Improving College Room Banding

Main problem with college ballot system at the moment is the room banding and pricing. The JCR claim to have ‘fixed’ it by giving each room a set of variables (size, ensuite/no ensuite, floor, etc.) and then have assigned some weights to each of those variables to form a linear equation. Each room is hence given a score and the bands and band prices have been arbitrarily chosen to cover some of these scores such that the total rent is equal to what is being demanded by the college accommodation office. 

Several problems with this:

Variable weights are clearly very subjective - for example some people may really want a big room on the top floor but don’t care about an ensuite whereas for other people an ensuite may be essential and they may be indifferent to what floor they are on. Hence the JCR just deciding what these weights should be feels arbitrary and unfair to the student body.

The number of rooms bands and prices of each room band also feel arbitrary and unfair. 

Potential fixes: 

Behavioural method - bidding system. Have thought about this any would maybe work if student body prepared to put in a bit more effort, but where it collapses is with syndicates. Cannot think of way to get it to work given that many people put a lot of value on staying with their syndicate.

Perceptual method - i.e. surveying students. This could work. See method outline below.

Perceptual method:

Survey some of the students.

Give them a choice of a certain number of rooms (3 rooms might be good), with the attributes of each room clearly laid out, like on the college ballot system.

Continue to assign a set of variables common to all rooms (size, renovation date, floor, ensuite/no ensuite, etc.).

Initially randomise all the weights associated with these variables, such that the scores for each room are fairly random (weights are initially randomised to avoid bias in model, but might get better results if used initial distribution that is the same as current distribution). Generate an initial score for each room based on these randomised weights.

Let students choose which of the n rooms they would pick, if they were first in the ballot.

Update the model weights accordingly given the choice the student has made. 

Repeat this process for multiple room selections for each student and repeat for large number of students.

Final weight numbers are what should be used in the pricing model. Introduce a new weight, w, such that score*w = room price, so that the total sum of all the room rents equals the amount demanded by the college accommodation office. 

Thoughts:

Think this is the best way to do it rather than assigning a price to each room based on score and then updating model in real time and feeding back the results to students as they make next room choice. Want to keep model simple. However, surely there should be some sort of mechanism to take student input on prices? - i.e. to work out how the room price affects the student tendency to go for a particular room.

First step would be to think about how to get the information from the ballot to a website where you can gather student input.

Would students try to game this system by trying to bias model preferences to something they don’t care about? They could try, but hopefully would get a large number of students to participate so would have little effect. Also may be ambiguous/not obvious sometimes which room they should choose in order to try to bias the model.