final assignment research prorosal detecting and preventing cheating in the exam by ai

Background of the study Detecting and Preventing Cheating in the Exam by AI is a project that aims to use artificial intelligence AI techniques to enhance the quality and integrity of on

Detecting and Preventing Cheating in the

Exam by AI

Course Name: IT Research Methodology Student : Phạm Hải Đăng - 20070701

name/ID ~ ,

Nguyên Lâm Tường - 20070799 Hanoi, 12/2023

3 Purpose and signfficance 0ƒ the SÍH(ÏJ HH HH nà HH HH HH 3

4 Research questions and objectives cocci ccc ttt Hà nà HH HH HH HH tệ 3

Fl U.ng8 7g ng ốốốố.ố.ốố.ố.ố.ố.ố.ốốố 12

1.1 Traditional Miethods qnd Theïfr' LIHÍQfÍOHĂ ác nh nh HH HH nh xxx ra 12

1.2 Al-Based Cheating Detection: Addressing the ŒĐS LH HH TH HH này 12

1.3 Key AI Techmiques for Cheating Ï)€f€CfÏOfH cá nh HH HH HH HH 1 HH 13 1.4 Challenges and Ethical COfisideTQflOHHS, à scnnnnnnHH HH nH H HH HH HH ng 13 1.5 Moving Forward: Responsthle ImpÏ€I€HIGIÏOH SH HH HH HH Ho 14

2 Previous studies and findings relafed to the ÍOĐC à SH HH HH Hà Hà ớu 14

3 Gaps in the literature and the need for the proposed SÍHđÌÿ cua 15

4 Theoretical framework or conceptual framework 0000c cocci tices ttt ku 17 TIL Method ology ccccrccccvccccvevccvevccvevscvevecvevccvevecvcvecscvecccvecscvccsccccsccscscvscsssseel

DA CAN LAY | cr 18

2 Data collection methods and instrument 0000 ccc ccc HT HT HH TH Hà k1 1x rủ 19

3 Data analysis techniques ng 20

TV Referemces ccccoscscsccccsccsecccscccsccccsccsecccscccsccccsccssecescccsccccscessecessccsesscscessesse dl

I Introduction

1 Background of the study

Detecting and Preventing Cheating in the Exam by AI is a project that aims to use artificial intelligence (AI) techniques to enhance the quality and integrity of online exams Online exams are becoming more prevalent and convenient in the era of the COVID-19 pandemic, but they also pose significant challenges and risks for academic honesty and validity Cheating behaviors

in online exams can compromise the credibility and reliability of the assessment, and affect the

student’s learning outcomes and future opportunities Therefore, this project seeks to develop and evaluate an AI system that can detect and prevent cheating in online exams, and compare it with human proctors and traditional methods

2 Statement of the problem

In this day and age, Information Technology appears in many fields and one of them is education The education sector in Vietnam has been making progress recently, but cheating in exams is still an unanswered situation There have been many solutions and laws put forward to overcome cheating but it is inappropriate and impractical when many students despise it and they often argue back after destroying the evidence Information technology is seen as the right and heavily tested solution to prevent fraud Up to now, there are a number of software to respond to online exam cheating (EduNow) Besides, students of some famous universities such as Industrial University are proposing and perfecting products to fight offline exam cheating and this is also the product that I will study the purpose and reason about its birth

3 Purpose and significance of the study

The purpose of detecting and preventing cheating in exams by AI is to ensure the integrity and fairness of the assessment process Cheating undermines the credibility of exams and can lead to unfair advantages for those who engage in it AI can be used to detect cheating in a variety of ways The use of AI to detect and prevent cheating has several advantages over traditional methods, such as human proctoring AI can examine enormous volumes of data rapidly and efficiently, whereas human proctors cannot Personal ties or other things that might influence human judgment do not impact AI Regardless of their history or circumstances, AI may apply the same rules and standards to all pupils

