Latest / AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence / When Your AI Boss Decides Your Next Shift
Transcript
- Lucas: If you've worked in retail or fast food over the past few years, you've probably had your shifts generated by an algorithm. No manager writing a weekly schedule on paper — it's all automated now. Luna: Right, and the pitch is efficiency. The AI can optimize for labor costs, customer traffic, inventory needs — all at once. Lucas: Exactly. But what happens when that optimization ignores basic human constraints? I'm thinking of a specific case from 2025 — a large grocery chain based in California. They rolled out an algorithmic scheduling system from a major vendor, and it started giving workers back to back shifts at different store locations. Luna: Meaning an employee would close one store at 11 p.m. and then be scheduled to open another store at 6 a.m. the next day, with a 30-minute drive in between. Lucas: Precisely. The algorithm only looked at labor cost per store — it didn't check if a worker could actually make it from Store A to Store B in time, or if they'd be getting any rest. In early 2026, a class-action lawsuit was filed against the chain and the software vendor, citing wage violations and unsafe working conditions. Luna: It's a textbook example of a model optimizing for the wrong objective. The system had all the data about shift times and locations, but it just wasn't programmed to factor in travel time or mandatory rest periods. Lucas: And the thing is, these systems are used by dozens of major retailers and fast-food chains. They're sold on the promise of cutting labor costs by 10 to 15 percent. But the human cost — employee burnout, safety risks, wage theft — that's not in the sales pitch. Luna: So let's get into the nuts and bolts. How does one of these scheduling algorithms actually work? Lucas: At its core, it's a constrained optimization problem. You have variables: number of employees available, their hourly wages, their skill sets — who can work the register, who can stock shelves. Then you have demand forecasts: how many customers are expected each hour, how many cashiers you need, how many stockers. Lucas: The algorithm tries to assign shifts to meet demand while minimizing total labor cost. But the constraints it uses often only include things like 'no employee works more than 40 hours a week' or 'each shift is at least 4 hours long.' Things like 'travel time between locations' or 'minimum 10 hours between shifts' are left out. Luna: And in this case, they were left out explicitly because they'd increase labor costs. The vendor even marketed the system as 'unconstrained optimization' to maximize savings. Lucas: That's the ethical crux. The algorithm wasn't making a mistake — it was doing exactly what it was designed to do. The problem was the design itself. And the workers who suffered were mostly hourly employees with little bargaining power. Luna: There's actually a study from the Economic Policy Institute back in 2023 that found nearly 20 percent of hourly retail workers receive their schedules less than a week in advance. Algorithmic scheduling can make that even worse. Lucas: Right, and unpredictable schedules make it nearly impossible to arrange childcare, hold a second job, or even plan meals. So the lawsuit in California is about more than just clopenings — it's about a system that treats employees as interchangeable units. Lucas: Now, if today's conversation gave you something useful — maybe a new way to think about the software behind your own work schedule — I want to mention that the way we keep this show ad-free is through listener support. Luna: It's true. We don't run ads, we don't have sponsors, and that's intentional. It lets us talk about cases like this without any corporate filter. Lucas: So if you value that independence, you can support the show at buy me a coffee dot com slash fexingo. That's buy me a coffee dot com slash fexingo. It literally helps us pay for research and hosting. Luna: And we really appreciate it. Okay, back to the scheduling problem — what were the specific allegations in the lawsuit? Lucas: The plaintiffs alleged that the algorithm systematically violated California's 'predictive scheduling' law, which requires employers to provide schedules at least two weeks in advance and to pay premiums for last-minute changes. They also claimed that back to back shifts without adequate rest violated state wage and hour laws. Luna: And the vendor — what was their defense? Lucas: The vendor argued that they provided a tool, and it was up to the retailer to configure the constraints. They said the algorithm was capable of including travel time and rest periods, but the retailer chose not to enable those features because it would reduce savings. Luna: So the retailer is caught between legal compliance and cost-cutting, and the vendor shifts responsibility. It's a classic accountability gap. Lucas: Exactly. And this isn't an isolated case. In 2024, a fast-food chain in New York faced a similar lawsuit over its scheduling algorithm, and a federal investigation found that the system had scheduled minors during school hours. Luna: So what's the fix here? Is it regulation, better algorithm design, or both? Lucas: I think it's both. Some states have already passed predictive scheduling laws, but enforcement is weak — especially when algorithms make it easy to claim it was an 'error.' On the design side, some companies are starting to build 'worker well-being' metrics into their models. Lucas: For example, a startup called ShiftWell offers a scheduling algorithm that penalizes back to back shifts and unpredictable hours. They claim it reduces turnover by 15 percent while only increasing labor costs by 2 percent. Luna: That's a compelling data point. It suggests that treating workers better isn't necessarily a huge cost — and it might actually save money in the long run through lower turnover. Lucas: Right. But the problem is that most companies are still optimizing for short-term quarterly results. The algorithm that saves 10 percent on labor today looks great on a spreadsheet, even if it leads to lawsuits and high turnover next year. Luna: And that's where transparency comes in. If workers and regulators could audit the algorithm's decision rules, companies would be forced to justify excluding things like rest periods. Lucas: Absolutely. Some cities, like Seattle, have proposed 'algorithmic accountability' ordinances that would require companies to disclose how their scheduling systems work. But the tech industry has pushed back hard, arguing it's proprietary. Luna: Proprietary or not, if an algorithm is making decisions that affect people's safety and income, there should be a way to challenge it. That's a core principle of AI ethics. Lucas: And it's not just scheduling. The same pattern plays out in hiring, credit scoring, and healthcare. Optimize for one metric, ignore the human context, and you get harm. Luna: So to bring it back to the grocery chain case — what's the status of the lawsuit now? Lucas: As of this month, June 2026, the case is still in discovery. But it's already had an impact. The retailer announced they would pause the algorithmic scheduling system and review their practices. And the vendor released a software update that now includes a 'rest period constraint' option — though it's off by default. Luna: Off by default — that's a telling design choice. It puts the burden on the buyer to opt in to worker protection. Lucas: Exactly. And that's why I think the real fix has to come from regulation. Because as long as cost savings are the default, companies will keep using algorithms that treat people like numbers. Luna: And we, as listeners, can pay attention to the software that shapes our work lives. Ask your employer how your schedule is generated. It's a small question, but it can start a conversation. Lucas: Well said. So that's the state of algorithmic scheduling in 2026. A technology that promises efficiency, but too often delivers exploitation — unless we demand better design and stronger rules. Luna: And that's a conversation worth having. Thanks for listening.