student attitudes towards partially automated peer grading ... · • th4 year undergraduate course...
Post on 19-Jul-2020
1 Views
Preview:
TRANSCRIPT
Student Attitudes towards ���Partially Automated Peer Grading ���
in Mechanical TA
Jessica Q. Dawson ���James R. Wright���
Kevin Leyton-Brown
BACKGROUNDCPSC 430 – Computers and Society
• 4th year undergraduate course – focus on critical reasoning about social implications of
computational advances
• focus on frequent short, frequent writing assignments – effective way to teach writing skills [Seabrook et al 2005]
– provides many opportunities to practice
• students complete a 300-word essay every week – Problem: inefficient and expensive for manual TA
marking – Solution: peer grading
BACKGROUNDPeer Grading and Mechanical TA
• in peer grading: students grade each others’ assignments
• peer grading often negatively perceived by students – tend to believe lower quality/less fair than TA grading
[Kaufman & Schunn 2011]
• a solution: Mechanical TA (see companion poster for details.) – software system for partially automated peer grading,
developed by CPSC 430 course staff – TAs remain in the loop:
• mark essays/reviews before students graduate to ‘independent’ • for ‘independent’ students: manage appeals and spot-checks
– results over 3 offerings found evidence that MTA helped improve student learning and grading ability [Wright et al. 2015]
MEASURING PERCEPTIONS OF MTAResearch Questions
• what are the students’ perceptions of the peer grading? – how do students perceive the quality, appropriateness,
fairness helpfulness and accuracy of: 1. reviews they gave their peers on their writing, 2. reviews they received from peers
. . . and how to did these compare to TA reviews?
– how helpful was the calibration (built in practice reviewing) and peer grading and in MTA for learning?
Data Collection
• End-of-term survey conducted in CPSC 430 (2014 W1) • n = 76 (response rate 83%)
Reviews students gave their peers
The majority of students rated the reviews they wrote favorably with respect to each factor.
11%
22%
5%
17%
24%
66%
58%
61%
49%
53%
22%
17%
34%
30%
17%
0% 20% 40% 60% 80% 100%
were accurate with respect to the grading rubric.
would have been helpful when they wrote future essays.
were fair.
were appropriate in tone.
were of high quality.
Q: The reviews I gave my peers on their writing . . .
Strongly disagree Disagree Neither agree nor disagree Agree Strongly Agree
RESULTS
Reviews students gave, compared to TAs
The majority of students the reviews they wrote were about the same as how a TA would have reviewed the same paper.
16%
31%
4%
17%
28%
72%
55%
75%
68%
58%
5%
12%
19%
11%
9%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
accuracy, with repect to the grading rubric
helpfulness, for wriHng future lectures
fairness
appropriateness in tone
quality
Q: Compared to a TA review, the _________ of the reviews that I gave my peers
Much worse Somewhat worse About the same Somewhat beKer Much beKer
RESULTS
Reviews received from peers
The students’ perceptions of the reviews written by peers were more mixed.
7%
12%
7%
11%
31%
28%
24%
18%
28%
35%
26%
43%
36%
38%
22%
28%
18%
36%
16%
5%
5%
8%
8%
7%
0% 20% 40% 60% 80% 100%
were accurate with respect to the grading
were helpful when I wrote future essays
were fair.
were appropriate in tone.
were of high quality.
Q: The reviews my peers gave me on my writing . . .
Strongly disagree Disagree Neither agree nor disagree Agree Strongly Agree
RESULTS
Reviews received, compared to TAs
For most factors, majority felt their peers’ reviews were worse than how a TA would have graded the same paper.
11%
16%
15%
10%
18%
49%
48%
35%
22%
39%
25%
25%
32%
49%
31%
11%
5%
14%
14%
8%
4%
5%
4%
6%
4%
0% 20% 40% 60% 80% 100%
accuracy, with repect to the grading rubric
helpfulness, for wriHng future lectures
fairness
appropriateness in tone
quality
Q: Compared to a TA review, the __________ of the reviews I received were
Much worse Somewhat worse About the same Somewhat beKer Much beKer
RESULTS
Helpfulness of peer reviewing for learning
Perceptions of helpfulness were mixed for activities asked about. For learning content and concepts, a small majority found peer review somewhat or very helpful (57%).
11%
12%
8%
17%
17%
14%
28%
22%
21%
33%
37%
36%
12%
12%
21%
0% 20% 40% 60% 80% 100%
Improving WriHng Ability
Obtaining large amounts of feedback
Learning content / concepts
Q: Please rate the extent to which you found peer reviewing helpful for the following ac9vi9es:
Very unhelpful Somewhat unhelpful Neither helpful nor unhelpful Somewhat helpful Very helpful
RESULTS
Helpfulness of calibration for learning
Majority of students found calibration helpful for all activities asked about – considered most helpful for activities specifically tied to reviewing.
5%
4%
5%
5%
18%
16%
25%
32%
55%
43%
48%
48%
20%
35%
19%
12%
0% 20% 40% 60% 80% 100%
obtaining a lot of pracHce reviewing
becoming an independent reviewer more quickly
improving the quality of your reviews of peers' essays
improving the quality of your essays
Q. How helpful was calibra9on [built in prac9ce reviews] for . . . .
Very unhelpful Somewhat unhelpful Neither helpful or unhelpful Somewhat helpful Very helpful
RESULTS
CONCLUSIONand possible next steps
• Students feel positively about their own reviewing ability, perceived to be similar to TAs. – calibration helpful as expected for learning how to review.
• But many still doubt peers’ abilities. – even though course staff also satisfied with reviewing ability. è how can we bridge this gap?
• Perceived helpfulness of peer reviewing for learning and improving writing was mixed. – could adjust types of feedback reviewers expected to
provide e.g., more qualitative and focused on writing skills.
References
• Seabrook, R., Brown, G. D., Solity, J. (2005). Distributed and massed practice: from laboratory to classroom. Applied Cognitive Psychology, 19(1):107–122.
• Kaufman J.H. & Schunn C.D. (2011). Students’ perceptions about peer assessment for writing: their origin and impact on revision work. Instructional Science, 3, 387–406.
• Wright, J.R., Thorton C., Leyton-Brown, K. (2015). Mechanical TA: Partially Automatic High-Stakes Peer Grading. Proc. SIGSCE 2015, 96–101.
top related