Crowd-sourced Policing Program SUBMITTED BY ENODO Global, Inc. Name: Drew O Sullivan Email: drew@enodoglobal.com Phone: +1 (703) 508-0321 Attn: This white paper includes data that shall not be duplicated, used, or disclosed -- in whole or in part -- for any purpose other than to evaluate the content. If, however, a contract is awarded to this offer or as a result of -- or in connection with -- the submission of this data, the client shall have the right to duplicate, use, or disclose the data to the extent provided in the resulting contract. This restriction does not limit the client s right to use information contained in this data if it is obtained from another source without restriction. 1
Contents 1 Background... 3 2... 4 2.1 Project Description... 4 2.2 Phase I: Assessment (6 months)... 5 2.2.1 Social Media Analysis... 5 2.2.2 Social Risk Analysis... 6 2.2.3 Social Risk Action Plan... 6 2.2.4 Phase I Deliverables... 7 2.3 Phase II: Implementation (24 months)... 8 2.3.1 ipeople... 9 2.3.2 Phase II Deliverables... 10 3 Conclusion... 11 4 Appendix A... 12 4.1 Methodology... 12 4.2 Technology... 13 Data Collection... 14 Specialized Technologies... 14 Topic Modeling... 14 Sentiment Analysis... 15 Demographic Analysis... 15 Narrative Analysis... 15 Social Network Analysis... 15 2
ENODO Global applies its population-centric methodology, advanced data analytics, and Social Listening tools to repair community-police relations, enhance policing techniques, and reduce homicides and violent crime. 1 Background From Oakland to Baltimore, violent crime and homicide rates continue to plague cities across the United States. Failures to address public safety have prolonged community suffering and reinforced harmful stereotypes of both police and citizens. Public officials have attempted to solve these problems through investments in high-tech projects like Chicago s ShotSpotter System or psychological programs like Newark s feel-good Trauma Training. Policing however, remains an inherently human problem that cannot be addressed through technology alone. Social media amplifies unaddressed social tensions and has changed the nature of policing itself. Platforms like Facebook, Twitter, and Instagram enable people to express their views online, which highlight ineffective policing programs and fuel social movements like #BlackLivesMatter and #HandsUpDontShoot. At the same time, recurring reports and videos of aggressive police tactics, often attributed to racial bias, deepen the discord with underprivileged communities. Shootings by law enforcement lead to widespread protests, and tensions rise further when police respond to protesters with ambivalence or aggression. Police forces are increasingly being pushed onto the defensive due to heightened public scrutiny, as they struggle to do their jobs in the new information environment. Meanwhile, community grievances go unaddressed, citizens are outraged that public funds continue to be applied toward seemingly ineffective initiatives, and communities continue to suffer. The result is a polarized society that no longer trusts the police and public officials. 3
ENODO s Crowd-sourced Policing Program provides the missing link to existing law enforcement approaches. 2 2.1 Project Description ENODO s is a state-of-the-art program based upon proven policing strategies that is designed to combat violent crime, enhance existing policing initiatives, and improve community-police relations. It combines ENODO s population-centric methodology, cutting-edge data analytics, and customized Social Listening tools to shape public perception, influence behavior, and build trust. CPP places the community at the center of the approach. It uncovers social tensions and underlying grievances, and measures citizens perceptions of key and trending topics, which are used to develop communications strategies and engagement activities that reconnect community members with public officials. It is informed by ENODO s ipeople platform, which enables citizens to participate in community-based policing initiatives. ipeople continually collects and analyzes information to ensure that police activities address citizens evolving concerns, based on changes in the environment. ENODO s CPP enables law enforcement and government officials to measure public sentiment regarding police activities in real-time, optimize current community policing techniques, and develop programs to improve socio-economic conditions. It reveals opportunities to influence behavior and restore trust at the individual and community levels. Addressing citizens' demands establishes and reinforces a more positive relationship between law enforcement and the community, enabling these two groups to solve issues together. The outcome is a reduction of violent crime in a given area by 20% in less than twenty-four months. ENODO s is divided into two phases: Assessment and Implementation. This multi-phased approach allows for a detailed evaluation of the client s specific operating environment, which is used to inform the creation of tailored analyses. Analytical products are designed to create a community-based policing strategy to adjust policing tactics with ipeople imputs, minimizing the possibility of practices being viewed as invasive or dehumanizing, which improve community-police relations and stem criminal activity. Note: ENODO produced Capability Demonstrations in Chicago, Newark, and St. Louis to illustrate our analytical rigor and insights clients can expect to receive. ENODO Global also offers online training and education courses such as Social Risk for Law Enforcement, available through the Texas A&M Extension Service (TEEX). 4
