We talked about, how AI technology could be a threat to democracy, but there are also ongoing initiatives, with a specific aim to create AI tools to improve democracy. Enthusiastic, AI developers are elaborating democratic ideas and models, based on how existing and future AI applications could be applied. What do these ideas look like? And how could they work in a real world setting? Would they make democracy better? Often these ideas are directed at different ways to innovate, democratic procedures. One such example, is the idea to make use of an AI avatar, or agent, for each individual. Imagine, all the data about you that is recorded on all the world's computers; emails, documents, texts, social media, search history, CLECs, purchases, Creditex phone and health records, where you drew, where you walked, all the pictures you've ever taken, video from security cameras, and of course, your Fitbit statistics. What picture of you, does all this data form? Probably, a quite accurate and detailed one in many ways, but also one where some essential things would be missing. Why did you once decide to change careers? Could that be predicted based on this data? How about the person you once met secretly, and never forgot? Could that be identified in the data? Even if someone had access to all this data, they wouldn't know, how to turn it into a real likeness of you. But, imagine, a future super algorithm, seeded with everything we could teach it about human life, and the ability to learn from its experiences. Imagine, that we equip this algorithm, with all this information about you. It would be something, like your, digital half, a digital agent, that's a model of you, that you could use for everything you liked. It could do your searching, communicate with other people's digital halves, and everything else to do in the digital world, but much faster. Thinking about democracy, what could this entail? One idea, is that AI agents, like your digital half, could act in your place, in a direct democracy, in the digital space. In a direct democracy, people are asked to give their voice on issues, also between elections. That's of course great, your opinion matters. But a direct democracy like that, also leads to a great number of questions, for each individual to have a say on. Sometimes, you may not have a very clear opinion on things, or you don't have the energy, to find out everything about this issue, or you simply want to spend your time with something else. Then, your digital half, could make the choice for you. Your AI agent, never gets tired, and it could easily conclude, but you would have thought in the matter, had you taken the time to think through the options. By the help of your digital half, your voice will be heard, even in the case, when you choose to do something other than making up your mind about these matters. That seems convenient. But, what happens to responsibility, and democratic engagement, that comes with participation, if you leave it to an AI agent to decide where you stand? Who's really responsible, for decisions taken by AI agent in your name? That's something to think about. So far, this idea is not a reality, but it's important, that we always reflect on the consequences of technological innovations, and what society we want. Now we will look at some other ideas, on how AI tools could be used to develop democracy, ideas that are not as futuristic as AI avatar voting in your place. One idea, that has existing AI technology as its points of departure, being some preference learning, that's essential to recommender systems. Think for instance, of streaming services, giving personalized suggestions, of what music to listen to, or what films to watch. In a system of electronic voting, AI could make recommendations on what candidate to vote for, in a similar way. The idea is that an algorithm, could assist voters, by providing information, about candidates close to their preferences. Something it's learned by their online behavior. By reducing information overload, the democratic process would be improved, the argument goes. Another suggestion, is something its proponents call, liquid democracy, which is presented as something between direct democracy, where voters vote directly on issues, a representative democracy, where voters elect representatives to vote on issues for them. The point of departure for liquid democracy, is a direct form of democracy, in which every decision is submitted to the voters. Again, from a democratic perspective, it's a good thing that we have a chance to make an influence. But voters generally, don't have time or expertise, to give their preferences on everything. If we are expected to participate in every decision, we will have to spend a lot of time, considering all the options and making up our minds. Even if we really wanted to engage, in every question, we wouldn't have time to do so. Many people are not interested in participation, in politics between elections at all, other people are. This is where the representative part enters. In this liquid democracy, you can delegate your vote. Not to an AI avatar this time, but to another human. This is how it's supposed to work. As a voter, you can either vote directly, or choose to delegate your vote to some other voter. Let's call this other voter, John. John can in turn delegate his own vote and the votes that have been delegated to him, by you and by other voters, to yet another voter. Let's call this voter, Lisa. The process continues until an individual decides not to delegate further, but to be the one who makes the actual vote. Say, Lisa decides not to delegate further. Then, she has set voting weight that corresponds to the number of people she represents. Meaning, herself and the voters who directly or indirectly, they're joining this case, delegated their votes to her. Should you change your mind and prefer to vote directly yourself, you can do that at any time. Your vote is then reduced from the delegate, Lisa, in this case. Delegation is not definite. Your voice always remains your voice, even if you, at times, want to delegate it. All these counting of votes and voting weights and delegations and recaptures of delegations, are taken care of by AI technology. Perhaps most important, AI also helps you pick the delegates. You solely want to delegate your vote to someone holding views similar to yours. AI algorithms keep track of this, and depending on what the issue is about, the algorithm will suggest delegates that are inclined to vote as you would have done, had you chosen to vote yourself. Proponents of this AI-aided combination of direct and representative democracy, claim that it gives every members of a society an equal opportunity to participate in the democratic process. Some even call it the future of democracy in the digital age. But from a democratic perspective, there are also some concerns. For instance, how could you know that the proposed delegate will vote as you would have done? If delegates choose to delegate further, the repeated process of delegation may lead to a gradual shift of opinion of the delegates to finally votes, compared to your opinion. This is one of several technical difficulties with this liquid democracy, that researchers and developers are currently dealing with. But there are also problems of principle with this and other ideas of how AI could develop democracy. Proponents of using AI tools to develop democracy, often arguing for more direct democracy, referring to direct participation of people to be a stronger democracy, than the more practical representative models. But direct democracy doesn't always mean more democratic. Direct democracy tends to advantage groups with strong educational backgrounds and other resources, and maybe a disadvantage for underprivileged groups. The strength with representative democratic models is the built-in mechanism for accountability, through general elections. In a direct democracy, it's more difficult to hold someone accountable. In addition, proponents of using AI tools to develop democracy, take their point of departure in the worldwide digital network and thereby see opportunities for a global democracy, beyond the nation-state. Democracy at the global level certainly is something that democratic theorists discuss. Many Eastchester societies have to deal with, pollutions, terrorism, pandemics, migration, climate change. They don't stop at the national borders. We need to find ways to deal with these questions on a global scale, the argument goes. But why democratic models built on AI tools often see how software code, can be a way to escape the power concentration of the state, democratic theorists see a need for public institutions coordinating their action. We've looked at ideas on how AI tools could be used to develop democracy, some very futuristic, others based on existing technology. We've seen that AI tools could facilitate the democratic process, often in the direction of direct democracy. We've also seen that there are practical and principle problems with these ideas. But the aim of this lesson has been to point out that there is a lively engagement among developers, to make use of AI technology to improve democracy, not to dig deep into weaknesses of their still premature ideas. We will finish this discussion on AI and democracy in an optimistic and forward-looking way. Imagine what experts on AI and experts on democracy could achieve, if they collaborated to develop democracy with the help of AI tools.