Risk Managed Demand: Proof of Concept for an Evidence-Based Framework, Operationalized to meet the Demand for Order in a Civil Society

Abstract

Is it safe to send a non-police “responder” to calls traditionally handled by police (911 in the US, 999 in the UK)? If so, what calls (categories and/or types) should be considered? How should they be identified (criteria) and with what operational procedures or systems? This study explores the viability of a diversified response (e.g., differential police response, co-response, alternative first response) to the more than 240 million calls received by 911 in the United States, annually. Observations made as early as the 1960s suggest the answer, at least with the benefit of hindsight, is unequivocally “yes.’’ The practicality of which is another matter. How and when to send a diversified response requires a systematic approach to risk management and call triage. A solution to what we term the “Ratcliffe Paradox,” whereby opportunities for a non-police response require a police response to identify them. This study leverages principles of risk management from commercial aviation and commercial technology from “intelligent call centers” to achieve three objectives, each of which is the subject of peer-reviewed publication either accepted or pending decision after having completed requested revision:

The first study analyzes three years of data from Seattle, Washington (USA) (n=727,423). Nearly half (48.9%) of the 356 types of call conventionally handled by police could be safely managed by a system of diversified response, rather than the current “all-hazards responder” approach. This Risk Managed Demand (Risk Managed Demand) framework could optimize resource allocation, potentially recovering up to 26% of police capacity expended on what are ultimately determined to be non-police calls. The second study demonstrated “proof-of-concept” that the Seattle Police Department’s newly developed Intelligent Risk Management (IRM) system and established it could safely and effectively triage (i.e., identify) opportunities for a diversified response. This system is based on a “human-in-the-loop” application of Natural Language Processing (NLP), formed using Machine Learning (ML), to effectively support the professional judgment of 911 “call takers” with a low-latency risk forecast presented as “intelligent decision support.” This system has been awarded $690,000 by the US Bureau of Justice Assistance (Award Number: 15PBJA-24-GK-05445-JAGP) to implement and study the efficacy of the system at the “primary Public Safety Answering Point” in Seattle, WA. The third and final work presented in this dissertation reports a single-case study, a multi-level reflection on the struggles of large-scale change in a modern police service. Despite the apparent opportunity for alternatives to police, this reflection finds implementing a diversified response to be deceptively challenging. Effective change requires more than an unequivocal evidence-basis. Similarly, it is not enough to passionately advocate for righteous change. Comparative analysis of two prevalent modes of change competing to evolve policing, “populist” and “bureaucratic,” suggests a systematic, incremental, and deliberative approach to public policy (i.e., bureaucratic) to be more effective and sustainable than radical reform (i.e., populist). This dissertation concludes by reflecting on its findings and their implications for the future of policing and the promise of a diversified response to public order emergencies in the United States and abroad.