AI can detect 90% of crimes before they happen

Question

An article by Jack Dunhill a Social Media Coordinator and Staff Writer for IFL Science, posted an article claiming that a "new" type of AI was able to detect 90% of crimes before they happen. The title of the article and claim happen to be misleading, essentially the AI doesn't detect crimes before they happen but uses data from past crimes in a 1,000-square-foot area of Chicago to predict when and where crimes are more likely to occur based on previous criminal data. Additionally, the "90% accuracy" they mention refers only to how well the AI predicts general crime hotspots and not exact events. The article also leaves out the important limitation of the fact that not all crimes are reported, which can affect the accuracy of its 90% prediction rate. So therefore, this AI doesn't detect crimes before they happen but instead predicts the likelihood of crimes occurring, giving police and other first responders a head start in responding to potential crime rather than stopping specific crimes before they occur.

Answer 1

I believe that the claim AI can predict "90% of crimes before they even happen" is misleading or even exaggerated. The primary source simplifies and exaggerated the finding from a really study that was conducted by actual researchers at the University of Chicago and was then published in Nature Human Behavior. The study showed that the AI analysis past crime data in small geographic areas in order to predict patterns and hotspots, which is not specific crime. or individuals. I think that the "90%" is referring to  how model predicts general crime and not exact events. The researchers involve are credible academics who had a goal of studying patterns in crime. The down side of this however, is the article IFLScience simplified and exaggerated their findings to attract the attentions. A primary source would be the original study where it explains the actual purpose and limits of the AI. Second source is the University of Chicago summary where it clearly states that the AI predict trends about a week in advance, but not individual crimes.

The evidence supporting the claim is that the AI can identify crime hotspots using past data, which may help police prepare. But, it does not predict specific crimes and relies only on data that is reported, which can limit some accuracy. The researchers also noted that concerns about bias in policing. The IFLScience article may be biased because it simplifies a lot of facts just for more attraction. 

https://www.nature.com/articles/s41562-022-01372-0

https://biologicalsciences.uchicago.edu/news/algorithm-predicts-crime-police-bias

https://www.iflscience.com/ai-predicts-90-percent-of-crime-before-it-happens-creator-argues-it-wont-be-misused-65025

Answer 2

The claim that AI can “detect 90% of crimes before they happen” is misleading and not true in the way it is stated. According to the University of Chicago study published in Nature Human Behavior https://www.nature.com/articles/s41562-022-01372-0, the AI system does not actually predict specific crimes before they occur, instead, it analyzes historical crime data to identify geographic areas where crime is statistically more likely to happen. The reported “90% accuracy” refers only to how well the model predicts general crime hotspots over time, not individual criminal events or prevention of crime itself. Because it relies on past police data which can be incomplete and biased the system is limited to forecasting patterns rather than foreseeing or stopping crimes.

Answer 3

This claim that "AI can detect 90% of crimes before they happen" is extremely misleading to those who see this. AI doesn't see the future, it probably knows after the crime happens but it would not be able to tell before. In fact after doing research on this some of the wrong people have been arrested you were innocent due to AI saying it was an 100% to the person who did it. I read that it is more like a weather report, and weather reports are not always accurate and neither is AI.

https://www.joneswalker.com/en/insights/blogs/ai-law-blog/ai-police-surveillance-bias-the-minority-report-impacting-constitutional-right.html?id=102lqdv

Answer 4

The claim that AI can "detect 90% of crimes before they happen" is a classic case of a sensationalized headline oversimplifying complex science. While the 90% figure is technically accurate, it doesn't mean the AI is a "Minority Report" style precog. Instead, researchers from the University of Chicago developed a model that identifies statistical "hotspots" within 1,000-foot areas about a week in advance. It’s a tool for geographic risk assessment—predicting where and when a crime might occur based on historical patterns—rather than identifying specific individuals or preventing an exact event from unfolding.

A deep dive into the primary research published in Nature Human Behaviour reveals that while the algorithm is highly effective at spatial forecasting, it is limited by the data it consumes. Since it relies on reported crime records, it inherently reflects existing enforcement biases and misses crimes that go unreported. Popular secondary sources like IFL Science often strip away these crucial nuances to grab attention. Ultimately, the tech is an impressive leap in urban data science, but it's far from the infallible crime-detection system that viral articles suggest.

Answer 5

This claim is a very misleading topic. Stating that "AI is detecting 90% of crimes before they happen" is certainty not true and its giving AI too much credit. When I look this claim up nothing of the sort comes up. What I did find was that AI uses algorithm to predict that their will be more crimes in certain areas but it cant give us any specific details on how? For example, according to the University Of Chicago " Algorithm predicts crime a week in advance, but reveles bias in police responce." This doesnt state anything about AI but it does mention algorithm hich is what helps determine these crimes.

