By: Jaspreet Momi
Published on: April 14, 2025
Predictive policing is the practice of policing locations according to artificial intelligence (“AI”)-generated forecasts of when and where crime incidents are likely to occur.[1] Location-based predictive policing software like Geolitica, established in 2012 as PredPol, examines an input of up to ten years of recorded crime data, including dates, times, locations, and types of crimes. While location-based predictive policing software was not originally designed to include information about specific individuals in its input data,[3] it can integrate such personal data into its analysis.[4] Analyzing the input data, predictive policing software attempts to predict, over a twelve-hour period, the location and timing of future crimes by marking “hotspot” locations on a map, each of which represents a geographical area deemed high risk.[5]
Several police departments nationwide formerly used and implemented software like Geolitica in their policing systems. For example, the Los Angeles Police Department (“LAPD”) adopted the use of Geolitica in the 2000’s.[6] Similarly, jurisdictions in Seattle, Atlanta, and New Jersey employed Geolitica or the similar predictive policing software, Palantir, in their policing systems.[7] By 2018, more than fifty police jurisdictions were using location-based predictive policing software.[8]
While the rise in popularity of location-based predictive policing software was swift, so was its downfall. Reports in 2018 and 2019 revealed findings that predictive policing at best produces “small, incremental” improvements in crime prediction in comparison to traditional hot-spot mapping.[9] Additionally, predictive policing programs received criticism that they would lead to aggressive over policing of select communities, notably communities of color.[10] As a result, various police departments dropped their use of predictive policing software.[11]
While location-based predictive policing software seemingly fell out of favor, predictive policing software company SoundThinking appears to be the exception. Formerly known as ShotSpotter,[12] SoundThinking was created in 1998 to detect and inform police of recorded instances of gunshots using planted microphones in various locations of a policed jurisdiction.[13] Despite heavy criticism due to ShotSpotter’s inaccuracies,[14] SoundThinking is currently still using ShotSpotter in conjunction with its location-based predictive policing software, ResourceReuters.[15] ResourceReuters uses recorded crime patterns along with “objective non-crime data” to identify specific types of terrain, such as vacant lots, laundromats, etc., that are deemed at higher risk for crime.[16]
In 2023, SoundThinking announced its acquisition of Geolitica.[17] Since this acquisition, SoundThinking integrated Geolitica’s software with its ShotSpotter and ResourceReuters systems.[18] In 2024, SoundThinking introduced its Crime Tracer person-based predictive policing software, which adapted its preexisting systems to integrate inputs such as personal information for the purpose of “detect[ing] relationships between people, places, and things” to “accelerat[e] investigations.”[19] Currently, SoundThinking’s integrated systems are in operation in more than 250 policing jurisdictions, overtaking Geolitica’s prior success as the leader in predictive policing systems.[20]
SoundThinking’s integrated software systems pose Fourth Amendment concerns, particularly with regards to the use of its hotspot reports as a basis for Terry stop and frisks.[21] SoundThinking uses both personal and location-based data inputs to create predictive hotspot reports.[22] The question of whether the combination of personal and location-based data may be used to substantiate the requisite reasonable suspicion for a Terry stop remains largely unanswered by courts.[23]
While there is no national legal standard governing the use of location-based predictive policing technology (“PPT”) to find reasonable suspicion for a Terry stop, the Fourth Circuit disfavored such a use in United States v. Curry.[24] In Curry, the court maintained that a person’s presence in a PPT hotspot cannot alone create reasonable suspicion.[25] In his concurrence, Judge Gregory cautioned against such use of PPT, concluding that while PPT may be an effective tool for law enforcement, it could cause over policing, which would in turn create greater tension between police performing their duties and the communities being policed.[26] Others noted that reliance on PPT may have a detrimental impact on the Fourth Amendment rights of people living in high crime areas, perpetuating racial bias and profiling within the criminal justice system.[27]
As SoundThinking continues to be used in hundreds of jurisdictions to predict hotspot locations with increasingly more data inputs, some of which are person-based, the questions around predictive policing’s constitutionality become increasingly more contentious. This is particularly true with regards to the Terry stop “reasonable suspicion” test. In particular, it remains unclear whether hotspot reports integrating both location and person-based inputs, which may be more accurate than just location-based inputs, are sufficient to establish reasonable suspicion under the Fourth Amendment. While the Fourth Circuit is the only appellate court to address the role of PPT in establishing reasonable suspicion under the Fourth Amendment, the increasing use of software like SoundThinking as well as its increased sophistication, data inputs, and applications in policing warrants more legal development on this issue.
[1] Aaron Sankin & Surya Mattu, Predictive Policing Software Terrible At Predicting Crimes, The Markup (Oct. 2, 2023, 10:00 AM), https://themarkup.org/prediction-bias/2023/10/02/predictive-policing-software-terrible-at-predicting-crimes; Maha Ahmed, Aided by Palantir, the LAPD Uses Predictive Policing to Monitor Specific People and Neighborhoods, The Intercept (May 11, 2018, 9:15 AM), https://theintercept.com/2018/05/11/predictive-policing-surveillance-los-angeles/.
