Open-Book Exams & AI Proctoring: Balancing Integrity & Flexibility
                     Open-Book Exams & AI 
Proctoring: Balancing 
Integrity & Flexibility 
Open-book exams (OBEs) and AI-driven proctoring are reshaping assessment 
design by trading memorization for application while trying to preserve 
academic integrity. This blog explains how institutions can combine flexible 
OBEs with ethical, privacy-aware AI proctoring to deter misconduct, improve 
learning outcomes, and meet regulatory expectations. 
 
Why Open Book Exams 
(OBEs)? 
The pandemic accelerated a shift from locked, closed-book exams toward more 
authentic assessments — and open-book exams (OBEs) are now mainstream 
because they test higher-order thinking rather than recall. 
Research shows OBEs reduce anxiety and encourage application and analysis, 
and they can improve performance for lower-performing students when well-
designed. But flexibility brings new risks. Multiple meta-analyses and 
systematic reviews of online assessments report high rates of dishonest 
behavior in remote exams. 
 
Pooled estimates find that roughly 40–45% of surveyed students admitted to 
some form of cheating in online exams, with some variation that is discipline- 
and context-dependent. That scale has prompted institutions to rethink both 
aAst stehses msaemnet  dtiemsieg,n t haen dg lmoboanli toonrliinge  teoxoalsm.  proctoring market is growing fast as 
universities, credentialing bodies, and certification providers invest in tools that 
use AI to detect suspicious behavior (eye movement, multiple faces, device 
 switching, audio cues, and browser anomalies). 
Industry reports estimate the market was worth about US$836M in 2023 and 
could approach US$2B by 2029 (≈16% CAGR), while other forecasts show 
even steeper trajectories depending on automation adoption assumptions. 
This growth reflects demand for scalable integrity solutions — but also raises 
privacy, bias, and reliability questions. 
 
Strengths and Limits: What AI Proctoring Can and Cannot 
ADIo p roctoring helps scale monitoring, provides audit trails, and frees human 
proctors for edge cases. For standardized, time-bound assessments, it can 
reliably flag many anomalies that warrant review. However, regulators and 
watchdogs are issuing warnings: some authorities now say AI detectors can 
miss or misclassify behavior (and that certain AI-enabled cheating is “very 
hard to detect”), pushing universities to diversify assessment approaches 
rather than doubling down on surveillance alone. 
False positives (legitimate behavior flagged as cheating) and fairness 
concerns—especially for neurodiverse students, those with unstable internet, 
or different cultural norms—mean AI flags should trigger human review and 
clear appeal routes. Transparency about algorithms, data retention, and 
student rights is essential. 
 
Best-Practice Blueprint: Design + Tech + Policy 
1. Design Assessments For Application, Not Retrieval: Make questions 
scenario-based, problem-solving, or open-ended so answers require reasoning, 
even with materials available. OBEs work best when questions are novel and 
require synthesis rather than look-up. 
2. Use Staged Monitoring, Not Blanket Surveillance: Combine non-
invasive AI detection (browser-lock, screen capture, keystroke heuristics) with 
targeted live review for flagged cases. Keep human-in-the-loop adjudication 
mandatory. 
3. Prioritize Student Privacy And Fairness: Publish data retention policies, 
offer alternatives for students with accessibility or tech constraints, and allow 
appeals. Consider on-campus secure assessments for high-stakes parts as 
recommended by some regulators. 
4. Invest In Academic Integrity Education: Honor codes, formative low-
stakes practice OBEs, and explicit AI-use policies reduce the incentive to cheat 
and create shared norms. 
5. Measure And Iterate: Track metrics: flag rate, human-confirmed cheating 
rate, false positive ratio, student complaints, completion success, and learning 
outcomes (e.g., mastery on later assessments). 
 
Technology Trends and Evidence 
● Market Momentum: proctoring vendors are adding explainable-AI 
fepartiuvraecsy, -by-design modes, and differential-proctoring (mixing live and 
recorded proctoring) to reduce friction. Industry reports show robust 
CAGR estimates and wide geographic adoption across certification and 
higher-ed markets. 
● Research: contemporary trials comparing open- vs closed-book formats 
fintde sOt BaEnsx ileotwye ar nd encourage deeper learning when combined with 
crafted questions — but also show score inflation risk if questions are 
only knowledge-based. 
● Policy: national and institutional regulators increasingly recommend 
diversified secure assessments (oral, in-person practicals, proctored 
components) rather than sole reliance on remote proctoring. 
 
Conclusi
on 
Open-book exams paired with ethical, transparent AI proctoring can deliver 
both flexibility and integrity — but only when assessment design, technology, 
and policy work together. Institutions should prioritize application-focused 
OBEs, layered and explainable proctoring with human review, clear privacy 
safeguards, and ongoing evaluation of outcomes and fairness. 
With the online proctoring market and AI tools maturing rapidly, the smartest 
path is a hybrid model that uses AI to assist, not replace, human judgment. 
Contact EnFuse Solutions to start a privacy-first, integrity-focused 
assessment transformation today. 
  
Read moRred:ucing False Positives In AI Proctoring WIth LLM-Based 
Contextual Analysis