Learning to Cheat: Quantifying Changes in Score Advantage of Unproctored Assessments...
summary
Within a within-subjects study of N = 510 students taking 5 proctored and 4 unproctored exams on the same platform, a modified IRT/logistic model estimates an average 3.32 percentage points score advantage on unproctored items (p < 0.001), rising from ~0 to ~7 points by semester end. The dataset has 347 questions (245 shared) and 128k item responses; practice-exam submissions fall for unproctored tests while programming questions show the largest advantage, suggesting rising cheating and less study.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:20The need of proctoring is taken as granted
- 0:39What this work is about
- 1:51Setup of proctored exams and unproctored exams
- 3:05Exam schedule, and weights
- 3:26Towards a way to compare proctored exam and unproctored exam
- 3:58Exam randomization
- 5:22The logistic model, overall unproctored advantage
- 6:42Convert model coefficient to percentage points
- 7:49Capturing unproctored advantage over time
- 8:24Distribution of unproctored advantage became more positive
- 8:58Mean of unproctored advantage increased throughout the semester
- 9:29Evidence that this unproctored advantage is due to cheating
- 10:28Capturing unproctored advantage on different question types
- 10:58Unproctored advantage over time on different question types
- 12:42Limitations
related talks
41mHackuracy: Boosting AST accuracy through hacking
12mIncreasing security without decreasing usability: A comparison of various verifiable...
11mAn open door may tempt a saint: Examining situational and individual determinants...
21mOnline Pen Testing
19mReplication: No One Can Hack My Mind Revisiting a Study on Expert and Non-Expert
12mProspects for Improving Password Selection
23mWhat cyber security can teach us about COVID-19 testing
35mBreaking Boundaries with Adversarial Testing