Souls is a side project I started hacking together a few months ago. It’s basically an MMI friend matcher (MMI tinder for some). It takes your MMI measurements, matches them up with other users in the database and those with the closest cross-correlation score can chat with each other if they want to.
The iOS/Android apps use CameraRNG and for now web users use some pre-generated entropy. The project is more end-user oriented right now in terms of its wording/description rather than as a technical experiment/research project.
Technical issues are not important in an “end-user” oriented test. At some point real time MMI generator data will most likely have to be used to get real results. There are certainly a lot of possibilities on how to produce a template for matching people. I can imagine an n-dimensional space with coordinates generated from the end points of n random walks and taking – in the simplest processing method – the distance between each person. Distance is just the square root of the sum of the squares of the differences between two people for every dimension in the n-dimensional space. See: Euclidean distance - Wikipedia
I can also think of generating each dimension individually in response to a number of questions about personality or interests as the user intends the result to represent the answer to each question. Then calculate the overall distance using the distance formula. This could be more engaging to the users and give more specific results.
Thanks for the feedback. Interesting ideas on ways to possibly match up people. The thing with a system like this is it really does depend on a large user base and users chatting with each other and sharing their experiences for us to really be able to validate the system. Will see if it grows…
How about to make a bit of benchmark of that system?
I was thinking a lot, how to test that algorithm and see if it really help to find matching mind.
So, I thought about 2 types of metric:
Length of text between matched persons. If minds are connected, probably, they have something to discuss, right?
Topic match. That might be kinda subjective, but we can count common interests between persons, which we found.
I actually had an idea like this but didn’t implement it. You can see a sneak preview in the screenshot of it on the webpage below. I was envisioning some MMI-selected topic from a list of topics, and then start the ball rolling between two people with “Suggested Topic Chat”. They could decide to make it public and maybe build up a group of users talking about that topic and we could compare their collective MMI measurements. Then users could see trending topics and stuff. All basic social network type elements to foster the growth of human connections.
I mean lenght on chain of messages in comments or cumulative lenght of personal messages.
Probably, we shouldn’t invent bycycle and try to find social media analytic for consultation. There are algorithms how to find match without MMI. So, we can try to use them to measure effectiveness of MMI.
