Malcolm K. Sparrow Page 2 11/3/2023
students who share the same ethnicity (assuming we prefer heterogeneous
groups).
students assigned to the same dormitory suites, or sharing accommodation (who
therefore spend time together anyway).
students who are known to be particularly talented, or particularly weak (and
where any clustering within groups along lines of ability might upset the overall
balance, making some groups potentially dominant, or potentially vulnerable).
“ad-hoc conflicts,” e.g. two or more students who just don’t get along, or for
whatever reason need to be kept apart!!
1.3: Important Note: Sensitive Data. Student Data is, of course, confidential in any
case. By defining conflicts as described above, administrators may well be creating new
data of a highly sensitive nature, incorporating judgments about ability or character.
Even the underlying choices about which variables to use or not to use as the basis for a
group assignment process could be contentious. Defining ad-hoc conflicts may be
particularly sensitive. All spreadsheets containing such data, including the GRumbler
itself, should obviously be treated as highly confidential. Such data will exist within the
GRumbler (unless you clear it out), and within any Workbooks that the GRumbler creates
(i.e., results files). The detailed instructions below will remind you when and how to
safeguard this information.
1.4: Capabilities and Technical Constraints. This version of the GRumbler:
allows users to define up to 20 different types of conflict, to be used
simultaneously.
allows class sizes up to 5000 students.
permits division of the class into between 2 and 2500 groups.
can create either gender-balanced or gender-clustered groups, assuming that data
on students’ gender is available.
allows for different weights to be applied to different types of conflict.
2
permits use of the TurboGrumbler for faster search times, as an optional extra.
See section 9 below for details and installation instructions.
1.5: Multiple, Sequential, Group-Assignments. Students can be assigned to one set of
groups for exercises (say) on Monday. Then, for Tuesday’s class, new groups can be
formed that have minimal overlap with Monday’s groupings. And so on, for Wednesday,
Thursday, etc., or for different projects or assignments. If it is mathematically possible,
the GRumbler will avoid assigning any two students to the same group today if they have
been together before, in any group, on any previous day.
3
This version of the GRumbler
can generate up to 50 sequential Group Assignments.
2
By default, all conflicts will be scored equally (value 1), but users may choose to apply greater weights to
particular types of conflict, in which case the algorithm will give these priority attention.
3
When searching for a new grouping, the GRumbler algorithm treats all earlier group assignments as a
source of additional conflicts. It also continues to recognize and accommodate the set of conflicts
originally defined by the user.