Raw LLM Responses

I think another strong pillar in any such system MUST be an evidence-based attitude to everything. Facts are what we can agree on, so this system should encourage a transparent logical and factual foundation for proposals, arguments, ideas, discussion, etc. You could embed links to supporting evidence in specific units of text (specific claims, perhaps). That much is easy, and a good design would encourage adoption of this kind of methodology (and really, it's very much like the traditional method in any scholarly article). You could also maybe create an AI that evaluates the probability of certain claims, taking in curated archives of factual information and this system's generated, meta-opinions. Various concepts and techniques from Machine Learning would be very helpful in sorting through the torrent of data and help individuals make informed choices much more quickly. So, while typing this out, I thought of a scenario where all the above could really shine: Let's say you wake up and go to this worldwide, open-source, transparent, constantly evolving web of information that has clear links between facts and arguments. You've setup your personal "dashboard" or "assistant" to sift through recent developments in your policy interests. There's a list of about 10 discrete units representing only the most important developments. Each has associated opinions and threads of arguments and discussion, video feeds, picture feeds, audio feeds; all of them with a weighting that is decided by an AI that ranks things according to their plausibility. The entire system is constantly receiving human input around the world, so it really would be a voice of the people, but also clearly rational, organized, and meaningful. So, now you would have a good, easy way to consume meaningful information while also participating in policy discussions. Also, let's say that this system would allow the accumulation of policy reports, complete with facts, public polling, the whole nine yard