harm appears to arise from the interaction of technical design, organizational workflows, and governance structures rather than model performance alone.The Risk–Danger Fallacy clarifies the limits of predictive risk tools and identifies conditions under which evidentiary safeguards may be weakened, with implications for preserving the distinction between probabilistic inference and evidentiary proof.
Authors
- Megan McCarthy (ORCID: https://orcid.org/0009-0006-3054-5361)
Publication Details
- Journal
- Journal of Public Child Welfare
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1080/15548732.2026.2764956
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00