Talk:Monster Girl Quest: Paradox/@comment-25601391-20151119153346/@comment-26418405-20151120195932

Little bit of Comp Sci, since I like teaching people. I'm pretty confident that Akinator works because of a field in computing known as 'Machine Learning' (specifically a 'neural network') which basically relies on a constantly building table of statistics to predict what the person is thinking of: https://en.wikipedia.org/wiki/Machine_learning

It's usually something used to predict future events, based on patterns in the past but it can be applied here by noticing patterns people follow when answering questions. That's why it sometimes asks weird questions, because a lot of the patterns haven't been fully defined yet and it's still trying to figure out the best set of questions to get the answer in the quickest way, by slightly adjusting which questions it asks for each pattern it detects.

So it's basically a field of Statistics and Computing.

The reason if knows MGQ characters is because, at some point in the past, someone tried for an MGQ character, didn't get it and then if Akinator can't figure one out, it asks you to tell them about it and give a pic of it. This then adds it to its database and it starts figuring patterns for people playing in the future. It'll get it wrong a few times at first but from these mistakes it realises what question orders bring up what results and updates itself each game.

'spretty cool. Have some fancy stackexchange talk about it: http://stats.stackexchange.com/questions/6074/akinator-com-and-naive-bayes-classifier