New Analyses

The theory in How to Predict Bad Cops in Chicago suggests that if we can identify the repeaters and devise intervention mechanism then the number of complaints will be reduced because the majority of the complaints come from a small fraction of police officers. Our theme is to challenge the assumption that the number of allegations is a reliable metric to flag a police officer as “bad apple”. Our purpose is to remind the readers that we are not in the shoes of the police officers, that they sometimes risk their lives performing duties that a less courageous one would shy away from and that it would require extreme caution to judge them and their work based on numbers from a spreadsheet, otherwise we do not pay our due respect. The theory that advocates intervention on repeaters made another assumption that the personal traits of the police officer are the driving cause for a complaint, not the situation and environment. This assumption is problematic because personality can change based on past experience. For example, officers who are placed in a crime-prone neighborhood can become aggressive and violent not necessarily because of their wish, but the intense environment. Even if these repeaters are removed, it is not guaranteed that other police officers will not become new repeaters once given the old repeaters’ tasks. Another reason we think only using number of complaints to judge an officer is problematic because we don’t know the intention for one to file a complaint; ie. we don’t know whether they file a complaint just to reflect the truth, or serving their own benefit like a city payout, or just retaliating the officer who make them unhappy. On some occasions, the complaint is about an officer accidentally drops her taser; on other occasions, the complaint is about subject feelings such as “handcuffs are too tight”. We believe these complaints do not amount to the “badness” of a police officer. However, our purpose is not to say that every cop is innocent of misconduct and that there is no room for improvement for the law enforcement agency. We aim to explore a collection of attributes associated with a police officer rather than focusing on a single one. In our work, we examined attributes such as awards, number of allegations, settlements that can possibly be used to evaluate the professionalism and integrity of an officer.

In 2014, we established the legal principle in Illinois that records that relate to complaints of police misconduct belong to the public. As a result, police misconduct complaints, including identifying information about the involved police officers, demographic information about officers, complainants, & witnesses, the investigations, outcomes, and any resulting discipline are available to the public. This principle holds irrespective of whether the City found that its officers had or had not engaged in misconduct. In November 2015, we launched the Citizens Police Data Project, which operationalized this principle making this information widely available to the public. The City/CPD fought vigorously against transparency. They predicted that the sky would fall if the public were given access to information about police misconduct complaints. The City’s and police unions’ predicted parade of horrible included that (1) people would use this information to target police officers for violence, (2) violence and assaults against the police would increase, (3) officers would be deterred from investigating crime, (4) crime & violence would go up, and (5) false complaints against officers would rise. We have now had a 4 year experiment in Chicago since this information has been widely available to assess whether the CPD’s dire warnings have come true. At the same time, we can also assess positive developments in policing and police accountability since police misconduct records have become public. While we may not seek to establish a causal relationship between transparency and a number of positive developments, it would be great to document some of the encouraging things that have transpired in the wake of this unprecedented transparency. It would be great to put some solid annual numbers on these claims. For example, I believe that: • There have been fewer negative police/citizen encounters • Stops are down (street but not traffic) (may want to exclude this because it is more likely tied to the CPD agreement to report stop-and-frisks for external review) • Uses of force are down • Police shootings are down • Police misconduct complaints are down overall • Brutality (force) complaints are down • Arrests are down • Complaint sustained rates are up (greater accountability) • Higher quality investigations into police misconduct • Release of Laquan McDonald video pursuant to this principle led to the first murder prosecution & conviction of on-duty CPO for killing an African-American. • Lots of positive stories about how people have used this information (academic research to produce knowledge; investigative journalism; improved public discourse; reforms to police policy & practice; USDOJ pattern & practice investigation; IL Torture Commission to investigate claims of CPD torture; people abused by police have used this information to defend against false arrests, support affirmative civil rights claims, & free people who have been wrongly convicted as a result of police abuse. It would be great to highlight a few of these powerful stories. At the same time, it is important to examine what hasn’t happened. • To my knowledge, there has not been a single instance in which a police officer has been targeted for violence as a result of the release of this information (Jason Van Dyke stuff is independent & related to his criminal trial) • The number of violent attacks against police has decreased • Officer safety has increased—fewer killings of police officers • Violent crime has not increased. Indeed, apart from the 2016 blip (Laquan McDonald code of silence effect), violent crime in Chicago has steadily decreased since then. Again, it would be great to attach numbers to this side as well.

