​London Voice is a democratically built AI, trained by the real opinions of real Londoners, designed to serve as a collective voice for the community. In a world where AI often sparks debate and concern, London Voice aims to demonstrate the positive potential of technology when placed into the hands of the people. It offers a new way of collecting and interpreting vast amounts of information while ensuring maximal privacy, security and full ownership of data for its users.

A citizens' assembly is a group of citizens, typically chosen to be representative of a larger population, that comes together to deliberate on a specific issue or set of issues. In the context of London VOICE, the citizens assemblies are designed to engage diverse groups of people in discussions about he issues that directly impact their lives and communities. The events are intended to be inclusive, transparent, and democratic, allowing participants to have a meaningful voice in shaping the ideas, issues and dialog . Through facilitated discussions and informed deliberation, we aim to generate diverse perspectives and insights that can inform an AI data set in order to create London VOICE and AI bot which will reresent the voice of London. We believe in the power of citizen engagement and the potential for citizens assemblies to foster collaborative, inclusive, and effective governance.

London is known for its rich tapestry of cultures, religions, and opinions. With diversity comes complexity and truly understanding and addressing the needs of its inhabitants is of crucial importance. Traditional methods of gathering public opinion are often skewed by the agendas of those who commission them and are costly to administer. London Voice seeks to change that narrative by providing a platform built by Londoners, for Londoners, ensuring that every voice is heard and considered in the democratic process.

London VOICE is the output from this democracy experiment. By building a collaborative environment with the community and technologists, the community will be presented with some LLM chat bots to explore and provide feedback to. This feedback can be further used to dial in the responses of the community to the LLM, aligning it with the thoughts and feelings of the group. As London VOICE develops out in the open, it will be progressively trained to generate a series of proposals for projects that the community thinks will improve London, and we will use Web3 technologies to explore if and how to get these funded. The input and the output of the AI bot will be gated through Web3 technology to the participants. They, and they alone, can decide whether to open the access to others.

If you attend one of the citizen's assemblies or join the online experiment you will be given a unique access link to the London VOICE web-app. You will onboard using an email address (or a web3 wallet if you already have one) and claim your unique, non-transferable voting token. The voting system uses web3 technologies to gate the app and it's contents to those who have proven the are Londoners and have claimed their token. This stops outside parties from changing the results away from the true consensus or event viewing them. All Londoners are invited to participate. They will be asked to prove they are a Londoner by turning up to one of the citizens assemblies, or through social validation from the Factory Labs team of helpers. The vote begins with a submission phase. Here, Londoners are invited to share 3 issues that they care about e.g. Cost of Living. These suggestions are compiled into a feed, where they will be voted on by the group to sort by perceived importance. The top 12-15 issues will be pushed to the final vote on May 2nd, where every voter will have a budget of voting credits to spend. Voters will spend their voting credits on the issues that they care about. There is an increasing cost to vote on each issue. This limits the “strongly agree / disagree on everything” effect we see in conventional survey design and promotes people to rebalance their opinions relative to all the other issues. At the end of the vote, everyone is asked to provide their detailed opinions on why they care about those issues. The result is a sorted list of issues generated by a broad consensus from the voting community attached to a body of language data that can be used to train an AI.