U.S. flag

An official website of the United States government, Department of Justice.

Innovations in Measuring Community Perceptions Challenge

Opportunity ID
fy23-measuring-community-perceptions-challenge
Solicitation Status
Closed
Fiscal Year
2023
Closing Date
Posting Date
Description

NIJ has launched an open competition to develop new methods for capturing community perceptions of police and public safety that are representative, cost effective, accurate across microgeographies, and capable of being administered frequently.

A total of $175,000 in prizes is available, and will be divided between the two submission categories: 

  • Category 1 is for approaches that use a survey instrument whether the sampling design is probability, nonprobability, or some combination of the two. 
  • Category 2 is for approaches that use non-survey instrument methods to measure constructs.

Review the Challenge.

Challenge Winners By Category and Subcategory

Note that winners were not selected in all subcategories or in each place within each category.

Category 1: Survey - Probability

No prizes awarded

Category 1: Survey - Nonprobability

PrizeAmountTeam NameTeam MembersDescription
First Prize$25,000MCHawksAndrew Wheeler and Giovanni CircoThis entry proposes the use of Every Door Direct Mail address list as a sampling frame, along with an innovative use of multi-level regression with post-stratification (MRP). Postcards are mailed to addresses. The postcards have a URL and/or a QR code to allow access to a web-based survey. Entrants propose that MRP be used in conjunction with data from the American Community Survey to reweight survey estimates of average attitudes in microgeographies. This approach allows measurement of microlevel variation in perceptions.
Second Prize$10,000Micro-Community Policing PlansJacqueline B. Helfgott 

Seattle Police Department/Brian Maxey 

Denver Police Department/ Commander Jacob Herrera 

Hello Lamp Post/Isabel Loos		This entry proposes to build on the Micro-Community Policing Plans neighborhood level, nonprobability survey methodology. A key innovation is the use of Hello Lamp Post to invite members of the public to engage in text-based mobile phone chats with the intention of reaching potential respondents who might not respond to traditionally fielded (e.g., mail or web-based) surveys. Hello Lamp Post employs signs with QR codes, that are placed in various public spaces, to engage place users in chats via their mobile phones. This submission presents a novel method of exposing potential respondents to the existence of a survey with the goal of increasing participation.
Third Prize$5,000Johnston & CompanyBritnee JohnstonThis entry proposes using a geographically enabled web-based survey platform, ArcGIS Survey 123, in combination with existing email and physical address lists collected and maintained by city departments as a sampling frame. This submission represents a creative use of existing email and physical address lists to enable geographic specificity and reduce the cost of data collection.
Fourth Prize$2,500Just PoliceShichun McCammond, Sara-Laure Faraji, and Derek Michael McCammondThis entry proposes to partner with each state’s Department of Motor Vehicles (DMV) to collect survey data when residents use the site for routine tasks such as renewing vehicle registrations. Surveys would be deployed via a chat and participation would be incentivized by a reduction in the DMV vehicle registration fee. Judges appreciated the creativity of proposing a partnership with a commonly used government agency to engage potential respondents as they go about a legally required activity and enable the development of estimates of perceptions for microgeographic areas across the entirety of each state.
Fifth Prize$2,500Policing Accountability and Policy Evaluation Research LabJoshua McCrain, Kaylyn Jackson Schiff, Ian Adams, and Daniel S. SchiffThis entry proposes to use multi-level regression with post-stratification (MRP) and validates its use with a series of simulations. MRP is a robust methodology for generating small area estimations with relatively small sample sizes -- even in geographies without any survey responses.