Focusing on a subset of officers with increasing allegations year to year, preferably with few or no allegations at the start of the complete dataset (2000). From there exploring available generalized events in an officer's career, environment, allegation type. The goal would be to find key factors of early allegations, or the environment surrounding the allegations, and relate them to a path towards repeated behavior. Specifically focusing on relationships in these officers to types of allegations and increased severity of the allegations. Examples of possible areas to explore are: Age of officer/time on the force Types of allegations over time Awards Major media events Settlements, punishment, dismissals or previous allegations Arrests

Our project will explore the influence that existing or promoted commanding officers have on the teams and individuals they lead. We will investigate this relationship by looking at complaint histories and percentiles of both the commanding officers and their teams before and after the commanding officer has been instated. We will discern if the teams’ misconduct goes up in any predictable way as well as compare this data to department averages, or to commanding officers who do not have a history of complaints, and the records of the teams they are managing. We also seek to determine if a supervisor is “lenient” on officer misconduct punishment, and if that perpetuates continued misconduct. From our analysis, we will be able to predict whether an officer who is promoted to a supervisor position will have an overall positive or negative effect on the department. Considering the allegation settlements, we can learn how these supervisors and police units financially impact the city.

Our initial research has brought our group to investigate which factors are the greatest determinants of the level of force used by police officers during incidents. Once arriving at a set of variables we would like to understand the narrative behind why changes in certain factors (for example: demographics/area/time_of_day/number_of_officers/type_of_resistance) influence the use of force. In addition, a case study of several officers will be used in order to analyze whether the same officer is meting out the same level of force regardless of changing environment or whether the same factors applicable to groups of police officers are influencing individual officer decisions. Research Questions 1. Relational Analytics a. What is the breakdown into categories of force used as a percentage of the total incidents? b. Demographics shares of incidents involving the use of force in comparison to the demographics of the beat/area/neighborhood where incident occurred. c. What are the possible values that can be used to quantify force within the dataset i. Which action categories/force types are we considering at what “level” d. What geographic group are most affected by a given type of action? e. What are the most common type of resistance levels among incidents of use-of-force? 2. Visualization a. A highlight table can be used to illustrate the high impact one factor can have on the level of force used regardless of changes in other variables such as geographic location b. A scatterplot can be used to show our analysis if our findings reveal two continuous variables of interest being correlated i.e. number of commendations/medals is directly correlated with a high proportion of an officer’s misconduct allegations being related to use of excessive force. c. A circle view chart can also be used for comparative analysis of the different variables of interest and how they underlie the number of use of force incidents. d. One interactive visual we can create is similar to the NCAA bracket predictor tagged below with the difference being that instead of each node showing chance of winning the tournament broken down by team, it would instead show chance of variable being present or co-present with others in a use-of-force incident i. https://thepowerrank.com/visual/NCAA_Tournament_Predictions 3. Data Cleaning and Integration a. Does the number of allegations filed previously against an officer (or group of officers) affect the monetary compensation of a settlement b. Do previous uses of force by an officer (or group) affect the monetary compensation of a settlement c. What percentage of law enforcement officers who are named in settlements have been named in previous settlements d. What percentage of settlements are early in an officer’s career? 4. Graph Analytics a. Are officers who are accused of using force alone more likely to reoffend than officers accused with other officers? b. If groups are more likely, is there a group size that leads to the largest proportion of officers reoffending? 5. Machine Learning and Text Analytics a. Can we predict how aggressive an officer (what level of force) will act following an allegation that was sustained? i. Specifically, to answer this question, we will look at the force_type used by officers following an allegation that was sustained. We will take into consideration to make sure the type of charge and resistance_subjects/resistance_levels are similar before and after a sustained allegation to draw a predictive analysis. b. Can we predict if a use-of-force incident is likely to happen? i. For this, we would like to see if we can come up with a decision-tree like system comparing all the attributes in the trr_data to create a predictive analysis for use-of-force incidents. c. Are certain complaint report tags correlated to relationships between officer and complainant demographics? d. We would like to cluster levels-of-force tags based on how common they are used with police officers with a complaints in the 99th percentile.