The use of AI to detect and prevent cheating is not without its challenges One concern is that

AI may be used to unfairly target certain groups of students Another concern is that AI may be used to invade student privacy It is important to develop and use Al-based cheating detection systems in a way that is fair, transparent, and respectful of student privacy

4, Research questions and objectives

Online Exams: AI can analyze patterns of unusual behavior, such as suspiciously rapid answer submissions or consistent identical answers, to identify potential cheating during online exams

Plagiarism Detection: Al-powered algorithms can compare written documents to a vast database of existing content to identify instances of plagiarism and academic dishonesty

=> Yes, Al can effectively identify suspicious patterns of behavior that may indicate cheating in exams AI has rapidly become a powerful tool for identifying cheating behavior during online examinations, offering a range of advanced technologies to monitor and flag suspicious activities These Al-driven solutions leverage various techniques to detect cheating, including facial recognition, eye-tracking, keystroke analysis, and plagiarism detection

For instance, one of the key AI technologies used to detect cheating on online exams is facial recognition This involves capturing and analyzing students’ facial features through webcams to monitor their behavior during the exam Al-powered eye-tracking technology plays

a vital role in detecting cheating during online exams By continuously monitoring a test-taker’s eye movements, AI can identify suspicious behaviors such as looking at unauthorized materials

or glancing off-screen Keystroke analysis is another powerful tool used by AI to detect cheating

on online exams By analyzing typing patterns and keystrokes, AI can identify unusual behavior indicating potential cheating 1

However, it’s important to note that while AI has proven to be an effective tool for detecting cheating on online exams, it is essential to consider the ethical implications and privacy concerns associated with these technologies Educators and institutions must establish clear guidelines and policies regarding the use of AI for monitoring online exams, ensuring transparency and addressing any potential privacy issues

e How can AI be used to flag suspicious activity for further review during exams?

Here are several ways in which AI can be used for this purpose:

® Behavioral Analysis:

- Facial Recognition: AI can analyze facial expressions for signs of stress or unusual behavior, flagging instances where a test-taker appears distressed or engages in suspicious activities

- Eye Movement Tracking: Monitoring eye movements can help identify behaviors like excessive looking away from the screen, which may suggest cheating

@ Keystroke Analysis:

- Typing Patterns: AI algorithms can analyze typing patterns and detect anomalies, such as sudden changes in typing speed or irregular keystroke sequences

e Screen Monitoring:

- Content Analysis: AI can analyze the content on the screen to identify unauthorized

resources, multiple open tabs, or the use of external tools during the exam

- Activity Detection: Unusual mouse movements or patterns of interaction with the computer interface can be flagged for review

e Audio Analysis:

- Voice Stress Analysis: AI systems can analyze changes in voice patterns that may

indicate stress or nervousness, potentially signaling suspicious behavior

e Biometric Data Analysis:

- Physiological Monitoring: Monitoring physiological responses, such as heart rate and perspiration, can provide insights into a test-taker's stress levels

e Pattern Recognition:

- Machine Learning Models: Train AI models with data on normal exam behavior to detect patterns that deviate from the expected norm, flagging instances that are statistically significant

e System Logs and Metadata:

- Access Logs: Monitor system logs and metadata to detect unauthorized access or activities that are inconsistent with normal exam procedures

AI can analyze the student's voice If there are sudden changes in pitch, tone, or other voice characteristics that suggest stress or collaboration, the system can flag the session for review

e What are the most effective AI techniques for detecting plagiarism in student exam responses?