[2] L.A. Police Comm’n, Review of Selected Los Angeles Police Department Data-Driven Policing Strategies (2019).
[3] Id.
[4] See Chandler Harris, Richmond, Virginia, Police Department Helps Lower Crime Rates with Crime Prediction Software, Gov’t Tech. (July 27, 2010), https://www.govtech.com/public-safety/richmond-virginia-police-department-helps-lower.html (pointing out that personal information such as suspect identification may be used to generate hotspot reports).
[5] Id.; L.A. Police Comm’n supra note 2.
[6] Eva Ruth Moravec, Do Algorithms Have a Place in Policing?, The Atl. (Sept. 5, 2019), https://www.theatlantic.com/politics/archive/2019/09/do-algorithms-have-place-policing/596851/.
[7] Id.; Sankin & Mattu, supra note 1.
[8] Mark Smith, Can We Predict When and Where a Crime Will Take Place?, BBC (Oct. 29, 2018), https://www.bbc.com/news/business-46017239.
[9] Id.; see Tim Lau, Predictive Policing Explained, Brennan Cent. for Just. (Apr. 1, 2020), https://www.brennancenter.org/our-work/research-reports/predictive-policing-explained (describing reports that found reliance on false data as inputs for predictive policing hotspots).
[10] Moravec, supra note 6.
[11] See id. (noting the cancellation of Geolitica and Palantir contracts by police departments in Arizona, New Orleans, New Jersey, and Santa Cruz); Sankin & Mattu, supra note 1 (“[The New Jersey Police Department] wanted to be more effective when it came to reducing crime. And having a prediction where we should be would help us to do that. I don’t know that [Geolitica] did that.”)
[12] Sankin & Mattu, supra note 1.
[13] Ralph Clark, ShotSpotter is Now SoundThinking, SoundThinking (Apr. 10, 2023), https://www.soundthinking.com/blog/shotspotter-is-now-soundthinking/.
[14] See Jay Stanley, Four Problems with the ShotSpotter Gunshot Detection System, ACLU (Aug. 24, 2021), https://www.aclu.org/news/privacy-technology/four-problems-with-the-shotspotter-gunshot-detection-system (pointing out reported inaccuracies in ShotSpotter’s system).
[15] Anthony Nunley, Does My Agency Need Patrol Software?, SoundThinking (Oct. 4, 2023), https://www.soundthinking.com/blog/does-my-agency-need-police-patrol-software/?utm_term=public%20safety%20software&utm_campaign=Search+%7C+BOF&utm_source=adwords&utm_medium=ppc&hsa_acc=8557512895&hsa_cam=20488582296&hsa_grp=154527773996&hsa_ad=676743695209&hsa_src=s&hsa_tgt=kwd-1064804015913&hsa_kw=public%20safety%20software&hsa_mt=b&hsa_net=adwords&hsa_ver=3&gad_source=5&gclid=EAIaIQobChMImZfQh9OphQMVqG1HAR0f-gKbEAAYAyAAEgJdm_D_BwE.
[16] Id.
[17] SoundThinking Annual Report 2023 (Apr. 29, 2024) [hereinafter Annual Report].
[18] See Sankin & Mattu, supra note 1 (pointing out that Geolitica acquired some of Geolitica’s intellectual property and transitioned Geolitica’s customers to SoundThinking); see generally Law Enforcement, SoundThinking (2025), https://www.soundthinking.com/law-enforcement/ [hereinafter Law Enforcement, SoundThinking] (describing the various predictive policing software that SoundThinking offers).
[19] Law Enforcement, SoundThinking, supra note 18; Annual Report, supra note 17 (“[Crime Tracer] enables investigators to search through more than 1 billion criminal justice records across jurisdictions to generate tactical leads and quickly make intelligent connections.”).
[20] Law Enforcement, SoundThinking, supra note 18; Annual Report, supra note 17.
[21] See generally United States v. Curry, 965 F.3d 313, 326 (4th Cir. 2020) (discussing the use of predictive policing software to establish reasonable suspicion); Terry v. Ohio, 392 U.S. 1 (1968) (establishing the requisite reasonable suspicion for stop and frisks).
[22] See Law Enforcement, SoundThinking, supra note 18 (describing the use of predictive software to “detect[] relationships between people, places, and things.”).
[23] See Lau, supra note 9 (emphasizing that legal experts continue to argue questions of constitutionality around the issue of predictive policing and the Fourth Amendment).
[24] Curry, 965 F.3d at 326.
[25] Id.
[26] Id. at 332 (Gregory, J., concurring).
[27] Id. at 344 (Thacker, J., concurring).
Pls link the source (just so i have it and can verify – thank you!!)
https://www.oig.lacity.org/_files/ugd/b2dd23_21f6fe20f1b84c179abf440d4c049219.pdf