Strengthening Data Science Methods for Department of Defense ...

Overview: Strengthening Data Science Methods for DOD P&R Privacy and Employee Data Workshop June 6, 2017 Sallie Keller Professor of Statistics & Director Committee Members

STEPHEN M. ROBINSON, Chair, NAE, University of Wisconsin-Madison PASCALE CARAYON, University of Wisconsin-Madison DAVID CHU, Institute for Defense Analyses CYNTHIA DWORK, NAS/NAE, Microsoft Research TERRY P. HARRISON, Pennsylvania State University ALAN F. KARR, RTI International SALLIE KELLER, Virginia Polytechnic and State University ALAIR MACLEAN, Washington State University, Vancouver DAVID MAIER, Portland State University STEPHEN M. POLLOCK, NAE, University of Michigan (until June 2015)

PAUL R. SACKETT, University of Minnesota MARK S. SQUILLANTE, IBM Research WILLIAM J. STRICKLAND, HumRRO STEVEN TADELIS, University of California, Berkeley Study Statement of Task Charge: Develop a roadmap and implementation plan for integrating data analytics in support of decisions within the purview of the Office of the Under Secretary of Defense (Personnel & Readiness). Interpretation: The committee interpreted the charge as calling for highlevel conceptual advice, acknowledging P&R must have the latitude to craft specific plans for strengthening its capabilities in, and use of, data science.

Principle outcomes for which P&R is responsible Ensuring DoD can recruit, train, motivate, and retain the necessary numbers of qualified personnel Creating incentives that guide the department to an optimal mix of personnel Ensuring DoD creates a force that is ready to carry out directed actions Influencing DoDs decisions that affect the shape of military careers Ensuring the services provided to support DoD personnel and their dependents are properly structured. Anticipating and responding to sensitive behavioral issues

P&R Policy and Oversight Areas Staffing decisions for both service members and civil service employees Recruiting standards for both service members and civil service

employees Selection criteria for military recruits Selection criteria for civil service positions Job assignment Compensation standards for service members Training and education programs for personnel Promotion criteria for service members and civil service Access to DoD buildings and locations Process for and time of transitions out of the armed forces;

Provisions for mental and physical health care to service members, civil service employees, retirees, and dependents; Suicide prevention among service members Provisions for the needs of families Responses to problematic behavior Provisions for health and retirement benfits to separated service members Responses to congressional requests for information The Data Need to open the data aperture to enable more

research "Data need to be easily accessible and shared across groups in a way that reduces the hurdles currently faced when researchers and analysts seek to find or share data while ensuring proper privacy and security protections." "Analytic methods available to P&R need to be expanded to enable stronger and more rapid responses to significant P&R research and analysis questions." Person-Event Data Environment Relevant findings for today's discussions Finding: Analyses developed to support the Secretary of Defense are often disjoint, one-off activities developed to be responsive to immediate questions and may lack a continuity plan for future use of data or analytic methods. Finding: The reuse of operational data for analytic purposes can expose issues in data collection, recording, transmission, cleaning, coding, and loading. Problems are

often not detected until the point of analysis when anomalies crop up in results. Finding: Enhanced data sharing within DoD, across agencies, and with the research community could promote the creation of new statistical methods, tools, and products. 8 Recommendations around privacy and data governance Recommendation: DoD should consider adopting or adapting the privacy and governance structure developed by the Office of Management and Budget for civilian statistical agencies. In particular, DoD should follow the guidance on use of administrative records and establishing of statistical units under the Confidential Information Protection and Statistical Efficiency Act (CIPSEA) for both military and civil service personnel. In doing so, DoD should examine the applicability of Fair Information Practice Principles in the treatment of DMDC data. Recommendation: In its role as steward of the Person-Event Data Environment, the Defense Manpower Data Center should consider ways to adapt and use privacy and governance

practices that the Office of Management and Budget has created for civilian use. Recommendation: DoD should carry out research on the feasibility of differential privacy methods for its personnel analytics. Adopt a paradigm based on trust, policy, and governance? Data governance versus data access Policy analytics versus use for administration Control use versus controlling data access Trusted data commons versus data enclaves