For hundreds of years, philosophers like Thomas Hobbes and economists like Adam Smith have been asking: How can we make our society fair? How can we distribute the wealth we have generated together in such a way that everyone can participate and individual achievements are appreciated.
This question can be answered in different ways, for example by the tax system. This is so complicated in Germany that your own tax return will never fit on the beer coaster that Friedrich Merz has been carrying in his pocket for 20 years. We’ve been talking about tax reforms for years, but nothing happens. Democracy in practice is a challenge that we often master rather poorly than well. Is that the topic, or is it the people?
An exciting experiment by the Google-owned research company Deep Mind shows that people are always part of the problem, but not necessarily part of the solution. A team of researchers has now used an AI system to develop a distribution mechanism for public funds. The majority of the more than 4,000 people involved in the experiment found this to be the best solution. Better than any man-made suggestions. AI apparently has the potential to reinvent democracy as well.
The Deep Mind team had trained the artificial intelligence in a series of experiments in such a way that it could learn from more than 4000 people as well as from computer simulations in an online economy game with four players. Players start with different amounts of money and must decide how much to contribute to a pool of public funds in exchange for a share of the pot.
Top jobs of the day
Find the best jobs now and
be notified by email.
Where AI would have won an election
The distribution model calculated by the AI manages to reduce the wealth gap between the players by redistributing the public funds according to the share that the individual participants used from their seed capital. It also penalizes free riders: Players who don’t contribute at least half of their starting capital get nothing back.
The AI has thus succeeded in resolving one of the significant contradictions of a just society: that between an egalitarian distribution approach (public funds are distributed equally, no matter how much someone has contributed) and a libertarian distribution approach (everyone gets their share according to their own contribution). funds repaid).
People liked that too. If the AI’s solution had been up for election in a direct vote, the AI would have won the election. That’s not the point, the research team hastens to assure. With the experiment, one does not want to speak up for an “AI government”. But AI is helping to control more and more areas of life, and it does a good job of solving a society’s distribution problem. So what does this mean for the future of our democracy and voting rights?
AI voting system collects data and develops political proposals
In a research project at the University of St. Gallen, we took a closer look at this question. We wanted to know how high people’s acceptance of an automated AI voting system would be and conducted an online survey in four countries: Switzerland, Singapore, the USA and Greece. Four different political systems with varying degrees of propensity to use technology.
The participants were presented with this scenario: “The government of your country is introducing a new voting system controlled by artificial intelligence. This new system constantly collects various digital data about you in order to find out your actual opinions, ideas and political preferences. Based on the information available about each individual citizen, the AI develops policy proposals. The proposals, which represent the majority of the people, could automatically be translated into practical policies. So the AI voting system would continuously ‘vote’ on your behalf instead of you actively voting.”
Acceptance of such a system is highest among Singaporeans: 39 percent acceptance versus 34 percent non-acceptance. Apparently, the Swiss always remain neutral in polls: 37 percent agree versus 37 percent disagree. A significant proportion of American respondents oppose the system: 45 percent disagree versus 37 percent agree. The Greek respondents adopt a rather hesitant attitude: 25 percent acceptance versus 50 percent rejection.
In technology-related states, the acceptance of an AI voting system is higher
Neither do we want to endorse an automated AI policy or government. In view of the many problems with distortions and risks of discrimination in the data that AI systems then update, such an election system raises many critical questions. But the results show that in some countries more than a third of the population could imagine such a system.
The reasons for this are obvious and can be found in our data: Acceptance is higher in technology-oriented countries such as Singapore. Many citizens there agree that more and more important questions should be solved by AI. People who have lost faith in politics and governments also reject such an AI voting system.
>> Read also: Siemens wants to create a digital platform for industry – and bring companies into the “metaverse”.
The results show the social acceptance of a hypothetical AI democracy. You don’t have to want it, and it doesn’t have to come either. But with the spurt that AI has taken across all living environments in recent years, we should deal with this question.
The former federal constitutional judge Ernst-Wolfgang Böckenförde once said: “The free, secularized state lives on conditions that it cannot guarantee itself.” . Today it is about the centrifugal forces that humans face in view of the possible uses of AI. And the sentence could be: Democracy thrives on conditions that artificial intelligence can now possibly guarantee better than human intelligence.
In this column, Miriam Meckel writes fortnightly about ideas, innovations and interpretations that make progress and a better life possible. Because what the caterpillar calls the end of the world, the rest of the world calls a butterfly. ada-magazin.com
More: The parrot problem of language models