Initial Feedback #1
Replies: 3 comments
-
|
Jack wrote via email: Dear Sam, Rebecca, Michaela, and Conor, (cc’d to the committee to avoid duplicating comments) Thank you for your hard work in putting these changes together — it’s greatly appreciated! In the current form, I’d be likely to vote for the price control change and against the new ballot system, but this is based on particular aspects of the changes/implementation rather than the overall ideas. While these changes do have benefits to the MCR community, there are also downsides to. I would like to suggest that the vote is made on these changes in a later meeting or online poll (e.g. Doodle/Google Forms to privately vote), to allow for a useful discussion and time to make any necessary amendments to these (although I appreciate that this needs to be decided on very soon). The two votes cover lots of details, and I’d appreciate some time to reflect on these. I’ve included my comments below on this, and I’d be grateful for any feedback you have. I apologise if it unintentionally comes across a little bluntly, as I’ve quickly put this together before the next meeting and haven’t had much time to proofread it. Also, if it would be easier to turn this into a google doc to address each part, do let me know. Kind regards, A. MCR asking College to set the prices of individual rooms. This is a good step forward to making College more responsible to rent issues. In regards to rent increases, why has the increase not been shared among all bands (i.e. applying a X% increase evenly across all rooms instead of bands having different % increases)? I understand that previously we have increased lower bands more than higher bands to keep accommodation affordable overall, but this does not work long-term, and causes unexpected rises. However, I would like to stress that we should approach this with some caution, as it’s possible that we end up with College setting prices on our behalf, while continuing to "pass the buck" back to us if we make changes to these. We also don’t know if College would let us return control back to us in future if we give it to them now. B. Changes to MCR committee benefits I have mixed feelings on these changes. Much of this has been historical: President/secretary as a reward for the role, Welfare officers due to running events and being near most students, and the Bar Treasurer to be able to quickly receive deliveries for the bar when asked to by the Porters. I agree that the president and secretary should continue to get first pick, and the treasurer added to the list. I also agree that the welfare officers should get a priority in the ballot but not necessarily first pick. However, I don’t agree with what this means. I’m aware I have a conflict of interest here, but I would like to point out that welfare officers should at least be near the site — e.g. having Men’s Welfare living at Rook Road near Addenbrookes is not sensible. Also, I don’t agree with the wording of “a lot less likely to be kicked off the ballot”, which is a result of the how the new system works (kicking people off to improve fairness). C. Changes to timing of ballot These are sensible changes to make if we use the new system D. Changes to priority system (+ change of code) Historically, the priority system is setup such that first years should get rooms on site, second years may get rooms on site or in the hostels, and third years onwards may get a room or not. The ordering of the priority groups ensures that those staying on will get a room as long as there is space in one of their preferences. The new system may “kick” people from the ballot if it improves overall fairness. I strongly disagree that we should do this: the MCR committee should work to ensure that we can provide accommodation for as many students as possible, and not remove people from the ballot if it helps to improve overall fairness. I agree that the MCR should improve fairness — this is very important to improve wellbeing — but not at the expense of removing MCR members from the ballot who would have otherwise been offered a room. I don’t agree with the change that the system in regards to first years: the selection process should ensure first years will get accommodation, rather than assigning a likelihood of getting something. This would turn the priority system into a weighting system, and completely overhaul this. Overall, I don’t agree with these changes to the priority system. While the previous ballot could be understood by all of the MCR community, you need an understanding of linear programming to properly reason about how the new ballot works. In addition, you could risk not being offered accommodation based on the ordering of your preferences rather than room availability. The new system will improve overall happiness, while putting those in priority group 4 at a higher risk of not being allocated a room. I think we should prioritise (a) that as many members as possible get offered a room, and (b) that allocation is as fair as possible, rather than prioritising (b) first and (a) is a side-effect. I don’t think the new system should kick people from the ballot. However, I think there is definitely room to use the code with the ballot: could we instead use this code to assign rooms within the priority groups? We currently select randomly within these groups, and using this new approach could provide a fairer assignment. I’ve sent across a previous email with two plots comparing the ballot systems, which show significant global changes, but this could be more useful within priority groups. E. Filling outer hostels I agree that we should encourage existing students to live in the outer hostels. However, this is tricky as the outer hostels have not been renovated in quite a while, and are not very appealing to live in. Could the MCR provide more details about the outer hostels to members before the room ballot, such as noting that they are not that far away despite their name (“outer”), and mandating that members select 4 or more outer hostels (rather than the current 2). The argument is made that the new ballot code will force the outer hostels to be filled. This opposes the idea that the new system will improve the wellbeing of members, since they will have to live in a hostel room which they have put last. Also, it would be tricky to apply a mechanism to the previous system too, for favouring outer hostels. F. Restricting the number of room options I don’t agree that we should reduce the number of options. By reducing the options, members may be less likely to be offered any accommodation as lower preferences will now not be part of the assignment process (i.e. room Z is available but Alice only had 6 options and therefore had to drop Z from her top 10 rooms). I’m not sure what “honey-pot” rooms are referring to here? |
Beta Was this translation helpful? Give feedback.