Al techniques for detecting plagiarism in student exam responses employ various methods and algorithms to compare and analyze text Here's an overview of the key techniques:

Algorithms like Levenshtein

distance, Jaccard similarity, or

cosine similarity compare the

similarity between text passages

Specialized tools compare entire

documents and highlight

similarities

NLP techniques analyze the

semantic structure of text

Train machine learning models

on datasets of known plagiarized

and non-plagiarized content

Algorithms compare source code

for similarities

Examine multiple exam

responses collectively to identify

patterns of similarities across a

group of students

Analyze references and citations

within exam responses

Examine metadata associated

with documents to identify

similarities in authorship, file

creation dates, or other relevant

information

AT-driven tutoring systems track

and analyze student progress

Online plagiarism detection

services compare exam responses

against vast databases of

academic and online content

How it Works These algorithms measure the similarity

of words or characters between two pieces of text Higher similarity scores indicate a greater likelihood of plagiarism

These tools often use advanced algorithms to identify matches, even when there are slight modifications, paraphrasing, or changes in the order of words

NLP can identify similar meaning in

different phrasing, making it effective in

detecting paraphrased or reworded content

These models learn patterns and features indicative of plagiarism, enabling them

to classify new exam responses based on the learned patterns

Token-based comparisons or abstract syntax tree (AST) analysis can identify similarities in coding structures, helping

to detect plagiarism in coding assignments

This approach helps detect collusion or cheating by analyzing similarities in responses across different documents Lack of proper citations, incorrect

citation formats, or the use of identical

citation patterns can be indicative of plagiarism

Consistency or anomalies in metadata

can be used to identify potential cases of plagiarism

Sudden improvements or changes in writing style may trigger suspicion, as

these systems can establish a baseline of

a student's typical performance These services leverage advanced algorithms and databases to identify similarities, providing a comprehensive approach to plagiarism detection

.Table 1; AI techniques for detecting plagiarism in student exam

e How can AI-based cheating detection systems be implemented in a fair, transparent, and privacy-preserving manner?

Implementing Al-based cheating detection systems in a fair, transparent, and privacy-preserving manner requires careful planning and adherence to ethical principles Here are some examples of how such systems can be implemented:

Transparency and Communication:

@ Example: Clearly communicate to students and educators about the use of Al-based cheating

detection systems in educational institutions Provide detailed information on what behaviors

are monitored and how the system operates

Informed Consent:

@ Example: Obtain informed consent from students before implementing Al-based cheating detection During the onboarding process, provide students with clear explanations about the purpose of the system and the data it will collect

Data Minimization:

@ Example: Collect only the necessary data for cheating detection For instance, focus on behavioral patterns during exams rather than collecting extensive personal information

Anonymization and Pseudonymization:

@ Example: Implement techniques like anonymization or pseudonymization to protect the

identities of students Ensure that the AI system works with de-identified data

Encryption:

@ &xample: Encrypt all data, especially sensitive information, during transmission and storage This prevents unauthorized access and protects the privacy of individuals

Access Controls:

@ £xample: Implement strict access controls to limit access to data collected by the cheating detection system Only authorized personnel with a legitimate need should have access Purpose Limitation:

@ &xample: Clearly define the purpose of collecting data for cheating detection and ensure that the data is not repurposed for unrelated activities, maintaining a focused and limited scope

Ethical Use of Data:

@ &xample: Establish and follow ethical guidelines for data usage, avoiding discriminatory practices and ensuring that the data is used solely for the intended purpose of maintaining academic integrity

Continuous Improvement:

@ &xample: Regularly update and improve the AI models based on feedback and evolving standards This ensures that the system remains accurate, fair, and up-to-date

Regular Audits:

@ &xample: Conduct regular audits of the cheating detection system to assess its performance, identify potential biases, and ensure compliance with privacy and ethical standards

Legal Compliance:

@ &xample: Ensure compliance with relevant privacy laws and regulations Understand and adhere to data protection and student privacy laws in the jurisdiction where the institution operates

e What are the potential ethical implications of using AI to detect and prevent

cheating in exams?

The use of AI to detect and prevent cheating in exams has several potential ethical implications:

1 Privacy Concerns: AI systems often require access to personal data, such as video feeds, audio recordings, and keystroke patterns, to detect cheating This raises concerns about

student privacy and data security

10