See sooner
Move from reaction to anticipation.
Use responsible, unit-level evidence to help leaders recognize patterns they might otherwise miss.
AI-assisted predictive analytics
Prevent harm through human care.
MP2 is a community of researchers, prevention practitioners, clinicians, and military leaders developing and testing responsible, AI-assisted, unit-level predictive analytics. The goal is to recognize changing organizational conditions, focus support sooner, and learn what works.
Human-led by design. AI helps surface patterns; people provide context, judgment, and care. MP2 does not score individuals or automate decisions.
An independent military prevention research collaborative.
Protect people
Strengthen teams
Support leaders
Advance prevention
01 The mission
Military leaders across the joint force are often asked to respond after a harmful event has already occurred. Prevention starts earlier: recognizing changing conditions inside units, listening closely, and connecting people to the right support at the right time.
Our purpose: connect military leaders, researchers, clinicians, prevention practitioners, and data experts to generate, test, preserve, and translate evidence that helps military organizations prevent suicide and other harmful behaviors before crises occur.
The technology can surface a signal. People must supply context, care, accountability, and judgment.
See sooner
Use responsible, unit-level evidence to help leaders recognize patterns they might otherwise miss.
Act wisely
Pair analysis with conversations, local knowledge, and professional expertise before choosing a response.
Learn continuously
Study what helps, what does not, and what causes unintended harm—then share those lessons with the field.
02 How we work
MP2 connects researchers, practitioners, clinicians, data experts, and military leaders as a learning community—rigorous enough to test new methods, practical enough to improve decisions, and accountable to the people it exists to protect.
01
Bring together military leaders, practitioners, researchers, clinicians, data stewards, and people with lived experience around shared prevention problems.
02
Sponsor rigorous pilots, replications, and evaluations that examine outcomes, implementation, and unintended effects.
03
Keep useful measures, methods, code, and findings accessible so hard-won learning survives turnover and can be tested again.
04
Translate credible research into decision briefs, convenings, and tools that help leaders focus attention and support responsibly.
03 Prevention in practice
MP2 grows from years of field research on harmful behaviors, organizational culture, readiness, and human performance. Across that work, one lesson is consistent: data should inform leadership, never replace it.
Unit-level warning signals can help leaders direct scarce time and attention. But a responsible prevention system also requires diagnosis, tailored support, follow-through, and honest evaluation.
04 Research principles
Credible prevention demands rigorous methods, explicit limits, careful stewardship, and respect for every person represented in the evidence.
Study modifiable organizational conditions and protective factors—not who an algorithm claims may cause harm.
Use unit-level or deidentified information by default. Never build individual risk scores for adverse personnel action.
Collect and retain only what a responsible research or prevention purpose truly requires.
Share uncertainty, null results, model drift, implementation failure, and credible alternative explanations.
Protect research questions and findings from sponsor, donor, vendor, or command pressure.
Put leaders, practitioners, and people with lived experience at the center of interpretation and action.
05 Selected research
MP2’s agenda is joint: develop evidence and methods that can be tested responsibly across every military service. These public case studies retain the context in which the research began.
The clearest statement of the prevention mission: use unit-level evidence to help leaders intervene earlier while preserving human judgment.
Read the articleThe predictive base model for understanding short-term unit-level incident patterns—not an individual risk score.
Read the paperA field-tested method for turning operational questions and messy data into decision-relevant insight.
Read the articleEvidence on how culture, inspiration, and leader behavior relate to cohesion, performance, and harmful behaviors.
Read the articleA case for using human-performance data carefully to inform training, recovery, and leader decisions.
Read the article