Answer 6

The claim that AI can detect 90% of crimes before they happen is misleading. First, the primary source of this claim comes from an article on IFLScience which has several other attention grabbing titles and even seem like "click bait". The article by Jack Dunhill summarizes research from the University of Chicago but frames it in a way that study does not actually support. It suggests a pre-crime capability that isn't actually supported by the research. Looking into more trustworthy sources, an article from New Scientist reports that the AI predicts crime levels and locations a week in advance with up to 90% accuracy, rather than specific incidents (https://www.newscientist.com/article/2326297-ai-predicts-crime-a-week-in-advance-with-90-per-cent-accuracy/?utm_source=copilot.com). Even then, the article describes how there is a questionable bias to some of the statistics and predictions that make it hard to completely trust. To support this even more, an article from Bloomberg News similarly explains that the model divides cities into 1,000 sq. ft. tiles and forecasts crime patterns using historical data (https://www.police1.com/investigations/articles/algorithm-claims-90-accuracy-in-predicting-crime-in-us-cities-before-it-happens-gpndkTXQaybYPxX3/?utm_source=copilot.com). It does not identify individuals or specific events. This proves that the original claim is misleading and exaggerated.Tracing the claim back to its original context shows that the research published by Nature Human Behavior uses historical crime data to predict the likelihood of crime occurring in specific geological areas. The 90% accuracy refers only to how well the model predicts hotspots, not individual events or precise future incidents. The model's accuracy is also limited since it only relies on reported crimes, which are only a portion of actual criminal activity.The evidence shows that the AI does not detect crimes before they happen and instead identifies statistical patterns that may help police anticipate where crime is more likely occur. The original claim exaggerates the model's capabilities and is therefore misleading.

Answer 7

The claim that AI can detect 90% of crimes before they happen is extremely misleading and is based off a 2022 study published by researchers at the University of Chicago in a study called Nature Human Behavior. It is based on patterns in reported crime of a 1000 foot radius in actions for the week prior. The area under the curve measures the models ability to distinguish high and low risk zones, rather than its ability to outirght oredict the crime to happen. this model also relies on crime that is already reported and still allows police bias to happen and limits the actual use of AI and prediciting it.

Answer 8

After applying the SIFT method, I found that the claim "AI can detect 90% of crimes before they happen" is misleading

At first the original article form IFLScience exaggerates what the technology actually does. The AI system, developed using research from the University of Chicago does not detect or stop crimes before they occur. Instead, it analyzes past crime data within small geographic areas to predict where crime is more likely to happen in the near future.

When investigating the source further, it became clear that the "90% accuracy" refers only to predicting general crime hotspot, not specific crimes, people or events. Additionally the claim is weakened by the limitation that not all crimes are reported. Because the AI relies entirely on recorded data, any gaps or biases in reporting will directly affect its predictions. This means the system cannot fully represent actual crime patterns, which further challenges the accuracy of this claim.

Comments

Always include links to your sources going forward. Thanks!

Answer 9

I looked into this claim and it is pretty misleading. The website IFLScience is a pretty popular science website. It's not a super academic source so it is not super accurate. I looked in the article called BBC and it makes this claim more misleading. It talked about how AI is not actually detecting crimes before they happen. It just uses past crime data to predict where crime is more likely to happen. It is more about locations and not specific crimes and exactly what's going to happen. The 90% accuracy part is also confusing. It does not mean the AI is right about 90% of future crimes it just means that it is good at predicting where crimes could be based on old data. That makes 90% seem better than it really is. I think the claim is exaggerated, the AI can help predict where crime might happen, but it definitely can not detect other stop 90% of crimes before they happen.

Goodyer, J. (2022, August 19). An algorithm can predict future crimes with 90% accuracy. here’s why the creator thinks the Tech won’t be abused. BBC Science Focus Magazine. https://www.sciencefocus.com/news/algorithm-predict-future-crimes-90-accuracy-heres-why-creator-thinks-tech-wont-be-abused 

Answer 10

The claim that AI can "detect 90% of crimes before they happen" is a big exaggeration. While the headline sounds like a movie, a study from the University of Chicago shows that the AI is really just looking at patterns in old data. It doesn't actually see a crime about to happen, it just points out small areas where crimes are most likely to occur based on what has happened there before.

There could be bias as AI companies are trying to give out more exaggerated claims to give them more value. The 90% accuracy part is also pretty misleading. According to The Verge, this doesn't mean the AI knows exactly who will do something or what will happen. It just means the AI is good at picking what they call "hotspot" areas. If the AI predicts a certain block is high  risk and a crime eventually happens there, it marks that as positive or correct, even if it didn't know any of the specific details regarding it.

After relooking at the article, the claim misses a huge problem with the data itself as Bloomberg Law points out that these AI models only know about crimes that are actually reported to the police. Since many crimes never get reported the data is incomplete by a lot. Instead of stopping crimes before they start, the tool is really just a digital map that helps police guess where they should spend their time patrolling.

The Verge: Predictive policing AI is more about patterns than precognition

Bloomberg Law: The bias and limitations of predictive algorithms in law enforcement