2.2 Phase I: Assessment (6 months) Figure 1. Crowd-sourced Policing Program In Phase I: Assessment, ENODO delivers a repeatable, measurable process and data-driven solutions for an individual community or area affected by violent crime. Data scientists build a customized data collection and analysis platform, unique to the community s informational, linguistic, and social environments. Analysts conduct an in-depth examination of crime related topics to pinpoint social tensions and measure public perception of existing policing techniques. Analysts continually monitor social media and other information sources, track public discourse over time, and produce a series of reports (e.g. Social Risk Analysis [SRA] and Social Media Analysis [SMA]) that enable the creation of a Social Risk Action Plan (SRAP) tailored specifically for each community s unique challenges. The SRAP is a community-based policing strategy which includes a communications plan and engagement activities. The strategy aligns community objectives with policing techniques, delivers unique insights into community perceptions of law enforcement activities, and predictive analysis of potential social unrest. 2.2.1 Social Media Analysis ENODO creates a Social Media Analysis (SMA) within the first 45 days, which details the specific language and narratives community members use in discussions about crime related topics (e.g. petty crime, robbery, police brutality, and violent crime). The SMA uncovers the most-used online platforms within the community and identifies the key influencers (e.g., individuals and organizations) that shape these discussions. 5
2.2.2 Social Risk Analysis Social Risk Analysis: ENODO analysts simultaneously conduct population-centric analysis of community-related data, using online information to supplement existing data sources. Analysts produce a Social Risk Analysis (SRA) by day 90 which pinpoints Social Risk factors that trigger community-police conflicts, measures the impact of each factor, and creates findings and recommendations that enable stakeholders to effectively address individual Social Risk factors. 2.2.3 Social Risk Action Plan Drawing upon the SMA and SRA analyses, ENODO develops a Social Risk Action Plan (SRAP) a community-based policing strategy that includes a detailed communications plan and tailored engagement activities designed to address negative narratives about law enforcement and positively engage the affected community. ENODO creates a Master Narrative, tailored content, and deploys customized micro-targeting that resonates with high-risk communities through the most effective means (e.g. specific social media platforms, key influencers, and popular events). Throughout Phase I, analysts conduct in-depth analysis, uncover unique insights, and provide timely findings and recommendations, which enable public officials, law enforcement, and citizens to enact effective engagements that mitigate the risk of community-police conflict. Insights from the analyses may also prove useful in stakeholders efforts to curb violence in other similarly affected communities. Moreover, Phase I analysis enables the client to assess CPP capabilities and its performance before continuing onto Phase II. Figure 2 provides an illustration of the Phase I timeline and deliverables distributed to the client(s) and appropriate stakeholders. Figure 2. Phase I Timeline 6
2.2.4 Phase I Deliverables Social Media Analysis (45 days) Analysts collect, filter, and synthesize varying data sets and online social media information over a forty-five (45) day period to create a detailed analysis that: Maps the information environment for a specific geographic area Defines social media use by online platform Identifies key individuals or groups influencing online discussions Measures public sentiment (e.g., positive, negative, and neutral) Pinpoints key and trending topics discussed by citizens online Measures public officials and law enforcement s social media reach in terms of saturation and penetration (i.e. ability to engage residents through online communications) Social Risk Analysis (90 days) Analysts identify and measure different Social Risk factors and their influence over a ninety (90) day period to: Provide a Community Violence Risk rating (i.e. high, medium, or low) based on the types, magnitude, and frequency of community policing events Quantify the impact of specific Social Risk Factors in a given community (e.g., drug use, gangs, public trust, school violence) Determine community sentiment about trending topics, issues, individuals, and events Measure perception of existing community police initiatives to determine effectiveness Uncover variations in perception amongst different demographic groups (e.g., race, religion, ethnicity, gender, and age) Provide recommendations to enhance existing violence prevention initiatives Social Risk Action Plan (180 days) ENODO designs a comprehensive community-based policing strategy comprised of a proactive communications plan and tailored engagement activities based on findings and recommendations from the SMA and SRA that: Develop clear, quantifiable policing objectives based on baseline assessments, Social Risk factors, and the Community Violence Risk rating Use data-backed findings to direct resources, focus efforts, and inform targeted engagement activities that address Social Risk factors 7
