There’s a big difference between what a person wants and what they need. All electoral wisdom is bottled down to what a person wants – the results of that kind of democracy are on display – Brexit, Trump, and God knows what’s next. The irrational part of the people is panicky, highly influenced by personal characteristics and easy to manipulate.
So, here is a crazy idea: two-step representation or avatar democracy. Imagine you have a personal avatar in a cloud somewhere. The personality of the avatar is created in three ways: observing your behaviour (big brother style), influencing your avatar by sharing your thoughts with it and your social image – what people are thinking of you. The avatar is your representation (first step) in a simulated society. Any politician trying to get an office must present a platform of intentions, which will be translated (by AI) into a model of new/modified rules of economics, social justice, etc. Now having all the avatars in place, the new model of the society could be played and estimate how “happy” the avatars are with the changes and most importantly which politician would make most people (avatars) “happy”. That way, AI playing all this will decide which platform is a winner.
Most of the system details, like: What is happiness? Does the degree of happiness matter and how it would influence the total score? How far ahead does should AI look to model the effect of changes? An election could be triggered by reaching a certain level of un-satisfaction.
The system seems to be highly corruptible (too much power to AI) and not a real democracy, but that is the only way I can see to achieve what democracy intends to do, but compensate for most of the individual stupidity (irrationality), which if not addressed will reach a point of self-destruction bigger than AI threat.