-
|
Sam wrote via email: Dear Jack, Thanks for taking the time to respond to this, especially with regards to investigating the code. Don’t worry about coming across as blunt, the primary reason for sending this out now is to get as much honest feedback as possible. I can’t speak to all the details of the code but I’ll briefly address your points individually. This is my personal response and I’m not speaking for Rebecca, Michaela or Conor in this. “Honey-pot” rooms are very popular rooms in nice houses. The idea was to ensure students select some less popular rooms to avoid oversubscription of the very popular rooms which is presumably behind people being kicked off the ballot. A - I’m not 100% sure as to the reasons behind historical discrepancies between bands. My guess is that the band system is fairly new and so it’s taken a few years to get to a reasonable distribution. Trying to avoid unequal changes is the reason behind suggesting the within 2% of the mean, while allowing for the fact that we want the total to be as low as possible within the bounds set by the bursar. With regards to college setting prices for us, they currently set the total price for us which is where the real power lies. My goal is to set in place a rigorous, documented system, possibly with approval of college council, which will formalise the system as being a collaborative effort. Whether this is possible remains to be seen. B - The issue of welfare living on site is in my mind, Rebecca has suggested allowing welfare officers to pick a list of rooms all of which are on site. This should avoid this issue subject to the kicking off the ballot which I’ll return to later. D - My understanding of how the code currently works is that anyone kicked off the ballot is still guaranteed a college room, just not one of their Top X choices. Obviously kicking people off and asking them to reselect is not ideal. With regards to first years and living onsite, this is something that has been noted but I believe is difficult for Conor to implement. I completely understand your concerns but I also agree with Rebecca’s argument that the current system of assigning all Priority 1 rooms before any Priority 2 can lead to non-optimal solutions. E - I agree that outer hostels are much less desirable but I don’t see an obvious solution to this when college want as many rooms to be filled (including the outer hostels). Moving forward we can certainly lobby for refurbishments on them but this is not a short term solution. The rationale behind requiring them to be filled is that it means no new students (who are not in the ballot) will be assigned to them and will instead live on site, something I think is very important. I think this has happened in the old ballot as well. F - I think this is something which should be played around with by running mock ballots and finding an optimal solution. I do agree though that increasing the number of choices should reduce the potential to kick people off the ballot. If I’ve understood correctly, most of your concern lies in the priority system and the cost-function in the code. The current suggestion is only a first draft and I am intending myself to take a closer look at it before Sunday. My hunch is that alteration of the cost-function (for example by heavily penalising kicking people off the ballot) will solve a number of these issues. I hope this is answers some of your questions. It might be useful to start a google doc to discuss the priority system in more detail. Best wishes, |
Beta Was this translation helpful? Give feedback.
-
|
Conor and Rebecca wrote via email: Dear Jack, Thank you very much for your comments! We really appreciate the time that you have spent going through our proposals. We think this really does need discussion. We’ve done this far in advance of the ballot to allow for iterations. We would like to start by making clear that the ballot should almost never kick anyone off unless there are more people than spaces. To address your points one at a time: B. This is a point for discussion in the committee meeting. Which exact positions should get preferential treatment in the ballot? We have discussed guaranteeing welfare accommodation on-site and can be trivially implemented. D. The cost of kicking someone from the ballot is so high that it almost never improves the overall fairness. If it does happen, the cost function needs to be tweaked. It is worthwhile noting that “kicking” people out is an inevitability of the system. If everyone applies for the same rooms or if there are more people than rooms, some people have to be kicked out. In the old code, students were prevented from being “kicked” out of the system by being randomly allocated a room. We would like to emphasise that this was a room that they did not pick. In our system, students would be asked to re-ballot for a room if there were rooms remaining. This means that they would get to choose their room. We anticipate that either: there will be more people than rooms, thus we will not need a second ballot; or there will be fewer people than rooms, in which case the code will assign everyone a room. In our simulations thus far, people never get kicked off the ballot unless the number of spaces are limited. This is a consequence of students picking from a very large number of rooms and thus significant collisions of choices are rare. We have done several tests of the code on past year data and we have never had a first year kicked out of the ballot. In your plots, we are assuming that you limited the overall number of rooms using the -m flag? If this was the case, we think the new code has produced a better distribution. If this was not the case, the cost function needs to be tweaked. If you send us the data from last year, we can attempt this before the meeting. In our opinion, the concept of first years is being incorrectly applied here. The first years have all spent at least one year at Churchill. In Conor’s case - four! We should not treat these students the same as those actually new to the College. Lastly, we disagree with your statement: “the previous ballot could be understood by all of the MCR community”. Does this include the mechanism, which assigned students a room that they did not pick closest to their top priority’s cost, or the limited (and incorrectly implemented) annealing step, or the hidden cost function that linearly weighted their choices? Most of the MCR community were not familiar with the mechanism, as it was not circulated to students. E. To be clear about the problem with hostels. The new ballot system is so good it gives everyone one of their top 3/4 choices (and most people get their first). This results in none of the hostels being filled as they are undesirable. If we make the decision, as a committee, that we WANT to fill the hostels the code can do this (fairly) with an extra command line argument :) It is our opinion that we should make this decision, as students new to the College (who are not in the ballot and also the majority) should get the full Churchill experience. F. This is a point for discussion in the committee meeting. It may be useful to have more choices. We have no strong feelings about this and - if it would rectify the Committee’s concerns about the code - we are happy to continue with 10 choices. — If you have any further comments, please do not hesitate to get in touch! Best, Rebecca and Conor |
Beta Was this translation helpful? Give feedback.
-
Jack wrote via email:
Hi all,
Thank you Connor for working on this!
I’m sharing the following two plots, as sharing an image in a meeting isn’t too easy, and it’s useful to visualise the changes here.
This compares the existing ballot (sorting by priority order) to the new ballot — the key takeaway is that priority groups 0/1/3 are less likely to get their top choice, in order for priority groups 2/4 (and overall) to get a better chance at their top choice (and this has also kicked 5 P4 people from the ballot).
(I have a few more comments on the changes, but it’ll be best to leave these to the meeting.)
Best,
Jack
Beta Was this translation helpful? Give feedback.
All reactions