4 ms·

Algo for teachers affected by cheating: 1) make questions brutal and unique (compared to previous year or current groups) if possible 2) make everything open-

by bitL 5y ago

Algo for teachers affected by cheating: 1) make questions brutal and unique (compared to previous year or current groups) if possible 2) make everything open-book 3) set a tough time limit 4) curve the results to a desired grade distribution (that is often imposed from above anyway) 5) use software that can the catch most obvious frauds (i.e. 360 degree room scanning, ban use of phones/second computers etc.) 6) if the exam requires algos, use software analyzing AST of the code produced to catch possible "cooperation", then decide on those cases individually (i.e. if an algo is common, then there is a high chance of very similar AST for non-cooperating people, but if it's unique, the chance is low)

wizzwizz4 5y ago
> i.e. 360 degree room scanning Tape a smartphone to the laptop screen. Out of view of the webcam at all times. (This isn't hard.) Such room scanning is harmful, and doesn't prevent cheating.
- op00to 5y ago
  The tests I’ve taken (amateur radio, tech certs) require a second device to show the rest of the first device.
  - bitL 5y ago
    Some tests I've taken required a 360-degree scan first including under-the-desk walk (to show no cheat sheets were glued anywhere) and then required a second device with a side-view of the desk I was sitting at. When I was taking an ETS test for a Stanford admission, they just required 360 room scan and the test itself only with a laptop camera.
ShroudedNight 5y ago
Do you have personal experience finding success with #6? My limited experience with detecting code plagiarism has been that detecting re-use that has been non-trivially obfuscated is a rather difficult problem.
- anonymousDan 5y ago
  For me what tends to happen is it becomes easy to catch the weaker students students who are cheating, since they often make weird/unusual mistakes that would be unlikely to occur in the same way (e.g. weird misspellings of certain keywords/variables).
- bitL 5y ago
  One of the Top 10 CS universities I studied at had some system for that, obviously it couldn't catch somebody who was a good programmer and could reshuffle AST with properly renamed variables. But it was still helpful for less-gifted students who cheated. Typically teachers had to set a threshold of similarity for their given assignments over which it was automatically reporting student names. The software scrapped GitHub as well.
noodlesUK 5y ago
I’m not sure about curving the grades. It creates a level of adversarial competition between students that I don’t think is healthy in an academic environment. Sometimes it can be necessary, but it’s not fair to punish people for having better peers than last year.
- sokoloff 5y ago
  In most cases, at least for medium and large courses (over 30) the curve is calibrating the exam contents more than the peer group.
  - noodlesUK 5y ago
    I agree with you on that point, especially with extremely large sample sizes (think national exams like SATs or A-Levels). My point is more that universities (in my opinion) should hope to foster a cooperative rather than ruthlessly competitive learning environment, and students know that curves mean that there’s effectively a quota of good grades available, and they’re competing with their friends for them. I feel this massively impacts peer learning dynamics.

by bitL 5y ago

wizzwizz4 5y ago

> i.e. 360 degree room scanning Tape a smartphone to the laptop screen. Out of view of the webcam at all times. (This isn't hard.) Such room scanning is harmful, and doesn't prevent cheating.

op00to 5y ago
The tests I’ve taken (amateur radio, tech certs) require a second device to show the rest of the first device.
- bitL 5y ago
  Some tests I've taken required a 360-degree scan first including under-the-desk walk (to show no cheat sheets were glued anywhere) and then required a second device with a side-view of the desk I was sitting at. When I was taking an ETS test for a Stanford admission, they just required 360 room scan and the test itself only with a laptop camera.

ShroudedNight 5y ago

Do you have personal experience finding success with #6? My limited experience with detecting code plagiarism has been that detecting re-use that has been non-trivially obfuscated is a rather difficult problem.

anonymousDan 5y ago
For me what tends to happen is it becomes easy to catch the weaker students students who are cheating, since they often make weird/unusual mistakes that would be unlikely to occur in the same way (e.g. weird misspellings of certain keywords/variables).
bitL 5y ago
One of the Top 10 CS universities I studied at had some system for that, obviously it couldn't catch somebody who was a good programmer and could reshuffle AST with properly renamed variables. But it was still helpful for less-gifted students who cheated. Typically teachers had to set a threshold of similarity for their given assignments over which it was automatically reporting student names. The software scrapped GitHub as well.

noodlesUK 5y ago

I’m not sure about curving the grades. It creates a level of adversarial competition between students that I don’t think is healthy in an academic environment. Sometimes it can be necessary, but it’s not fair to punish people for having better peers than last year.

sokoloff 5y ago
In most cases, at least for medium and large courses (over 30) the curve is calibrating the exam contents more than the peer group.
- noodlesUK 5y ago
  I agree with you on that point, especially with extremely large sample sizes (think national exams like SATs or A-Levels). My point is more that universities (in my opinion) should hope to foster a cooperative rather than ruthlessly competitive learning environment, and students know that curves mean that there’s effectively a quota of good grades available, and they’re competing with their friends for them. I feel this massively impacts peer learning dynamics.