-
Notifications
You must be signed in to change notification settings - Fork 0
Questions
This page is about in-game questions for the user, that make up the main flow.
Unbiased means no discrepancy between reality and answers.
Example: if there is a question type which has answers true/false at frequencies 50/50, it's biased as in reality you have much more false statements possible. Thus you miscalibrate the player, which is the opposite of the game's goal.
Balanced means that no knowledge should score maximum points with no-knowledge answer, wherein every option has the same probability.
E.g. let's say in reality 99% statements are false and we reflect it in the game, making a true/false questions with 99% of them having a "false" answer. Then user can see this and start answering 99%false every time and be perfectly calibrated!
To achieve this:
- Every question type used must have evenly distributed right answers. This must be guaranteed by form of question type, with any data and not by picking "right" data.
- No opinion questions, only factual.
It's theoretically possible to have different types of questions off the same data, but requires additional thought.
Objects have to be in the same group (e.g. celestial bodies, TV shows, psychological effects). Selection has to happen randomly. Can compare any number of objects.
In reality randomly comparing objects will yield you equal distribution of answers, which makes this type unbiased and balanced.
Examples
"Who of these billionaires is the richest? [Bill Gates] [Warrem Buffet]""Which of these cities has more skyscrapers(150m+ high)? [Tokyo] [New Work] [Dubai]"
"Which sorting algorithm of these is more time-consuming on mixed data with more than 10000 values? [Bubble sort] [Monkey sort] [Bucket sort]"
Structure
Examples
Structure
Examples
Structure
Biased: in reality the amount of false in such statements is exponentially bigger than true, so player will be miscalibrated.
Unbalanced: player, without knowing anything about subject topic, can put more probability on false and be reliably well-calibrated - unless questions are specifically balanced to have 50/50 true/false, which is ridiculously biased, and also requires constant moderating(impossible to guarantee).