Beyond the Risk Score: Elevating Survivor Strengths and Restoring Professional Judgment
By David Mandel, CEO and Founder, and Ruth Reymundo Mandel, Chief Business Development Officer and Credible Expert, Safe & Together Institute
In today’s child welfare systems, predictive analytics tools are increasingly viewed as a path toward efficiency and consistency. Algorithms crunch decades of administrative data to generate risk scores, supposedly guiding frontline workers toward better, faster decisions. Yet for families impacted by domestic abuse, these tools can function as engines of harm—reinforcing stereotypes, misreading safety efforts as risk, and missing or masking the behaviors of those actually responsible for creating danger.
This is not a glitch. It is a feature of systems that continue to ignore or misunderstand the dynamics of domestic abuse as they impact child safety and well-being. Algorithms often fail to account for how perpetrators as parents’ behaviors intersect with other risk factors like mental health, substance use, housing instability, and childhood injury, death, and trauma. Instead of elevating the professional judgment of practitioners or the lived reality of survivors, these systems often outsource decision-making to predictive models that are both flawed and opaque.
Survivor Strengths Hidden in Plain Sight
In data science, there is a concept called BIBO or “bias in, bias out.” Like its sibling concept, GIGO, which is “garbage in, garbage out,” it is shorthand for how bias that shapes the data that a system is trained on will produce bias in the results. The failure is especially stark in how predictive tools address—or fail to address—survivor protective behavior. Domestic abuse survivors are often under extraordinary stress: navigating threats, controlling behaviors, physical and emotional harm, and the fear of losing their children if they disclose abuse. Yet many demonstrate tremendous parenting strengths, including emotional support, safety planning, maintaining routines, and shielding children from the worst of the abuse.
Predictive models often don’t see this. Because most risk algorithms operate on historical system involvement, they treat previous calls to child protective services, emergency room visits, or mental health referrals as risk factors. These same data points, when interpreted through a domestic abuse–informed lens, might indicate a survivor trying to access support and safety—or may reveal her perpetrators’ pattern of harassment via systems manipulation.
Elevating Survivor Strengths, Not Penalizing Them
Many predictive models fail to capture protective efforts made by non-offending parents, especially domestic abuse survivors. Worse, these models may interpret survivors’ trauma responses, impoverishment resulting from the perpetrator’s coercion and violence, or justified systems avoidance as risk factors. This perpetuates a “failure to protect” framework that blames mothers for conditions they did not create and behaviors they did not perpetrate. It makes victims responsible for the criminal violence and danger created by their perpetrators and for the failure of systems.
A behaviorally grounded framework, by contrast, teaches and expects the documentation of survivors’ strengths. From safety planning to maintaining routines, seeking services, or emotionally supporting children under pressure, survivors demonstrate countless protective capacities. This survivor strengths approach actively resists the erasure of protective behaviors and instead builds collaborative pathways toward safety and healing.
It also helps reduce gender bias by giving mothers, usually but not always the survivor, full credit for what they do that promotes safety, stability, nurturance, and healing from trauma in the context of the perpetrator’s behaviors. Due to high, rarely articulated expectations for women as parents, these heroic efforts are often invisible as data points in practice models. The historical failure of social workers to document how a trip to the doctor for a medically fragile child took extra effort due to barriers and obstacles created by the perpetrator leads to a case file that simply reads “child medically up to date.” This is how domestic abuse–destructive or neglectful practice translates into “BIBO” data.
Correcting for Misattributed Risk Factors
This approach also corrects for the systemic failure to understand how perpetrators’ behaviors drive many of the other survivor-focused risk factors flagged in traditional assessments: substance use, mental health deterioration, homelessness, and even criminal legal system involvement. Rather than treating these as isolated or inherent characteristics of families, a behaviorally grounded framework trains practitioners to see the intersections and the context—how perpetrators’ patterns of coercive control cause, contribute to, and compound, a survivor’s life challenges.
Making the Invisible Visible
Instead of centering decision-making on statistical correlations from system records, this framework centers on behavior:
What has the protective parent done to promote the child’s safety, stability, nurturance, and healing from trauma?
How does an understanding of the perpetrator’s pattern help us see these actions in context?
How can we use our understanding of gendered expectations of parents to ensure that mothers who are survivors are given full credit for their protective actions?
Have we fully understood and documented a survivor’s efforts to repair the harm to the children of the perpetrator’s behaviors?
How have systems helped or hindered those efforts?
The approach provides a clear, evidence-informed structure for decision-making that values behavioral context, culture, and lived experience. It promotes child-centered, survivor-engaged, and perpetrator-focused assessment. And it invites practitioners to challenge their own assumptions while grounding decisions in specific, documentable behaviors.
Why This Shift Matters Now
In an era of rapid technological adoption, child welfare systems face an urgent decision: Do we double down on behaviorally decontextualized predictions based on risk assessments and historical system engagement? Or do we invest in practitioner development, system accountability, and domestic abuse–informed frameworks that document and understand the actual behaviors and dynamics driving danger and harm for children and families?
A behavior-based approach offers a path forward. It supports consistency not through mechanized yet flawed and incomplete data logic but through structured, reflective, and transparent behavioral information gathering which reflects the actual behaviors of the parent causing harm and danger. It empowers professionals with concrete knowledge and strategies and supports professional decision-making skills while validating and supporting survivors in their protective efforts. And at a time when practitioners are talking more and more about the moral harm created by the pressure to hold survivors accountable for the behaviors of perpetrators, it strengthens the entire system by aligning values with practice.
