When a red light stops the clock

For a food delivery rider, a red light could mean two different things at the same time: stop and hurry. This contradiction highlights the way algorithms have come to organize work. But this might change soon.

In recent months, companies such as Meituan, JD.com and Taobao Instant Commerce have begun piloting a simple adjustment in some cities.

When a delivery rider is stopped at a red light, the delivery timer on his app pauses. Once the light turns green, it resumes, and the order deadline is extended by the time spent waiting.

On a phone screen, this may look like a minor tweak to a stopwatch. But for the rider, it completely changes the significance of those 60 or 90 seconds. A red light is no longer time taken away from him.

For years, the delivery business has operated on a simple principle: speed is measurable, and what is measurable can be managed.

Algorithms estimate how long it should take to pick up an order, cross a neighborhood and reach a customer's doorstep. Deadlines are set down to the minute. Late deliveries can mean lower scores, lost bonuses or customer complaints.

But cities do not run on algorithms. A rider may have to wait for several traffic lights on a short trip. A restaurant may take longer than expected to prepare an order. An elevator may stop at every floor.

A security guard may ask for an ID before allowing entry into an office building. None of these delays is the rider's fault. Yet, under a rigid countdown, they all become his problem.

It is easy to see how this can lead to dangerous choices. When every second has a price, obeying a traffic light can feel like a personal loss. The rider who waits patiently may arrive later than one who takes a risk. What appears as reckless behavior could be a calculated decision from the rider's perspective.

That is what makes the new timer system more important than it seems. The real issue is not whether a delivery platform should grant riders another minute or two. It is about what the platform considers a legitimate part of work.

For a long time, the answer was essentially whatever the algorithm could measure easily. Riding time counted. Delivery time counted. Waiting at a red light did not. But algorithms do not merely record reality; they help shape it.

When a system rewards speed above everything else, riders learn to find speed wherever they can. When it fails to make a distinction between controllable and uncontrollable delays, it shifts the cost of uncertainty onto individual workers.

This is where algorithmic management becomes a matter of public governance. It is tempting to think of algorithms as neutral software — efficient, objective and free of the messy judgments that human managers tend to make.

But they are none of these things by default. Every system embodies choices about what matters, what can be ignored and who bears the consequences.

Consider the red light. Traffic signals are designed to slow everyone down for the sake of safety.

But if the delivery algorithm treats this pause as lost productivity, two systems of rules are working against each other: the traffic system says stop, while the labor system says keep moving.

No amount of rhetoric about road safety can fully resolve this contradiction. The incentives must change.

That does not mean efficiency has to be sacrificed. Quite the opposite. A delivery system that saves a minute but encourages dangerous riding, accidents, complaints and worker turnover may not be efficient at all.

There is a broader lesson here for the future of work. As algorithms take over more decisions about hiring, scheduling, evaluation and pay, a crucial question arises: can algorithms recognize the difference between what a person can control and what he cannot?

Bad weather cannot be negotiated with. Congestion is not always predictable. An elevator cannot be ordered to hurry. A red light cannot be persuaded to turn green.

A humane algorithm does not have to be sentimental. It simply needs to accurately reflect the real world.

That is also why the next step will be more challenging. Stopping the clock at a red light is relatively easy because the system can identify the event.

Far more difficult are the gray areas of urban life: a sudden downpour, an unexpected traffic jam, a hospital with complicated access procedures, an old residential compound where finding the right building takes several minutes.

The challenge is to teach algorithms to account for these realities without turning every exception into another complexity.

The goal should not be an algorithm that is endlessly lenient, nor one that is relentlessly strict. It should be one that can distinguish between delay and failure.

This may sound like a modest ambition, but it points to a larger question about the digital economy.

Amid all the excitement about artificial intelligence becoming faster, smarter and more autonomous, perhaps a measure of progress is simpler: whether technology can better understand when a human being is doing exactly what society expects him to do.

At a red light, the right thing is to stop.

The clock should know that too.

Li Cong is a professor at the School of Economics and Finance, Xi'an Jiaotong University; and Li Minglai is an assistant professor at the same institute.

The views don't necessarily reflect those of China Daily.

If you have a specific expertise, or would like to share your thought about our stories, then send us your writings at opinion@chinadaily.com.cn, and comment@chinadaily.com.cn.