Chicago Police Department deployed ShotSpotter gunshot detection AI technology across multiple districts, but internal analysis by analyst Scott DeDore found the system only correctly detected 47% of confirmed shootings where people were struck between October 2017 and July 2018.
Between 2017 and 2018, the Chicago Police Department expanded its ShotSpotter gunshot detection technology network across the city as part of newly established Strategic Decision Support Centers. Scott DeDore, a CPD analyst in the tenth district, conducted a nine-month accuracy analysis from October 2017 to July 2018, tracking the system's ability to detect confirmed shootings where people were struck. According to records obtained through public records requests, ShotSpotter correctly detected gunshots in only 63 of 135 instances where a person was shot, achieving an accuracy rate of approximately 47 percent. The technology's performance varied dramatically over time - detecting 9 of 10 shootings in one period but missing all 6 shootings in another month-long period. DeDore regularly shared his findings with Chicago police leadership and ShotSpotter representatives, and worked with the company to install additional sensors to improve accuracy. Despite the documented poor performance, Mayor Rahm Emanuel signed a new three-year, $33 million contract with ShotSpotter one month after DeDore's last available report. The contract covered 12 police districts and 100 square miles, making Chicago the company's largest customer at the time.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed