top of page

Can AI determine Safety level of neighborhoods with Google Street View?

Urban perception plays a critical role in understanding how people interact with and feel about their environment. By leveraging Google Street View images and artificial intelligence, such as ChatGPT, we can gain insights into how various urban design elements contribute to these perceptions.

Following up on our previous article, we selected two streets, Street A and Street B, in New York City to analyze using Google Street View images. We provided these images to ChatGPT and asked for an evaluation of safety perception, similar to our previous article. The scores ranged from Very Bad (1), Bad (2), Neutral (3), Good (4), to Very Good (5). We extracted 14 images from each side of both streets.

Street A received an average score of 3.42, while Street B scored 4.38 on average. Since the focus was on the visual information influencing safety perception, factors such as building and sidewalk maintenance, as well as the presence of greenery, were likely decisive, rather than actual crime rates.

Interestingly, the current score for the area around Street A, as reported by Marketstadium, is 60.6, whereas Street B's score is 99.6. The median rent for Street A is approximately $2,500, while for Street B it is about $6,400.

Given that urban studies traditionally focus on safety perception, this article also centers on this aspect. However, conducting an assessment of the physical condition of the streets could provide clearer insights into the correlation between Google Street View images and factors like rent prices and scores.

*It is important to note that this study does not reflect the actual crime rate. ChatGPT cited reasons related to maintenance, eyes on the street, lighting, and visibility as design elements influencing its reasoning, and emphasized that it is unrelated to the actual crime rate.

Figure. Street A

Figure. Street B

Never miss an update!

Subscribe to get all the latest news and exclusive analyses, delivered directly to your inbox. We'll only send occasional emails, typically no more than once a week! To experience our product firsthand, hit the button located in our menu bar above to schedule a demo anytime.

48 views0 comments


bottom of page