Our group hopes to investigate the relationship between the diversity of police units and the frequency and severity of misconduct complaints. We expect that partnerships with officers of different races and/or genders would have fewer complaints lodged against them. We plan to use demographic data from complaints naming more than one officer and compare the demographics of these co-accused officers with the broader demographics of the entire department. Our hypothesis is that groups of racially and/or gender homogenous officers are more likely to have complaints filed against them, especially by individuals from outside of their racial or gender group. It is common knowledge that diverse teams tend to outperform their homogenous counterparts, and it makes sense that this phenomenon would be particularly apparent in the high-stress type of interpersonal conflict resolution that defines police work. An investigation into this phenomenon is possible due to the data’s inclusion of both complainant and officer demographic data. Additionally, we would like to explore the impact that a complaint investigator’s demographic background might have on an investigation’s outcome since this data is also included in the CPDP data. A final phenomenon we want to investigate is whether members of the same unit or task force behave differently when working with their groups. While this phenomenon will likely not depend on demographic biases as strongly as the others we want to investigate, it is a potential source of in-group / out-group bias worth investigating all the same.

Our group is interested in exploring the systemic patterns within the CPD which lead to violence and higher officer allegations. Are we able to use data to understand patterns and trends in how officers begin to exhibit violent behavior and how the CPD responds to such behavior? See our proposal in the following link: https://docs.google.com/document/d/1N39nfzxgB22sVge_6MxeSYquMK5RxtbUz4zECtxNs0k/edit

Our project team will explore a change in misconduct rate of each department over time and also identifying key players in the force responsible for the improvement/deterioration of a department. Our goal is to research the relationship between "a change in misconduct rate of each department" and "reward/promotion of the officers in the same department" to understand how police officers can be/are placed, promoted and rewarded in police organizations effectively. We would also like to further analyze the departmental relationships of these cops to that of a network of bad cops to help us identify crucial players and relationships across these hierarchies.

According to the Chicago Police Data Project, a large number of complaints go without further discipline or proper investigation in the Chicago police force. It is very common for leaders, such as the police supervisors or even the superintendent, to overlook the complaint, and automatically assume the side of the officer. Even the new COPA, formed to solve this issue, struggles with the political motivations of the staff, as noted by the Chicago Police Files article. Therefore, we are curious as to which allegations are investigated and for which disciplinary actions are taken; are they simply more serious, or does there exist a correlation between sustained allegations and particular demographics? One driving factor for this was the recent traffic arrest of an African American Northwestern PhD student, which gained a lot of media attention. Is it possible that, in most cases, such a scenario would have been ignored with no disciplinary action taken had the individual not been a Northwestern student (i.e. someone with “higher” social standing)? We seek to explore the effect of demographics such as race, economic status, and geographic location on the likelihood of a complaint to receive disciplinary action.

Our group attempts to investigate how pressure stemming from society could play a role in the outcomes of police misconduct investigations. We look at this phenomenon from two complementary viewpoints: from the cases that are picked up by the press (both local and national) and from the social media landscape represented here by Twitter. Research Question: When a case of police misconduct causes indignation and broad discussion in society, how does it influence the case's investigation? This single question unfolds over a series of hypothetical, measurable effects on the investigation (part I of section "Feature Space" below); and we can measure pressure exerted by society using a number of digital channels (part II of section "Feature Space" below). Feature Space: We hypothesize that, to some extent, the following effects could happen: Effect 1: It could expedite a disciplinary investigation; Effect 2: Increase the likelihood of the officer(s) involved being disciplined; Effect 3: Increase the likelihood of the state granting compensation. And these effects could depend on the following features: Feature 1: The number of news stories reporting on the case; Feature 2: The involvement of certain media outlets (local vs national ones); Feature 3: The length of the news burst (from the first story to the last); Feature 4: Social engagement with the case as measured in Twitter. Methodology: Our proposed methodology comprises the following steps: Step 1. Manually analyze a few cases on Google: Try to use some allegation fields to search for stories in Google and map the relevant news sources (both local and national); Step 2. Manually analyze a few cases on Twitter: Try to use some allegation fields to search for tweets and map the Twitter handles that can be considered sources or good proxies for public sentiment; Step 3. Collect more data points: Use Python, the Newspaper library [1], and Twitter API [2] to automate the process in Steps 1 and 2 in order to collect a larger sample; Step 4. Run predictive analysis [3]: Use the features generated by Step 3 to predict the target features (Effects 1, 2, and 3 can be obtained directly from the Invisible Institute's Citizens Police Data Project). [1] The Newspaper Library for Python — https://newspaper.readthedocs.io/en/latest/ [2] Twitter API — https://developer.twitter.com/en/docs/developer-utilities/twitter-libraries [3] Hall, Mark, et al. "The WEKA data mining software: an update." ACM SIGKDD explorations newsletter 11.1 (2009): 10-18.