Amazon's AI plans: Driver safety or surveillance?

Amazon’s new AI-based driver monitoring system may be marketed in terms of driver safety, but the whole setup is clearly designed to put drivers under constant surveillance.

The driver information video [https://vimeo.com/504570835/e80ee265bc] is aimed at drivers, but the underlying message serves the interests of Amazon’s managers and PR machine.

There is little to no justification for using AI in this. Few if any of the benefits of this system require AI. “Stupid” tech will do the job just fine. For example, a simple dashcam would provide the driver with evidence in incidents like the collisions they mention, and anyone viewing the footage could tell the fault, it doesn’t need AI.

Assisted driving tools can also be useful – vehicles increasingly have sensors that help drivers manage the different hazards of the road, whether that is watching speed or judging safe distance from the vehicle in front. Even curbside cameras could be useful to protect drivers in the case of an incident, perhaps triggered by a driver-controlled panic button. But keeping even external cameras on while a driver is on a break highlights the concern for protecting their property over the safety of their people.

Control by the drivers themselves is key. Otherwise we are simply turning people into pieces of a data-driven machine with no agency of their own. All these tools need to be integrated in a way that supports drivers, not as an excuse for unnecessary AI and surveillance.

Excessive data collection and AI processing has costs for privacy, agency and the environment. We need to stop prioritising business interests when thinking about building and using invasive technologies, and as my recent paper at the Resistance AI workshop at NeurIPS2020 suggested, if we centre those who are most affected then we sometimes just have to “put it in the bin”.

With other options available, the Netradyne system doesn’t have a strong case for why it is necessary. Beneath the claims, the main reason will be increasing Amazon’s control over its workers, and converting those people into metrics that can be tracked to tick a box in their PR around safety.

Amazon drivers are already tied up with invasive technologies like the privacy-abusive Ring doorbell system. Like Amazon’s warehouse workers, they are put under enough pressure and surveillance as it is. They are given little job protections and gruelling targets. This often leads to them having to cut corners, but instead of resolving these issues, Amazon seems set on using surveillance to shift the blame for bad employment practices onto individual workers. The driver cam escalates this.

Netradyne, the makers of this “fleet management system” brag about capturing 100% of driving time, putting hassled drivers under the additional strain of being constantly watched. It may be that the added pressure of this always-on surveillance is more likely to make drivers make mistakes if they are worrying about that rather than concentrating on the road.

Amazon seem intent on turning every part of their business into a data-obsessed machine, treating their workers only as metrics to be controlled. With the increased reliance on people like Amazon delivery drivers throughout the pandemic, it’s a kick in the teeth to then turn around and say they need to be constantly watched as they take on risks for the company and consumers.

Dr Garfield Benjamin's research spans cultural theory and creative media practice, focusing on the relation between humans and (digital) technology.