This is an automated rejection. No LLM generated, assisted/co-written, or edited work.
Read full explanation
As AI begins to manage everything from traffic flow to job markets, staying autonomous might become a luxury few can afford. New research quantifies four "Prices of Autonomy" resulting from AI’s ability to manage our resources more efficiently than we can.As AI begins to manage everything from traffic flow to job markets, staying autonomous might become a luxury few can afford. New research quantifies four "Prices of Autonomy" resulting from AI’s ability to manage our resources more efficiently than we can.
The Autonomy Dilemma
Would you let an AI decide which lane you should drive in, which job you should apply for, or how to divide limited office space? Most of us value our “autonomy” — the right to make our own choices. But a new study, “The Prices of Autonomy in Resource Division,” (under review) reveals how holding onto that autonomy could become increasingly expensive.
Using AI tournaments to simulate social games like the “Lifeboat Problem” and the “Kolkata Paise Restaurant Problem,” the study examines what happens when we introduce “advised players” (or robots) into populations trying to divide scarce resources.
The Findings: 4 Ways Autonomy Costs Us
The research allows us to measure four specific prices associated with optimized AI advice:
The Benefit of Advice: Following AI advice (specifically a strategy called “turntaking”) would be more efficient and fair than current resource division. That potential improvement is the cost we currently pay for delaying or banning AI, and it falls mostly on people who find themselves in what amounts to a lower caste, repeatedly receiving less access to resources.
The Collateral Damage: As more people follow AI advice, those who don’t follow it see their relative access to resources plummet. In a world of efficient robots, former chess masters lose their edge to those who let AI tell them what to do.
The Price of Your Own Autonomy: This is the profit that an individual loses by making their own decisions. The study found that as more people follow AI, the cost of staying independent grows exponentially, creating a viral pressure to give in and let the AI lead.
The Price of Others’ Autonomy: This is AI users’ economic incentive to ban human autonomy, forcing everyone to let AI drive. Coordination benefits everyone, so you are not the only person hurt when you choose to be your own driver. Those who limit optimization and those who limit autonomy may each label the other “selfish” in a different way.
Why Can’t We Just Do It Ourselves?
You might wonder why we can’t just adopt efficient strategies without AI. The study found that the grandmaster strategy — turntaking — is too complex for the human mind to track. It requires keeping detailed accounts of “favors owed” and “debts incurred” across every interaction in a population. We humans simply aren’t built to be that fair or that efficient on our own. And the games included in this study probably represent just the tip of an iceberg of our imperfectability.
The Bottom Line
We are entering an era where maturity might mean recognizing our limitations.
The question isn’t whether AI is better at dividing resources — anyone with a computer can download the linked software from GitHub and reproduce the controlled experiments proving that it is. The real questions are: Are you willing to pay the price to keep your autonomy? and How can we build the empathy necessary to hold our society together when our differing answers to that first question threaten to turn us against each other?
More resent (unpublished) research includes actual human participants (using the same GitHub package), confirming that we do fail to take turns and we do fall into conflict over whether to follow AI advice.
The Autonomy Dilemma
Would you let an AI decide which lane you should drive in, which job you should apply for, or how to divide limited office space? Most of us value our “autonomy” — the right to make our own choices. But a new study, “The Prices of Autonomy in Resource Division,” (under review) reveals how holding onto that autonomy could become increasingly expensive.
Using AI tournaments to simulate social games like the “Lifeboat Problem” and the “Kolkata Paise Restaurant Problem,” the study examines what happens when we introduce “advised players” (or robots) into populations trying to divide scarce resources.
The Findings: 4 Ways Autonomy Costs Us
The research allows us to measure four specific prices associated with optimized AI advice:
Why Can’t We Just Do It Ourselves?
You might wonder why we can’t just adopt efficient strategies without AI. The study found that the grandmaster strategy — turntaking — is too complex for the human mind to track. It requires keeping detailed accounts of “favors owed” and “debts incurred” across every interaction in a population. We humans simply aren’t built to be that fair or that efficient on our own. And the games included in this study probably represent just the tip of an iceberg of our imperfectability.
The Bottom Line
We are entering an era where maturity might mean recognizing our limitations.
The question isn’t whether AI is better at dividing resources — anyone with a computer can download the linked software from GitHub and reproduce the controlled experiments proving that it is. The real questions are: Are you willing to pay the price to keep your autonomy? and How can we build the empathy necessary to hold our society together when our differing answers to that first question threaten to turn us against each other?
More resent (unpublished) research includes actual human participants (using the same GitHub package), confirming that we do fail to take turns and we do fall into conflict over whether to follow AI advice.
Read the full preprint here: https://dx.doi.org/10.2139/ssrn.6194078