Increase social media reach, penetration, and saturation to improve client s ability to connect with the community through tailored messages and key influencers Create engagement activities tailored to address community grievances and needs Positively shape sentiment, culture, and behavior to reduce crime Delivers a Master Narrative (e.g. Protect and Serve ) and sub-narratives that shape individual and community identity A policing strategy that relies on force will only fuel the law enforcementcommunity divide. 2.3 Phase II: Implementation (24 months) In Phase II: Implementation, ENODO helps the client execute the community-based policing strategy, disseminates tailored communications, and advises engagement activities. The customized data collection and analytics platform developed in Phase I is modified to conduct Active Social Media monitoring through the ipeople platform. The platform tracks public sentiment of crime related topics, monitors the impact of communication and engagement activities, and provides real-time updates to the community-based policing strategy based on changes in public perceptions of crime and the evolving social landscape. Throughout Phase II, ENODO gradually expands the complexity of analytical support and scope of the CPP. Analysts move beyond topic modeling and sentiment analysis to conduct in-depth demographic, narrative, and social network analysis (see appendix A) that enable law enforcement to develop targeted engagement activities customized for specific segments of the population. ENODO delivers daily updates, weekly reports, Emergency Alerts, and Ask ENODO s to provide key insights that are continuously informed by ipeople. Phase II activities are supported by a full-time data analyst and communications specialist who together produce targeted recommendations designed to inform policing activities and communications that maximize the impact of personnel and resources. CPP s design is predicated upon the inkblot approach that enables the program to quickly expand in scope from an assessment focused on a single community to widespread implementation of the Social Risk Action Plan across multiple communities throughout a larger geographic area. As new communities are introduced throughout Phase II, ENODO monitors public discourse and produces additional analyses to improve a department s ability to align policing initiatives and public services with community needs. 8
2.3.1 ipeople Progress is impossible without change. George Bernard Shaw ENODO s ipeople platform is an interactive, crowd-sourced Social Listening tool that creates a virtual dialogue between police and the communities they serve. It enables community members to discuss crime and crime-related topics in an open forum near real-time. It provides insights that continually update the strategy and engagement activities to ensure initiatives are aligned with the public s shifting perceptions and changes in the environment. The platform is backed by an analytical engine that is able to ingest virtually unlimited real-time structured and unstructured data from online and social media sources. The dashboard geographically organizes crime related data (e.g., burglary, homicide, violent crime, etc.) to render a geo-visualized output (e.g. heat maps) of online discussions and their associated public sentiment. ipeople allows police to pinpoint, measure, and monitor public sentiment surrounding key events and crime topics near real-time, enabling them to engage with the citizens directly. Moreover, information generated from ipeople informs predictive analysis and provides police departments with a smart policing capability to increase effectiveness of traditional policing techniques. Figure 3. ipeople Platform 9
2.3.2 Phase II Deliverables Daily Updates: Provide information to inform daily community-policing activities o One-page analyses of crime related news and discussions that have occurred within the community in the last twenty-four hours o Provides near real-time information that would otherwise go unreported Weekly Reports: Provide insights concerning key and trending topics and events, and their corresponding online narratives and sentiment o Measure changes in public sentiment due to communication and engagement activities o Provide recommendations for online and social media content and dissemination o Provide insights on key influencers and on differences in perceptions among demographic groups by conducting social network analysis and demographic analysis Social Media Messaging: Measure the performance of the communications plan over time o Create a Master Narrative (e.g. Protect and Serve), sub-narratives, and microtargeting based that align with individual and communal identity o Disseminate tailored messaging across various social media outlets based upon platform requirements (i.e. 280 character or less for Twitter) o Target different social and demographic groups through specific verbiage connected to identities Ask ENODOs: Answer a specific question related to a specific event or activity o Completed in less than one business day o Limited to one Ask ENODO per week Emergency Alerts: Near real-time notifications of significant changes in public sentiment in response to a specific event or activity (e.g. natural disaster, mass protest, school shooting) ipeople Platform o A fully functional customized data collection and analysis platform designed for the project s specific operating environment o Allow police departments and public officials to understand, measure, and monitor public perceptions of policing activities and social narratives 10