Category 2: Data - Overall

PrizeAmountTeam NameTeam MembersDescription
First Prize$25,000Community PoliSenseAnnie Chen, Cecilia Low-Weiner, Osama Qureshi, and Michael A. Keith, Jr.This entry proposes to use data from social media (X, formerly Twitter), administrative sources (voter registration), and Census to uncover community perceptions at microgeographies. Twitter names and self-reported locations would be matched with voter registration street address data to obtain individual-level demographics. This sample of geolocated Twitter users can then be enriched with Census block group information. Tweets and shared articles are processed with text classifiers to add tags, which would be used to train another natural language processing (NLP) model to identify the sentiment of tweets and calculate user-level sentiment scores measuring perceptions for each of the five constructs. This replicable methodology represents an innovative approach to identifying and measuring community perceptions of public safety-related constructs at microgeographies.
Second Prize$10,000Albrecht & SmirnovaKat Albrecht and Inna SmirnovaThis entry proposes to use a stratified sample of city-based comments from text-based discussion platforms (Reddit) and Natural Language Processing (NLP) as a proof of concept. The proposed approach would process city-level subreddits to generate measures of the five constructs. The innovative aspect of this submission is in the use of city-level subreddits and computational methods using several types of NLP analyses to reveal community perceptions of constructs.
Third Prize$5,000ScienceCastJohn Beverly and Andrew JiranekThis entry proposes to leverage information in the Police Score Card Database (PCSD), Wikidata, and Twitter to measure community perception constructs. The PSCD is a nationwide database of policing indicators gathered from state and federal databases as well as media reports and public records requests. A framework describing relationships within these sources would be developed and used to inform the creation of a knowledge graph. The proposed method employs a variety of machine learning tools to process the data and sentiment analysis to quantify community perceptions of the constructs, highlighting novel strategies and tools for analyzing existing data regarding community perceptions.
Fourth Prize$2,500Michelle MastersMichelle MastersThis entry proposes a qualitative method called Photovoice that uses a combination of photos and text to gather data. Photovoice is often deployed in conjunction with Community Based Participatory Research (CBPR) when used at the community level. The five constructs would be used as a series of prompts. Participants would upload a photograph and descriptive text in response to each of the prompts. Microgeographies of photos could be identified by the submitter. Qualitative analysis software would be used to identify themes in the text. This entry proposes a novel design for soliciting community perceptions.

Category 2: Data - Individual - Respect

No prizes awarded

Category 2: Data - Individual - Fear of Crime

Note that two first place, and no other, prizes were awarded. 

PrizeAmountTeam NameTeam MembersDescription
First Prize$5,000The EaglesRuilin ChenThis entry proposes to use data from Airbnb property listings and customer reviews, particularly those classified under the "location" subcategory, to assess and quantify levels of fear of crime within communities. Entrants propose the application of computational methods, with a primary reliance on Large Language Models (LLMs), to extract neighborhood-specific information from the Airbnb dataset. The proposed text analysis process includes text classification, sentiment analysis, and topic modeling. Both supervised and unsupervised LLM-based text mining techniques would be used. This proposal illustrates how an existing data set can be used to measure community fear of crime.
First Prize$5,000Duddon ResearchShana M. JudgeThis entry proposes measuring fear of crime by using a proxy measure comprising certain neighborhood and housing characteristics that local governments could compile from administrative records and publicly available sources. To test the proposed approach, the entry operationalizes the fear of crime construct by using residents’ perceptions of crime in their community, as reported in responses to questions from the U.S. Census Bureau’s biennial American Housing Survey. The entry then uses penalized regression models to determine which housing-related characteristics best predict whether residents agree that their neighborhoods have high levels of serious crime. Results using all available survey years indicate that nine distinct neighborhood features can be used to accurately predict residents’ perceptions of crime. Because local governments can readily measure these characteristics, the proxy measure could be used at frequent intervals and at different levels of geography. The entry illustrates how an existing national dataset may be used in a novel way to measure community-level fear of crime.

Category 2: Data - Individual - Police Accountability

No prizes awarded

Category 2: Data - Individual - Community Policing

PrizeAmountTeam NameTeam MembersDescription
First Prize$5,000Mood mappersShafaq Chaudhry, Murat Yuksel, and Nelson RoqueThis entry proposes to adopt a citizen science methodology to measure community emotions using an application named "Mood Diary." Individuals would be asked to install the app on their smartphones. "Mood Diary" is specially designed to gather information about mood and stress levels using smartphone sensors, including accelerometers, GPS, and phone usage patterns, along with self-reported surveys. The proposal outlines methods to analyze stress data in real-time and introduces a "Mood Map" dashboard to track mood fluctuations within a city. This proposal offers the potential to enable examination of how changes in community policing strategies affect community mood. 

Awards

Number of Awards: 12
Total Amount Awarded: $102,500

Date Created: May 15, 2023