Normal Source Test (100 questions, 10 sources) about confidis

[seveibar]

I think a test definition format could be helpful here:

{
  "numberQuestions": 100,
  "sources": [
     { "accuracy": 0.9 },
     { "accuracy": "0.7 + t * 0.1" }, // t is [0,1], corresponds to number_answered / total_answers
     { "accuracy": 0.5, "chanceOfAnswering": 0.3 },

     // If accuracy is an array, it corresponds to the chance they'll answer in each biased group
     { "accuracy": [0, 0.1, 0.7], "chanceOfAnswering": 0.3 },
     // ...
   ],
  "numberAnswers": 1000 
}

There are additional complexities that could be added, for example difficult questions, chances of answering different questions in different ways, number of unique answers etc. but this is a good start