Recent analysis, reported by Quanta Magazine, shows that even basic pricing algorithms, those used by online retailers and marketplaces, can steer prices higher without any human agreement.
In traditional economics, firms collude to raise prices; that behavior is illegal and detectable. But the new research uses game-theory models to demonstrate that when independent sellers deploy adaptive pricing algorithms, those algorithms can end up implicitly colluding. By repeatedly adjusting prices based on demand and competitor behavior, they gradually converge toward high-price equilibria that boost profits for sellers while hurting buyers.
One surprising insight: this “algorithmic collusion” doesn’t require complex AI or coordination. Even algorithms programmed simply to maximize profit over time can learn that undercutting competitors reduces long-term gains. So instead of undercutting, they effectively stabilize at higher prices.
What’s worrying is that this doesn’t leave the traditional evidence trail of collusion. There are no secret pacts, no joint price-setting meetings, just autonomous algorithms adapting independently to the same market signals. Regulators, accustomed to policing explicit collusion, may struggle to detect or prove wrongdoing.
For engineers and technologists building e-commerce platforms or dynamic-pricing systems, this research highlights a hidden risk: even “fair” algorithms can lead to unfair outcomes at scale. As more commerce moves online and dynamic pricing becomes the norm, market outcomes may depend less on human intent and more on the emergent behavior of interacting algorithms.