Professional Documents
Culture Documents
Abstract—This paper presents design guidelines for CS discipline [e.g., 10, 11, 12], with promising results
implementing a free-selection peer review protocol. “Free- reported.
selection” (FS) refers to the ability of students freely access all Researchers stress the fact that peer review offers to
available peer work and choose which of them to read and students the chance of developing a range of skills important
review. A series of two studies on the free-selection protocol in the development of language and writing ability, such as
has provided evidence on the efficiency of the method. In the meaningful interaction with peers, new perspectives on the
First study, the FS protocol was compared against the widely writing process, and a greater exposure to ideas [13, 14].
used assigned-pair one, where students work in instructor- Based on that, we focused our attention on how to enhance
defined dyads. In the Second study, further issues of the FS
the peer review method by increasing even further students’
approach were evaluated, with our attention focused on
students who, due to the freedom element of the protocol, do
exposure on new perspectives and ideas.
not receive reviews. Both studies paint a very promising As a reference point we used the widely used “assigned-
picture of free-selection. However, several issues were also pair” (AP) protocol. While there are studies that assign
raised on how to effectively apply such a protocol. As the use of multiple reviewers to a single peer [15, 16], in the context of
technology is necessary in the FS approach, we provide in this our study we use the term “assigned-pair” to refer to a more
paper the design implications derived from the two studies common research design where students in a author-reviewer
regarding, (a) peer work availability, (b) non-reviewed works, dyad are assigned exclusively to each other. Students can
(c) student population, (d) peer work visual representation, (e) play both roles and review each other’s work. The overhead
peer work length, (f) presentation order, (g) double-blinded for the instructor is well contained and the activity is
approach, (h) peer reviews availability, and (i) students’ straightforward for the students. This design, of course, is not
approaches in selecting peer work for reading and reviewing. without drawbacks. One review for a single peer might not
be enough for the author to get valuable comments and
CSCL, peer review, free-selection, assigned pair, technology- suggestions for improvement. Additionally, the benefit for
enhanced learning, computer network education the reviewer may be limited, since she gets just one more
point of view (the author’s). Finally, the method requires a
I. INTRODUCTION level of stability in the dyads formation throughout the
Peer review is an instructional method aiming to help activity, while a bad pairing may have negative results.
students elaborate on domain-specific knowledge, while Our focus was to (a) enhance the learning benefits of
simultaneously developing methodological review skills. peer review for the students, without increasing the
Typically, in a peer review cycle, an author drafts a piece of minimum amount of work that they had to do, and (b) keep
work which is then evaluated by a peer. The evaluation or instructor’s overhead low. Towards this direction, we
critique is carried out anonymously on the basis of explicitly decided to investigate the efficiency of a “free-selection”
defined criteria and is subsequently returned to the author. (FS) protocol, where there are no dyads and students are free
The author is free to review his or her final draft based on the to browse all peer work and select what to review.
given critique. Yet, when practicing peer review in the We tested the effectiveness of the FS protocol in a series
classroom the instructor has a number of alternative design of two studies conducted in the same context. Results were
selections to choose from. very promising suggesting that indeed such a protocol can
The literature abounds with relevant studies indicating enhance the learning outcome. However, the implementation
that the method is popular among educators inspired mainly of the FS protocol requires the use of technology and is more
by the constructivist and socio-constructivist paradigms for complex from its AP counterpart. Through the two studies,
learning [e.g., 1, 2, 3] who want to challenge their students to we were able to identify factors that can affect the impact of
think critically, synthesize information, and communicate the FS approach.
science in nontechnical language. The method has been used In the following, we present (a) information on the two
extensively in various fields [e.g., 4, 5, 6, 7, 8, 9] and in the studies on the FS protocol, (b) useful design implications for
implementing the protocol, and (c) conclusions on the
approach.
approach to circumvent this bias. Other representations may focus only on answers appearing in the first places. As a
be used, having in mind to present all submissions as equally good measure we applied a simple algorithm to randomly
important. The students had to click on an answer in the grid present the answers. We used students’ username in the
to view the whole answer text. system as the algorithm seed so that each student would view
The second design guideline is to give students the a different answer grid that would remain the same each time
chance to get a rough idea of what each answers says, the student logins in the system.
without reading the whole submission. In our case, the Usage analysis showed that both the readings and the
knowledge domain was Network Planning and Design, and reviews were widely spread throughout the grid and this was
students included characteristics of the network designs they one of the reasons we had so few non-reviewed answers in
proposed even in the first sentences of their answers. Thus, the First study.
including only the first 3-4 lines of each answer in the grid
boxes was usually enough to give students a glimpse of what G. Double-Blinded Review to Move Focus on Submission
their peers were suggesting (Fig. 1). Content
In both our studies, students’ answers were usually a We applied a double-blinded review process in both our
little less that a whole page. In settings were students studies. Our goal was to analyze students’ choice in selecting
produce lengthier deliverables, additional information should which answers to read and review when they had to base
be given to the students. For example, in case the students selection solely on the answer text. This allowed us to avoid
have to submit papers expanding in several pages, abstracts any biases. In the First study, we asked students whether they
should be required so that they can be presented in the grid. wanted to know the identity of the authors they reviewed and
The goal in FS is to present shortly the main points of each of their reviewers. Students answered using a 5-step Likert
answer without having the students reading the whole text. scale (1=No, 5=Yes). Students in the Assigned-Pair group
That way, they will be able to take an informed decision on were divided regarding the authors’ identities (M=2.50), but
which answers to read and which to skip. most of them wanted to know who commented on their
answers (M=3.43). On the contrary, students in the FS group
E. Preferred for short deliverables did not care about others’ identity (M=1.36, M=1.55,
It was very important that the students’ answers were respectively). Thus, one could choose to apply a non-blinded
relative short, rarely filling a whole page. This was approach. We believe, however, that students’ behavior
commented by the students in the interviews, saying that it could change in case peers’ identities are freely available.
was easy to read a lot of answers because of their length.
