Navigating the Exceptions: The Achilles' Heel of Algorithmic Decision-Making

As society becomes more technologically advanced, algorithms are taking on a greater role in completing tasks. However, with the increased reliance on these programmed protocols, common sense and intuition often take a backseat. These algorithms are designed to handle typical scenarios, but struggle when faced with exceptions. As a result, humans are increasingly out of the loop.

Recently, I decided to try two new service that promised swift delivery of groceries and beverages. However, my foray into the unknown left me with a sour taste, as I experienced firsthand how algorithmic "efficiency" can lead to more inefficiency than one might anticipate.

The other day, as I was eagerly awaiting my beverage delivery, I was staring out the window and finally spotted the van turning onto the street. But 5 minutes later, with no sign of movement from the person behind the wheel, I decided to go take a look. After greeting the driver and inquiring what was going on, he showed me his smart phone which displayed the following message: "mandatory pause". Apparently, Germany has regulatory breaks for commercial drivers. But since my apartment was only a hundred meters away, we could easily have carried over the boxes in the meantime. However, the algorithm hadn't taken that into consideration as this case had not been pre-programmed into its system. The algorithm disregarded the proximity to the client when it carried out its obligatory halt. So we patiently waited another 30 minutes before finally carrying over those precious boxes.

In a similar instance, I placed an order for groceries via an online delivery service. Unfortunately, I missed the doorbell by a mere two minutes. However, I caught sight of the delivery vehicle across the street. Eager to remedy the situation, I dashed out to speak with the driver, only to discover he'd just cancelled my order. We both made earnest attempts to resolve the issue with the headquarter, but there was no chance. My groceries drove away. As a result, I had to place an identical order, leading to another car redundantly covering the distance to deliver the very groceries that had been within arm's reach.

These two scenarios highlight the inherent limitations of algorithms, a challenge well understood by anyone involved in programming. Addressing every edge or exceptional case can easily turn into a long series of "if-else" conditionals. Given this, it's overly optimistic to buy into the hype surrounding "artificial intelligence" as some sort of magic bullet for resolving such nuanced issues. Machine learning algorithms, while advanced, are fundamentally dependent on the data they've been trained on. This data may include a range of edge cases, but the algorithms can only offer probabilistic solutions—they are not equipped to improvise in all such situations.

An algorithm cannot account for every possible scenario or exception, which is why human involvement and collaboration are crucial. This is especially evident in situations where desperate customers seek assistance. However, call agents are increasingly limited to following protocols dictated by algorithmic decision-making. In a recent post, Jamie Bartlett highlights the negative effects of automation, focusing on how it often prioritizes cost reduction over quality service and leaves customers to deal with issues themselves. Anyone who has ever had to dispute an invoice or deal with service problems can relate to this frustrating experience. It's like being trapped in a Kafkaesque nightmare, passed along from one call agent to another without any resolution in sight. And this issue is not limited to traditional industries; it applies to modern ones as well.

Startups often simplify a service and offer limited options, making it seem effortless. If you're just looking to book a cab, then using an app is convenient and provides a sense of security. However, services like Uber do not allow you to request accommodations for wheelchairs and make it difficult to voice complaints. And good luck getting an issue resolved with booking.com. These large-scale digital services only function smoothly if everything goes as planned. Companies are increasingly relying on bots to interact with customers, which can be frustrating when they lack common sense and empathy. Ultimately, these automations have not significantly increased overall productivity in the economy - but that's a different story altogether.