Group-Work and Assessment
Group work is a desirable part of the undergraduate physics curriculum because, inter alia:
- more more achieved collectively through co-operation than as isolated, or competing, individuals;
- students' learning benefits through social interaction
- employers value evidence that potential employees can work successfully work as a member of a team;
- it makes efficient use of scarce and/or expensive resources.
In order to achieve the fairest outcome for all students  mixed-ability groups are preferred.
The higher-ability and/or harder-working students have their greater contribution
recognised in their individual mark so that they are not penalised by being obliged to work
with lower-ability students.
Increasing group size is known to decrease individual motivation . Students also normally lack the management skills
to organise large groups and to cope with the increased incidence of 'social loafing' in them . Groups of four to
six students work best for undergraduate projects. Groups of eight have been found
to experience significant problems, e.g. when agreeing decisions, tracking progress, allocating tasks, and
coordinating multiple activities. Groups with even numbers of students are preferred because Physics programmes give student a lot of
practice at tackling tasks in pairs.
Numerous studies  have shown that group work can improve student performance, engagement, marks and retention.
To achieve its potential, however, the assessment mechanisms must reinforce appropriate student behaviour. The literature
yields three common approaches to assessment:
Minimalist: All the students get the same grade unless there is a complaint that one has not been pulling their
weight, in which case the tutor decides what to do.
Abdicatory: Award the students collectively the group mark multiplied by the number of students
and let them the students divide up the marks between them as they see fit.
Algorithmic: combine the students' ratings of each others' contributions with the group mark to arrive at individual marks.
Sharp  presented an algorithmic method with a statistical basis for deriving individual student marks for a piece of group work
from a single 'tutor' mark and ratings which students make of each other's contributions.
The method incorporates a mechanism for directly controlling the size of the adjustments made.
The original spreadsheet discussed by Sharp performed the the calculations necessary to
apply the method and inspired the version that we currently use, which is available from the link below .
Sharp made the following assumptions:
the tutor and not the students assess the overall quality of the work and that
this has been awarded a percentage grade by the tutor (i.e. this paper is not concerned
with peer assessment of quality but peer perception of contribution);
the students and not the tutor are in a position to assess the contribution
of each student and for each student a single number has been determined
to represent that student's contribution;
the students should not be asked to evaluate their own contributions;
each student's evaluation of each other student's contribution should be equally
weighted and input to the calculation independently and secretly.
and we use the following procedure, based on his recommendations:
- The final individual marks must lie within ±10% of the tutor's mark.
- A significance level of 25% is set for the statistic A, representing
the the strength of the evidence which must exist that there were differing
contributions between different the students before any changes are made.
- Students ratings:
should be made independently in secret (whether they follow injunction is ultimately a matter for them),
do not include self-assessment,
- The tutor mark and ratings from each group are analysed using the spreadsheet .
This yields values of final student marks and the statistic A.
If there is no variance in the ratings then A is undefined and each student
receives the tutor mark. Otherwise, the value of A is compared to a critical value corresponding to a user-specified significance level.
The table below gives some example values for reference purposes. If the observed value is below the critical value, then the spreadsheet
parameter φ is set equal to zero for that group and again each student receives the tutor mark.
Otherwise the individual marks calculated are examined to verify that they lie within
the preset limits decided in advance. If they do not, the value of φ applied
for that group is reduced until the range of individual marks is satisfactory.
Further details, notes and examples are provided with the spreadsheet linked-to by reference 5.
|Group size, N||Critical value of A|
Table 1. Critical values for A at the 25% significance level.
Group assessment in Systems Analysis and Design: a comparison of the performance of streamed and mixed-ability groups.
M. Lejk, M. Wyvill and S. Farrow.
Assessment and Evaluation in Higher Education (1999) 24(1) pp 5-14.
Dispensability of member effort and group motivation losses: Free-rider effects.
N. L. Kerr and S. E. Bruun.
Journal of Personality and Social Psychology (1983) 44(1) pp 78-94.
The assessment of group work: lessons from the literature.
G. Gibbs, ASKe, Oxford Brookes University (2009)
http://www.brookes.ac.uk/aske/documents/Brookes%20groupwork%20Gibbs%20Dec%2009.pdf [Accessed October 2011.]
Deriving Individual Student Marks from a Tutor's Assessment of Group Work.
Assessment and Evaluation in Higher Education (2006) 31(6) pp 329-343.
Spreadsheets for Assessing a Group Project Using Sharp's Method (v6)
C. D. H. Williams (2017).
We are grateful to Dr Stephen Sharp for permission to repackage and publish his Peer Assessment of Student Groups spreadsheet
(2004) licensed under a Creative Commons
Attribution-NonCommercial 3.0 Unported (CC BY-NC 3.0) License.