One should expect that as the answers become longer, the H. Control Access of Submitted Reviews
possibility of students reading more of them gets lower. This, In our studies, we chose to present the review comments
in turn, diminishes one of the main purposes of the FS only to the respective authors, thus controlling students’
approach, which is to make the students learn more by interaction. In that way we were able to better control the
analyzing different opinions. variables under investigation in each study. However, one
could opt to make the review comments public, thus
F. Random Appearance of Submissions allowing deeper interaction between the students. This could
One guideline that is connected to the previous section is affect the impact of the FS approach in several ways.
to randomly present the answer into the answer grid to First, students are better informed of each other’s
enhance the spread of reviews, even if the students decide to perspectives, since each review represents in a sense the
opinion of the reviewer on the subject. This can be beneficial REFERENCES
for the students. [1] Topping, K. (1998). Peer assessment between students in colleges and
On the other hand, available reviews may affect future universities. Review of Educational Research, 68, 249-276.
reviewers’ judgment. The first review may act as the [2] Falchikov, N. (2001). Learning together: Peer tutoring in higher
dominant one, while it is not likely that students would education. London: Routledge Falmer.
choose to submit conflicting reviews. It is more probable that [3] Liu, C. C., & Tsai, C. M. (2005). Peer assessment through web-based
students will submit converging reviews, and in that sense knowledge acquisition: tools to support conceptual awareness.
reproduce same comments and suggestions. Innovations in Education and Teaching International, 42, 43-59.
Finally, having the reviews publicly available diminishes [4] Falchikov, N., & Goldfinch, J. (2000). Student Peer Assessment in
Higher Education: A Meta-Analysis Comparing Peer and Teacher
the chance of non-reviewed answers. It will be clear which Marks. Review of Educational Research, 70(3), 287-322.
answers have already been reviewed and which ones are left [5] Dossin, M. M. (2003). Among Friends: Effective Peer Critiquing. The
non-reviewed. Thus, students will be able to provide one Clearing House, 76 (4), 206-207.
review to each answer. [6] Anewalt, K. (2005). Using peer review as a vehicle for
communication skill development and active learning. Journal of
I. Factors that Affect Students’ Selection on Selecting a Computing in Small Colleges, 21(2), 148-155.
Submission for Review [7] Turner, S., Pérez-Quiñones, M.A., Edwards, S., & Chase, J (2010).
Base on usage data and interview statements in the two Peer Review in CS2: Conceptual Learning. Proceedings of
studies, there are two main criteria students apply in SIGCSE’10, March 10–13, 2010, Milwaukee, Wisconsin, USA.
selecting which answers to review. Half of the students try to [8] Liou, H. C., & Peng, Z. Y. (2009). Training effects on computer-
mediated peer review. System, 37, 514–525
avoid conflict and select answers that, according to them, are
of high quality. This way the students are able to provide [9] Goldin, I. M., & Ashley, K. D., (2011). Peering Inside Peer Review
with Bayesian Models. In G. Biswas et al. (Eds.): AIED 2011, LNAI
complementing reviews saying only nice things for their 6738, pp. 90–97. Springer-Verlag: Berlin.
fellow students. On the contrary, another half of the students [10] Crespo, R. M., Pardo, A., & Kloos, C. D. (2004). An Adaptive
selects answers that, according to them, present obvious need Strategy for Peer Review. Paper presented at ASEE/IEEE Frontiers in
for improvement. This way it is easier for the students to Education Conference. Savannah, GA.
write more meaningful reviews. Of course, students’ [11] Ziu, E.Z.F., Lin, S., Chiu, C.H., & Yuan, S.M. (2001). Web-based
opinions about high and low quality answers are not always peer review: the learner as both adapter and reviewer. IEEE
in line with the instructor’s judgment. However, these two Transactions on Education, 44, 246-251.
ways of thinking are, most of the times, behind students’ [12] Demetriadis, S., Egerter, T., Hanisch, F., & Fischer, F. (2011). Peer
review-based scripted collaboration to support domain-specific and
behavior. Other factors that, according to students, affect the domain-general knowledge acquisition in computer science.
selection process are the length of an answer (medium size Computer Science Education, 21(1), 29-56.
answers are usually preferred), and how clear an answer is [13] Hansen, J., & Liu, J. (2005). Guiding principles for effective peer
(students in general refrain from reviewing answers that they response. ELT Journal, 59, 31-38.
do not understand). [14] Lundstrom, K., & Baker, W. (2009). To give is better than to receive:
The benefits of peer review to the reviewer’s own writing. Journal of
V. CONCLUSIONS Second Language Writing, 18, 30-43.
Results on the Free-Selection approach have been [15] Tseng, S., & Tsai, C-C. (2007). On-line peer assessment and the role
of peer feedback: A study of high school computer course. Computer
encouraging so far, and as the availability of the necessary & Education, 49, 1161–1174.
technological tools makes the implementation easy, we [16] Tsai, C.-C., & Liang, J.-C. (2009). The development of science
expect similar approach to gain researchers’ attention. In the activities via on-line peer assessment: The role of scientific
two studies, we tested a generic FS format. However, as we epistemological views. Instructional Science, 37, 293-310.
presented here, there is a number of factors that can be [17] XXXXXX (2012). Computers & Education.
modified and affect the learning outcome. The 9 implications [18] XXXXXX (in preparation).
presented in the previous sections can provide a useful
picture to the instructor who wishes to implement similar
peer review methods.