Uber patent details AI that can tell if riders are drunk or tired

Uber has submitted an application to patent an artificial intelligence technology that identifies drunk passengers. The patent, which is dated June 7 with the US Patent and Trademark Office, is titled "Predicting User State Using Machine Learning," and it hints that such a system may be used to accommodate riders who are outside of their "normal behavior" state.

The patent describes a system that is trained to identify passenger states so that it can determine when they're acting outside of their normal behavioral state. "The prediction is based on data associated with a request submitted by a user," Uber explains in the patent. Machine learning would be behind this system.

The company indicates that the system could be used as a way to prevent issues that could arise when "users and/or providers behave uncharacteristically." Being drunk is perhaps the most common instance in which a passenger may be in an altered state that results in "uncharacteristic" behaviors, but the patent also points toward tiredness as another possible issue.

One usage example is that a tired passenger could have trouble finding the driver's car compared to someone who is well rested. Knowing whether a passenger is behaving unusually can help mitigate any issues before they become problems. Based on the patent, the system appears to be one that depends on past user data to predict future behaviors.

The user's current state is also determined by this system based on how they handle their smartphone. If the user is clumsy with their handset, for example, and if they make a lot of typographical errors, the system would guess a level of impairment that might indicate abnormal behaviors. The patent explains things that may help the system make its determination:

For example, the user activity can include text input characteristics, interface interaction characteristics, and device handling characteristics. For example, text input characteristics may include the number of typographical errors entered by a user or the number of characters erased by a user while entering a search query. Examples of interface interaction characteristics include the amount of time for a user to interact with the user interface after new information (or a modified display) is shown to the user, or the user's accuracy in pressing an interface element on the device.

So, for example, a user who typically walks at 3MPH, who rarely makes typos while entering text on their phone, and who doesn't have to try to tap buttons multiple times, will have trained the system with those characteristics. Based on this, the system will guess the person is in an altered state if they're walking at an unusual pace, making typos, and having trouble tapping buttons.

How would Uber use this data? It could, for example, prevent a possibly drunk passenger from joining a ride that has multiple passengers, a way to avoid possible issues. The system could also pair riders with drivers who are experienced to handle them. Of course, a patent doesn't mean the system will actually ever be implemented.

SOURCE: USPTO