With respect to modeling I’d go even further. Given the debate uncovered some problems with trust and techniques trying to contain it (I hope) and given that your Hackaton is going to take place in Prague in a week from now, I suggest you use the chance and prepare at least a theoretical basis of what you actually want in terms of at least minimally reliable system based on trust. A conceptual model. Further I’ll describe what should you focus your attention on. Then you have two possible paths.
First, to prepare a model and a mutation mechanism on your own. I don’t know whether there are members of your community capable of doing it, but if there are let them write first a simplified version during your event. Simplified, as the more interconnected dimensions are involved the more computing power is needed. But even with a simplified one you could possibly spot some regularities.
The other path is to have some professionals interested in it. Your event may be a chance to meet them. By professionals I mainly refer to academics. They may come from the Uni in Prague for instance. My suggestion would be to invite guys from MPI in Tübingen. In particular their “Intelligent systems” division. Of course, you can invite guys from different academic centers around. They often have resources and knowledge you don’t, maybe some will be willing to have a weekend trip to Prague. That approach is more reasonable, as any more complex model will require machines you probably have no access to. Also, a cooperation could provide you with a solution that at least has some merit. Not to mention a label of being scientifically substantiated.
But to take that path you must prepare something more than just a wish list.
So, let’s see what you need to have your problem being properly described.
Before you start reading further please see this clip, as I’ll refer to it.
First of all, decide what base system you’d like to model. I mean the most abstract one, the one that could potentially be a base for others, more use case oriented systems. In the clip I wanted you to see it would be a bipedal walker.
Next step should address defining a set of things the system must cover. So, whether there should be a rating (I assume that yes), should it address hierarchy and leadership (@AdriaanB), should it be as tamper resistant as possible, or only to a certain extent limited by usability, should it force behaviors or respond to them (or both), etc. Analogy in the clip - two legs, erected body, muscles attached to joints, locking joints, etc.
Next step - defining goals. It’s your fitness function. Analogy in the clip - it walks firmly, doesn’t topple, is able to navigate obstacles. In your case it could be the fact that a user feels some kind of satisfaction. But pay attention to the fact that you must be able to somehow measure it. So, maybe the time spent on using it? The number of interactions? It’s your job. But the set of values must be clearly numerical and as concise as possible. So, no social sciences’ mumbo-jumbo and no multitude of factors.
Next step - defining an environment. Your environment is our population. Because it’s complex, you must build a simplified model. In the clip these are bumps on a path. In your case these could be popularity fluctuations for instance.
Next step - defining your environmental threats. In the clip - boxes thrown at walkers. In yours - malicious actors, aspiring leaders, non diverse community, etc.
Next step - defining interactions between different elements of a puzzle. In the clip these are constraints and routing (dependencies) of muscles. In your case… So, we talked about that. Human behaviors aren’t very simple, but when broken down into atomic operations, as I tried to explain above, some factors are clearly dependent on others. So, for instance, an attitude toward a person, relation with that person impacts on our willingness to trust. It’s probably not linear, but there is a correlation. Maybe there are some papers trying to grasp a scale of that correlation. Look for them. There are other relations as well, for instance between stratification and satisfaction (hierarchy). And many more. Some were researched. Find numbers.
Next step - define a size of your population. Start from small numbers. Analogy in the clip - muscles and bones. Mind that you must take into consideration malicious actors as well.
Next you should decide which factors are constant, which are variable and which a GA should fill with optimal values. Analogy in the clip - the value to look after was a sequence of firing muscles (or its generalized model). You can prepare different scenarios with different constants, variables and values to be looked after.
Before you call it an end, one more caveat. Whenever possible try finding some figures in research papers. There are lots of papers focused on social issues, but focus on those that contain numbers and analytical approach. No mumbo-jumbo, remember? If there is no paper on a particular matter, look for things that may be related. For instance, trust isn’t deeply researched, but belief is. From social, neurological, biological perspectives. Another example - the wisdom of the crowd was researched in many papers. While not directly related to your use case, some rules and values might be useful.
After completing the above steps you’re more or less ready to prepare your model and its evolution. Evolution, because you’d like to find a form that works the best way possible (optimally satisfies the fitness function).
Then you just write a code, run it and depending on equipment and complexity wait from hours to weeks to see the result.
I address this post to all of you, notably to @Brooks who seemed to care about the subject, to @jakob.winter who certainly cares, to @AdriaanB who wants to solve at least one issue and to whoever is interested in the subject of this debate. You don’t have to like me, you can even be emotionally distressed by what I wrote so far, but this advice is free. There is nothing better than a free candy, right?