All 2 entries tagged Plagiarism
View all 9 entries tagged Plagiarism on Warwick Blogs | View entries tagged Plagiarism at Technorati | There are no images tagged Plagiarism on this blog
October 12, 2005
I am allergic to jargon! It can be used to
- conceal the lack of substance or meaning in an idea
- inflate the commonplace or trivial
- exclude the unititiated
- confer pseudo-scientific authority and claim false superiority
So what about Latent Semantic Analysis and its inevitable contraction to LSA? I came across it in an interesting conversation with Mike Joy about current issues in e-learning and assessment. LSA is a statistical method designed to measure the commonality of meaning in a collection of text passages or documents. It compares the frequency of significant words, numerically conflates their meanings, applies some mathematical jiggery-pokery to the data (viewed as sparse matrices), and comes up with some numbers that may indicate how close the texts are in meaning. It can be used as an alternative to the more familiar comparison of strings (a la Google) in detecting likely plagiarism. I believe that it is used effectively in monitoring plagiarism in program source code submitted for assessment by students in the Department of Computer Science.
Clearly jargon is both necessary and useful to experts, and LSA meets this test. It also has the virtue of meaning what it says: the analysis of hidden meaning. It might be interesting to run LSA on this and other blogs on plagiarism!
October 11, 2005
I got an encouragingly lively reception when I climbed Gibbet Hill on Friday afternoon (a walk I miss since Mathematics moved to Central Campus) to talk to the Biological Sciences Undergraduate Teaching Committee about CAA. The discussion ate 15 minutes or more into their precious meeting time but brought me some positive outcomes:
- Two first-year modules were identified where CAA would help, and suitable materials already exist for one of them
- CAA could play a useful role in the continuous assessment of several second-year modules
- The departmental Chair, Professor Robert Freedman, did not entirely rule out the possibility of freeing up staff time to develop CAA resources
Two of the discussion points were:
- The value of using paper-based data-capture methods for continuous assessment
- The risks of plagiarism
I will discuss data capture in a separate blog but will make a couple of points here about the risks of cheating in computer-aided assessment when it is used in a 'summative' mode, that is to say, a mode where the assessment directly affects the outcome of a student's degree (for example, yielding marks for course credit or determining progression).
First some general points:
- The level of security in exams only needs to be proportional to the weight of the assessment
- A vanishingly small amount of course credit seems to get most learners engaged
- Credit representing say 10% of a first-year module affects at worst the second decimal place in the final percentage used to determine a student's degree class
- Students who cheat in continuous assessment shoot themselves in the foot; they are usually exposed in the final exam
- No system of invigilation offers 100% security against cheating
Now a few practical suggestions for raising the level of security in computer-aided tests and exams used in summative mode:
- Generate many variations of each question type and permute them so that each student answers a different question paper; several assessment packages are good at this for numerical questions
- When a student logs on to a test, display his or her university card on the table and screen to make identity confirmation easy
- A couple of random visits by an invigilator to the computer room during a 50-minute exam is a good deterrent
- Install CCTV in computer suites used for assessment
- Disable access to other computer applications and directories while the exam is in progress if the style of the questions makes this necessary (to prevent students using, for example, the internet or some mathematical sofware to answer the questions)
Other suggestions please.