Trusting AI to increase productivity? Perspectives Across the Global North and South
This exploratory study reveals a counterintuitive divergence between the Global North and South, where users in the Global South exhibit higher trust in generative AI but report fewer productivity gains compared to their Global North counterparts, suggesting that trust alone is insufficient for productivity without adequate access, task suitability, and verification needs.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of work as a giant, bustling kitchen. For years, chefs (developers and researchers) have been trying to figure out if a new, magical robot assistant (Generative AI) can actually help them cook faster. The big question isn't just "Can the robot chop the onions?" but "Do the chefs trust the robot enough to let it chop, and does that trust actually make the dinner come out of the kitchen sooner?"
Two researchers, Adam and Shalini, decided to investigate this by looking at chefs from two different neighborhoods: the "Global North" (places with lots of fancy, well-stocked kitchens) and the "Global South" (places where kitchens might be smaller, have fewer tools, or face different challenges).
The Great Search for Clues
First, the researchers went on a massive treasure hunt through the library of existing science. They looked for any study that connected three specific things: trusting the robot, getting work done faster, and the Global South. They dug through nearly 5,000 papers! But here's the twist: they found zero peer-reviewed studies that had all three pieces of the puzzle together. It was like looking for a specific type of blueberry in a forest and finding only red ones, green ones, or no berries at all. The only clues they found were in "grey literature"—reports from big companies and think tanks—which hinted that the robot is great for productivity but didn't explain how trust fits in.
The Survey: A Snapshot of 36 Chefs
Since the library was empty of the right answers, the researchers asked 36 people directly. These weren't just any people; they were a mix of folks born in the Global South working there, folks born there but working elsewhere, and folks born and working outside the Global South.
Here is what the data suggested, based on their answers:
The Trust Score: The researchers asked people how much they trusted the AI.
- People born and working in the Global South gave the robot a high trust score of 0.83 (on a scale where higher is more trust).
- People born and working outside the Global South gave it a lower trust score of 0.30.
- The mixed group (born in the South, working outside) was actually the most skeptical, with a score of -0.15.
The Productivity Score: Then they asked, "Does using this robot make you faster?"
- Surprisingly, the people who trusted the robot the most (the Global South group) reported a productivity score of 0.62.
- The people who trusted it the least (the non-Global South group) reported the highest productivity score of 0.68.
- The mixed group reported the lowest productivity at 0.22.
The Big Reveal: Trust Isn't the Magic Wand
The most exciting part of this story is that trusting the robot doesn't automatically mean you get more work done.
Think of it like this: Imagine you have a super-fast sports car (the AI).
- Group A (Global South) loves the car. They trust it completely! But maybe they are driving on a bumpy dirt road with a speed limit sign, or they can only afford the basic model. They trust the car, but they can't drive it as fast as they want.
- Group B (Global North) is a bit suspicious of the car. They think, "Is this thing safe? Is it lying to me?" But they are driving on a smooth, wide highway with no speed limits and a full tank of premium gas. Even though they are skeptical, they zoom past everyone because their environment lets them go fast.
The data suggests that access is the real secret sauce. The people outside the Global South reported saving an average of 21.48 minutes per task. The people in the Global South only saved about 10.83 minutes per task.
Why the Difference?
The researchers found that the Global South group faced more "roadblocks." The most common barrier was the high cost of subscriptions (paying for the premium car), followed by limited availability in their region and rules from their bosses or governments.
One participant put it perfectly: "I don't have to trust the system for it to increase my productivity. I cross-check for errors... as I don't use a paid service and use basic models." This suggests you can still get work done faster even if you don't fully trust the AI, as long as you have the right tools and the freedom to use them.
What This Means (And What It Doesn't)
The authors are careful to say this is just the beginning. With only 36 responses, and only 3 people from the specific group of "born in the South and working in the South," they can't say this is a final, proven law of the universe. They are just suggesting a pattern.
They argue that we can't just assume that if we teach people to trust AI, productivity will magically skyrocket. Instead, we need to fix the roads (infrastructure), lower the tolls (costs), and make sure everyone has a license to drive (training and permission). Until then, the robot might be trusted, but it might not be able to do its job.
In short: Trust is important, but it's not the whole story. You can trust a car all you want, but if you're stuck in traffic, you're still going to be late.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.