o Provide regular updates (e.g. daily, weekly, monthly, or quarterly) that track behavior and sentiment associated with critical Social Risk Factors o Identify trends and report deviations from baseline assessments o Deliver real-time indications and warnings of impending community violence (Emergency Alerts) and data-backed answers to specific questions submitted by law enforcement and public officials (Ask ENODOs) o Measure progress of the crowd-sourced policing strategy, communications plan and on-going engagement activities o Provide recommendations to modify or enhance stakeholder communication and engagement activities o Continuously update the CPP to ensure that police activities address citizens needs and evolve based on changes within the environment o Provide insight into the effectiveness of current policing tactics, techniques, and procedures TTPs o Fosters greater transparency and dialogue resulting in increased trust and improved community-police relations 3 Conclusion ENODO s delivers a community-based approach to forecast Social Risk and mitigate violent crime, aggressive police tactics, and reduce anti-law enforcement sentiment in communities across the country. It produces timely, quantifiable, data-driven solutions that enable stakeholders (e.g., local officials, law enforcement, citizens, and community groups) to understand and address the issue of community-policing through the lens of the citizens themselves. ENODO s approach delivers powerful insights by leveraging the voices of those most affected by community-policing, giving stakeholders the tools to uncover and address root causes and reduce the risk of violence between citizens and law enforcement. The ipeople platform produces valuable data analytics that are both quantifiable and measurable. The information gathered through the easily-accessible online platform can be used to improve decision-making and deliver efficient solutions, as well as inform communications and engagement activities. Guesswork is removed when executing engagement strategies, allowing local officials to manage online discourse and proactively engage key stakeholders. 11
4 Appendix A 4.1 Methodology ENODO Global s advantage over traditional consultants or data analytics firms lies in our population-centric methodology that allows us to drill down to the targeted population at risk. Our methodology is designed to systematically solve complex social problems in dynamic environments. ENODO s methodology leverages decades of international development experience, business expertise, and real-world military operations, along with social science competencies to identify and mitigate Social Risk. The methodology explains how people understand their reality and organize socially, why they support or resist movements, and how and why their identities transform over time. It examines complex social dynamics through a single variable identity. Identity is the key to unlocking beliefs, values, interests, and behaviors. It is the most fundamental human desire and is the principal source of most conflicts. Identity plays upon the norms, values, and traditions that exist within societies and is so fundamental to how people interpret themselves and the world that any threat to it produces an immediate and powerful response. By harnessing the power of identity, ENODO helps shape public sentiment, influence behavior, and facilitate inter-group negotiation and compromise. ENODO s methodology delivers a proactive approach to understand, measure, and support populations affected by community-police conflict with messages and channels with which they identify. This approach also helps shape public perceptions of local government and law enforcement regarding community policing. It uncovers the grievances, identities, key narratives, and social networks that drive the online discourse regarding attitudes towards public officials. It pinpoints the dominant influences within the community, quantifies the attitudes and beliefs that drive perceptions about policing techniques, and informs the development of effective narratives and engagement strategies that resonate with each effected community. Project findings can be used to inform strategic communications and reduce violent crime, manage community-police conflict, and increase public trust within each community. While these activities cannot remove the difficult realities facing underprivileged communities and their police departments, they can help reduce negative or false narratives and can encourage cooperative solutions to the challenges they face. 12
