Measuring the impact of instant high quality feedback presented at the 5th International Assessment in Higher Education Conference. Stephen Nutbrown, Su Beesley & Colin Higgins, 2015.
Measuring the impact of instant high quality feedback.
1. Measuring the impact of
instant high quality feedback.
Stephen Nutbrown: psxsn6@nottingham.ac.uk
Su Beesley: susan.beesley@ntu.ac.uk
Colin Higgins: colin.higgins@nottingham.ac.uk
University of Nottingham & Nottingham Trent University
2. Background
oAutomated assessment
oWhy do we assess work?
Assessment of Learning vs Assessment for learning
oExample - teaching programming – a practical subject, assessment for
learning is vital.
oFeedback core part of assessment for learning
3. Overview
oStudy of 141 Computer Science students
oAccording to the NSS, relative to other questions, students unhappy
with their feedback.
oWhat is good feedback?
oIntroduction to TMA (Automated marking system) to produce
feedback in line with ‘good feedback’.
oMeasure student performance – same exercise + multiple submissions
oMeasure student performance on following exercises without multiple
subs
4. Aim: Improve feedback, measure the result
oHelp student’s learn!
oGeneral properties of good feedback :
o Informative (How to improve!) & specific
o Reliable & Consistent
o Clearly communicated
o Timely
o The assessment (and feedback) should be useful for teachers
5. TMA – A brief introduction
(The Marker’s Apprentice)
Electronic framework for combining automated, semi-automated, manual marking.
Mark Scheme broken into many parts, a tool for each part which may be:
Automated:
e.g Spell checks, word counts, grammar checks for reports
functionality tests + convention tests for programming.
Semi-automated (help a human marker):
e.g Selecting options from a rubric or bank of feedback, each with
actions on how to improve.
Manual (entirely up the marker):
e.g Free text entry
6. Instant?
oIf all of the tools for an assignment are automated, feedback can be
given instantly.
oAlso allows for marking of drafts, or using a subset of tools for a
“pre-submission” to give an idea of how the student is doing.
7. Use in Computer Science
oSeveral thousand submissions so far at University of Nottingham UK
and University of Nottingham China.
oA range of automated tools have been developed which work with
TMA for the assessment of programming.
9. Rule based conventions tool
oOne of several tools – but this is the one we will focus on today.
o127 rules, based on PMD (http://pmd.sourceforge.net)
oSearches for particular patterns.
oCan identify common problems and their exact line number.
oFor each violation marks deducted & feedback given
o Difficulties due to repeat violations, weightings, difficulty.
10. Tool example - conventions
Extremely difficult to manually mark – imagine thousands of lines of
code each submission, for 140+ students.
Bad code, example with common issues Good code, example without common issues
int a = 99;
int b = 10;
int c;
if(a < b) c = b-a;
else if (a > b) c = a-b;
else c = 0;
int first = 99;
int second = 10;
int calculatedDifference;
if (first < second) {
calculatedDifference = second – first;
} else if (first > second) {
calculatedDifference = first - second;
} else {
calculatedDifference = 0;
}
11. Linking back to good feedback
Good example (How it should be done)
Bad example (Help identify the problem)
Reason
Hyperlink back to learning resource (Lecture slides)
Feedback given instantly
Perfectly consistent
Feedback sessions to discuss – keep communication open
Used in combination with functionality tests.
14. Programming is just an example. The same is
possible for reports, or other assignment types.
15. Programming – Coursework 0
First measurement:
oDoes not count towards final grade, as many submissions as they wish.
getting started with Java.
oInstant feedback (automated) makes feedback on drafts possible.
oAssessed using functionality and conventions tools from before.
oMeasure performance between submissions
16. Measuring improvements
within one exercise
CW0, does not count towards the final grade.
346 Submissions from 60 unique students (Avg. 5.7 each)
60.8
77.2
0
20
40
60
80
100
AverageGrade
First Submission, Last Submission
First submission
Last Submission
17. Results raise some questions
oDid they read the feedback and engage with it?
oDid they just mechanically fix problems?
oDid they learn anything?
Should fix
line 7. See
lecture 2
Submission 1
Well done
Submission n
Nothing
18. Survey of students (35 responses)
Did you read the instant feedback?
Yes (100%)
Did the feedback highlight areas which can be improved?
Yes (91%) No (9%)
Did you improve the quality of your submission based on your
feedback?
Yes (77%) No (23%)
Do you feel the feedback assisted in your learning and will help you in
future work?
Yes (68%) No (32%)
19. Coursework 1
oCounts towards final grade (100%!)
oSame coursework set 2 years ago.
oMade up of 3 parts:
o Part 1 – Ensures every student has received specific, instant feedback at least
once from TMA using functionality and convention tools..
Allowed 2 submissions.
o Part 2 – No pre-submissions. Compare results to previous years
(This part had a focus on their testing of their own work)
o Part 3 – 1 pre-submission [a subset of tests] containing very few tests, for
peace of mind.
20. Avg. Number of convention
mistakes
0
5
10
15
20
25
30
35
Avgconventionviolations
Cohort
Previous Year
(118)
This Year(141)
35.4% difference
22. Analysis
oSame course material
oSame lecturer
oSame assignment
oImproved feedback. Specific, instant, clearly communicated, timely,
useful for teachers too.
23. Conclusions
oThe feedback generally helped students to learn (assessment for
learning)
oThe feedback, as it was specific and broken down into sections, is
useful for teachers.
oThe assessment technique is vitally important to student experience
and has a huge impact on student learning
oSimilar findings in China, not formally analysed.
24. General: For other disciplines
This study strongly highlights the importance of good feedback and
well-considered assessment techniques. Students created submissions
of much higher quality than before.
TMA almost forces good feedback practises, even for semi-automated
tools
◦ Require action
◦ Split up mark scheme – specific
◦ Saves time
25. To take away
oFeedback is extremely important, considering all of the guidelines is a
good start, even if it isn’t automated.
oAutomated assessment may help, but is not a replacement for human
interaction.
oCan you do anything to improve the turnaround time or detail of your
feedback?
oTMA will be made generally available soon, if you would like a demo,
please feel free to ask.
Editor's Notes
Assessment of learning – measuring stick
Assessment for learning – use assessment as a tool for assisting in the learning process. Feedback is a key part of this, particular for practical subjects.
First question: What is good feedback?
TMA
Measuring performance – automatically marked, multiple submissions.
NSS:In particular, “Feedback was prompt” and “Feedback on my work has helped me clarify things I did not understand.” (NSS 2014)
Informative – example, “Well done, however to get a first you must make your essay sparkle”.
Reliable – Manual marking example, many markers, some with good feedback, some without.
Clearly communicated – handwriting
Timely – Studies showing sooner you get the feedback, more likely to engage with it. Forgetting what was done.
Useful for teachers.
Online submissions through a webpage.
Split into many parts. Each part has a tool associated with it.
Split into two main parts.
Functionality – Does it work. Test this by putting some input into their program and seeing what comes out.
Is the source code good quality – This is more difficult to test, but we want to give them feedback on this aspect too
both of these parts may also be broken down further, we may have many functionality tests and also many different tests for the quality of the sourcecode.
As we would expect.
Probably read their feedback.
Practiced.
Previous year (2012/13) – general feedback comments which were manually written. For example “Try to improve your style”.
Note that these are assessed in the exact same way using the new system.
Would not be fair to compare grades as they were assessed differently, so we compare the absolute number of convention violations.
Assessment technique and style of feedback – broken down into many parts means it’s easy to see where students are doing well or badly.