As AI becomes increasingly integrated into our lives, a new question is emerging: can AI help us live better lives, not just in terms of technology or creativity, but also in how humans make decisions and collaborate?
A recent study by Arend Hintze and Christoph Adami entitled 'Promoting Cooperation in the Public Goods Game Using Artificial Intelligence Agents' , published in the journal npj Complexity, attempted to answer this question.
'The tragedy of shared property' and the test of altruism
The research is based on a well-known economic theory called the 'tragedy of the commons'. According to this theory, when many people share a limited resource, each individual tends to exploit it to the maximum for their own benefit. As a result, the resource is depleted, and the entire group suffers.
To test whether AI can improve human cooperation, the research team used a 'public goods game' model. In this game, each player can contribute to a common fund that benefits the entire group, or keep their share.
In theory, if everyone contributes, the whole group benefits the most. But in reality, people often choose to keep their share and benefit from the contributions of others. Experimental results show that when only people are involved, they mainly act for personal gain rather than for the collective good.
Then, researchers brought AI into the game.
In the first scenario, the AI is programmed to always cooperate. That sounds positive, but in reality, it doesn't change human behavior. Players continue to act selfishly. Simply adding 'good guys' to the system isn't enough.
In the second scenario, the player has control over the AI. The result is even worse: they configure the AI to always cooperate, while they themselves do not. In other words, they 'outsource' good behavior to maximize their personal gain.
The third scenario is what truly yielded remarkable results. The AI was designed to mirror player behavior: if you cooperate, the AI cooperates; if you are selfish, the AI will do the same. This feedback mechanism creates a loop that encourages cooperation. When positive behavior is 'rewarded' with cooperation from the AI, the level of cooperation among players increases significantly.
What does this have to do with self-driving cars?
Although the study simplifies many aspects compared to real-world situations, scientists believe this principle can be applied to many scenarios, including self-driving cars.
Instead of simply adhering to rigid rules, autonomous vehicles could be designed to 'reward' cooperative driving behavior. If enough autonomous vehicles operate in this manner, a positive feedback loop could form, leading to smoother traffic flow.
Studies published in Transportation Research Part C also propose systems for routing and coordinating idle vehicles to serve passengers more efficiently. Meanwhile, a work in Robotics introduces a system for tracking and maintaining visual connections between autonomous vehicles to avoid collisions.
The principle of encouraging cooperation can also be applied to scheduling charging for self-driving electric vehicles, in order to reduce long waiting times and limit pressure on the power grid.
AI cannot 'erase' human selfishness. However, if designed correctly, it can create enough incentive for cooperation to become the most beneficial option.
In fact, current AI systems – including chatbots like ChatGPT and Gemini – learn based on a reward-and-punishment mechanism to optimize performance. This same approach could easily be extended to future robotaxi systems and intelligent transportation infrastructure.
If implemented properly, AI can be more than just a technical support tool; it can also become a 'behavioral regulator,' helping society function more efficiently and with less conflict.