think about preference elicitation in ScPo context

[floswald]

at wharton:

1. choose utility 0-100 for each course
2. can choose pairwise adjustment for courses that are preferrably taken together (choose 200 if very preferrable to take together, choose -200 if not preferrable at all)

do we need that? what else do we need that is specific to ScPo? may need to look at the ScPo report on coursematch for this.

[lsbosshart]

1. I see no reason why one should not adopt a similar preference range. Behavioral concerns aside, 0-100 is intuitive and people are 'familiar' with locating their preferences on such a
2. Building on the ScPo report which starts with a short elaboration on the class structure at PSIA:
   (a) Formation Commune: All ScPo Master students in their 1st and 2nd semester have to take one class each from a common pool --> clear preferences
   (b) Quantitative Skills: All PSIA students have to take one QS class, either in 1st or in 2nd-semester. Similiar to Wharton, one could adopt tokens (additional 'money') for 1st semester waiving the QS class in the 1st semester, which is then transferred to the second semester. This will ensure that people who need the class will get in.
   (c) Core Curriculum: The only case where there is a definite need for adopting a complementarity approach is the Core Curriculum. Students have to choose three courses. At Wharton, students are free to define as many bundles as they want. In the PSIA context, this should work just fine. But there might be problematic features due to cross-semester and -master availability. (i) For the International Security Master, for example, in the 1st-semester there are three distinct category pools out of which students have to choose three classes, each from a different category. 1st-semester students have their class choice entirely to themselves, as 3rd-semester students do not have any classes. Students in the 2nd and 4th semester share their Core Curriculum courses. In the 2nd- and 4th-semester they choose out of seven category pools --> As 4th-semester students are about to graduate, allocate extra tokens to their budget. (ii) Further, the classes are not exclusively available to the specific programs. CC classes of a certain master are often also offered in multiple concentrations or in multiple PSIA masters CC --> Allocate relatively more tokens to the CC budget than to the concentrations budget.
   (d) Concentrations: One regional, one thematic concentration. To be chosen before course registration. Within each concentration, students choose one class every semester. The amount of options varies per concentration and can range from 1 to 29. Assuming one leaves out the possibility of reporting complementarities between regional and thematic concentration classes and establishing the primacy of students for which the class is part of their CC, the problem becomes self-contained.
   (e) Language: Only relevant to the extent that it might block a certain timeslot --> Follow the rule in place of having two rounds. Students will internalize their given timetables for the other classes into their decision for their language class --> clear preference

Bottom line: The class registration problem can be broken down into smaller self-contained problems. Problems of intersectionality, as in the case of CC and concentration classes, might be solved by allocating additional tokens to the more important 'stakeholder'.

[JulieLen]

I completely agree to that, the problem can be broken down into smaller problems, which will make our job easier.
Each and every school in Sciences Po seems to have its own specificities, we have to be careful to make the problem sufficiently general so that any specificity of one school can be easily implemented (for instance in Public Affairs there are no classes that are common to 2nd and 4th year student).

Moreover I was asking myself a question :

• Should we allow student to express preferences on each and every couple of class ? If we do, should we allow student to express preferences only on couples of electives, or can they also express the fact that they want this "Tronc commun" class and this elective ?
• Or should we limit the number of cross-preferences they can express ?

[lsbosshart]

Good question concerning how many couples a student should be allowed to define. In general, I'd say that a richer set of information will yield a better fit for the student in question. However, one concern could be that students spend a lot of time on defining a lot of couples. To save them from this work, one could limit it.

Regarding the second question related to cross-complementaries between Tronc commun and CC: I think that the value added of it is marginal and it makes things more complicated.

[floswald]

can this issue be closed?

[lsbosshart]

If there is no veto, yes.