4.2 Technology ENODO builds customized technology platforms designed for specific language, social, and information environments to augment and accelerate our methodology. Our semi-automated social media collection and analysis platform delivers accurate, actionable insight tailored for specific topics and organizations in near real-time. We combine our proprietary populationcentric methodology that identifies underlying social dynamics with advanced computational techniques that include natural language processing (NLP), machine learning (ML) techniques, and social network analysis (SNA) to maximize the insights we extract from various data sources and quantitatively measure text-based narratives, perceptions, and sentiments. Our technology platform outperforms off-the-shelf technologies by providing increased sentiment and translation accuracy. Moreover, the platform is dynamic and evolves with the changing social and information conditions and events on the ground. Each tailored technology platform enables clients to make sense of digital information from online news outlets, social media, and proprietary datasets to identify, monitor, and forecast public sentiment and narratives to augment planning and execution within rapidly evolving social environments. Figure 4. Example Data Analytics Platform 13
We optimize open source technology with specialized tools. We use a variety of best-in-class libraries such as Google s TensorFlow for machine learning and natural language processing. These tools enable fast, accurate processing and classification of digital data. They are built on trusted and scientifically-vetted algorithms. We also use R an open source statistical platform for follow-on analysis derived from insights created by Python libraries. We augment our analysis with more specialized programs designed to perform Geographic Information System (GIS) and SNA data. Figure 2 (above) provides a flow diagram that visually illustrates the technologies used in a previous platform designed to measure public perception of corruption. Specific technologies and processes are outlined in greater detail below: Data Collection ENODO analysts will identify terms, topics, hashtags, and accounts to initiate the data collection. The platform automatically collects data on these social media information sources using a combination of public APIs and, where relevant, authorized data vendors. Specialized Technologies Natural language processing (NLP) and machine learning (ML) tools optimize the platform to the particular dialects, slang, and rhetoric used in social media. The platform is built on open-source NLP techniques that ENODO customizes using a mixture of formal and informal texts to create a linguistic model specific to the community. ENODO applies this model in conjunction with other tailored ML technologies to develop filters that measure sentiment, identify key topics of interest, and aggregate trends. Topic Modeling Analysts employ a topic model to identify and classify the most relevant and discussed topics concerning community policing. Insights from the model help understand the key issues associated with policing, the words and phrases most resonant in social media, and the ways in which these narratives change over time. The topic model can also be used in conjunction with sentiment analysis to uncover perceptions specific to particular aspects of community policing. 14
Sentiment Analysis Analysts use the customized Natural Language Processing (NLP) and Machine Learning (ML) tools to continually measure the positive versus negative sentiment of Twitter and Facebook posts associated with key issues, events, and individuals. Analysts track trends in sentiment and identify short-term changes and anomalies that may provide warnings of emerging threats and opportunities. They dissect these trends across social media platforms, hashtags, and different effected target groups to inform strategy. Demographic Analysis ENODO collaborates with clients to determine the most significant demographic groups and potential variables that influence their perceptions and behaviors. Based upon this, ENODO develops social identity filters across primary (e.g. age, ethnicity, religion, and gender) and secondary (e.g., education, family, and occupation) factors to measure the influence of individual and communal identity layers within the population. Narrative Analysis Analysts use findings from the analysis to identify a Master Narrative and sub-narratives. Narratives connect key groups and the topics they care most about through a common identity. Narrative analysis enables clients to determine where and how to engage each demographic group to increase positive perception and align with the identities of demographic groups. Analysts continuously track responses and changes in public perception to adjust messages and increase impact. Social Network Analysis As the understanding of sentiment and relevant topics emerge, analysts map out the most important communications conduits (key individuals) and patterns that drive these narratives. Social Network Analysis focuses on accounts whose messages generate the most retweets, likes, and other aggregate measures of usage or specific individuals identified as thought leaders. 15