Chicago Police recently released de-identified inputs and outputs of the Strategic Subject List (SSL), a model created a few years ago that assigns scores from 1-500 in attempt to identify individuals most likely to be "parties to violence"—as victims or perpetrators, with the stated intent of being a resource for social service outreach as well as investigations. Since the model is run on (what appears to be) everyone arrested in the city of Chicago in the past few years, 398,684 people have been assigned a score.
While there has been some news coverage of the SSL in the past few years, the information has changed over time and is still difficult to understand. How can we make this information more tangible and easily understood to the general public, in particular those who are most affected by it? We may just meet for tonight depending on what makes the most sense. Some initial takeaways:
Of the 398,684 people assigned a score, 287,404 (72.1%) have been assigned a score of 250 or more. For reference, a December 2016 LA Times story said: "Those people with a score in the upper 200s or higher are considered in danger of being shot or of shooting someone else."
56.25% of all Black men ages 20-29 in Chicago have scores of 250 or more (based on ACS 2015 demographics).
Only 5.9% of white men ages 20-29 in Chicago have scores of 250 or more.
Who we're looking for
Anyone concerned about the impact of scores generated by an algorithm on the relationship between law enforcement and communities. This is for thinking of the best ways to look at and communicate this information, tech-related or not.
Tools
Whatever medium is best for communicating the information clearly, whether it's writing, data visualization, or something else entirely. Additionally, any tools people feel comfortable to look at the dataset, whether that's Excel, Python, R, or the data portal itself.
About the group
Chicago Police recently released de-identified inputs and outputs of the Strategic Subject List (SSL), a model created a few years ago that assigns scores from 1-500 in attempt to identify individuals most likely to be "parties to violence"—as victims or perpetrators, with the stated intent of being a resource for social service outreach as well as investigations. Since the model is run on (what appears to be) everyone arrested in the city of Chicago in the past few years, 398,684 people have been assigned a score.
While there has been some news coverage of the SSL in the past few years, the information has changed over time and is still difficult to understand. How can we make this information more tangible and easily understood to the general public, in particular those who are most affected by it? We may just meet for tonight depending on what makes the most sense. Some initial takeaways:
Who we're looking for
Anyone concerned about the impact of scores generated by an algorithm on the relationship between law enforcement and communities. This is for thinking of the best ways to look at and communicate this information, tech-related or not.
Tools
Whatever medium is best for communicating the information clearly, whether it's writing, data visualization, or something else entirely. Additionally, any tools people feel comfortable to look at the dataset, whether that's Excel, Python, R, or the data portal itself.
Relevant Links
Initial repo looking at the dataset: strategic-subject-list
Where we meet
Cafeteria