SP-Rank-Dataset / README.md
amritpuhan's picture
Update README.md
be1cd94 verified
metadata
license: apache-2.0

The dataset contains questions from three domains - Geography, Movies, and Paintings where multiple options are shown to voters and they are asked to provide two information - 1) Their vote on what the ground truth ranking of the alternatives is 2) Their prediction on what everyone else thinks the ground truth ranking of the alternatives is. The criteria for ranking is decreasing order of population for Geography, decreasing order of gross box office lifetime earnings for Movies, and decreasing order of auction prizes for paintings.

Dataset attributes:

workerid - Individual voter ID

problem - Problem number each voter answered

treatment - numerical representation of elicitation format

domain - Geography=1, Movies=2, and Paintings=3

questions - Question number form the global number that the voter responded to

options - The options presented to voters. Note that while being presented, this is randomized. But the way it is presented in the dataset can be used as ground truth ordering.

votes - Vote of each voter on what their own opinion is about the ground truth ranking

predictions - Vote of each voter on what they predict everyone else thinks about the ground truth ranking

elicitation_format - elicitation format name