1.1 Development and Evolution of AI Tools: Machine Learning, Deep Learning, and Natural Language Processing Techniques

The field of artificial intelligence (AI) has seen significant advancements in recent years, particularly in its application to plagiarism detection. The evolution of AI tools—from rudimentary algorithms to sophisticated machine learning, deep learning, and natural language processing (NLP) techniques—has transformed the landscape of academic integrity. Early efforts in this domain primarily relied on basic string matching and naïve algorithms, which were often limited by their inability to detect semantic similarities and nuanced language use (Manzoor et al., 2023). The progression towards more complex models has subsequently led to more accurate and reliable plagiarism detection systems.

Machine learning has greatly enhanced the capabilities of plagiarism detection tools. Traditional algorithms could identify exact text matches, but they struggled with paraphrased or contextually altered content. Machine learning approaches address these gaps by learning from vast datasets to recognize patterns and anomalies (Moravvej et al., 2023). For example, supervised learning models can be trained on labeled datasets to distinguish between plagiarized and original work, incorporating a variety of features such as word frequency, sentence structure, and syntactic patterns. Unsupervised learning models can further refine these capabilities by identifying clusters of similar documents without requiring labeled data.

Deep learning has pushed the boundaries of what is possible in plagiarism detection. Techniques like long short-term memory (LSTM) networks and convolutional neural networks (CNNs) are adept at processing and understanding complex linguistic features. LSTM networks, combined with attention mechanisms, have proven particularly effective in context-aware text analysis (Moravvej et al., 2023). These models can handle long sequences of text and maintain context over multiple sentences, making them highly suitable for identifying subtle instances of plagiarism that simpler models might overlook.

The advent of large language models (LLMs), such as BERT (Bidirectional Encoder Representations from Transformers), has added another layer of sophistication to plagiarism detection (Quidwai et al., 2023). BERT-based models leverage bi-directional context, allowing them to understand words in relation to the entire sentence. This method contrasts with traditional unidirectional models, which only consider preceding or following words. The use of BERT enhances the detection of semantic similarities and contextual nuances, significantly improving the accuracy of plagiarism detection systems.

Natural Language Processing (NLP) plays a crucial role in this evolution. NLP techniques involve the application of algorithms to understand, interpret, and generate human language. In the context of plagiarism detection, NLP can be used to parse and analyze text, identifying paraphrasing, synonyms, and other linguistic variations that might indicate plagiarized content. For instance, advanced NLP models can decompose complex sentences into manageable parts, making it easier to detect subtle forms of plagiarism like synonym swapping and sentence reordering (Manzoor et al., 2023). The integration of NLP with machine learning and deep learning creates a powerful triad that enhances the robustness of plagiarism detection tools.

Much of the current research focuses on improving the robustness and accuracy of these models. Manzoor et al. (2023) conducted a Systematic Literature Review (SLR) that analyzes various facets of intrinsic plagiarism detection, including feature extraction techniques and detection methods. This exhaustive review highlights the challenges posed by low-resource languages and the need for more inclusive and adaptable models. Similarly, Moravvej et al. (2023) proposed a novel approach combining BERT-based word embeddings with attention-based LSTMs and an improved differential evolution algorithm, demonstrating that their model outperforms conventional methods in several benchmark datasets.

The future of AI in plagiarism detection looks promising, with ongoing research aimed at reducing false positives and enhancing interpretability. Quidwai et al. (2023) underscore the importance of transparent AI systems, advocating for NLP techniques that offer quantifiable metrics at both the sentence and document levels. Their proposed method, which achieves up to 94% accuracy, provides a robust solution that is adaptable to advancements in LLM technology.

In conclusion, the evolution of AI tools for plagiarism detection—from basic algorithms to sophisticated machine learning, deep learning, and NLP techniques—has significantly improved the accuracy and reliability of these systems. Continued advancements in these areas, coupled with comprehensive research and development, promise to further enhance the detection of plagiarism, thereby upholding academic integrity and the quality of scholarly publications.

1.2 Key Mechanisms AI Uses to Identify Plagiarism: Pattern Recognition, Semantic Analysis, and Synonym Swapping Detection

AI has revolutionized the field of plagiarism detection through various sophisticated mechanisms such as pattern recognition, semantic analysis, and synonym swapping detection. These advanced technologies have aimed to overcome the limitations of traditional plagiarism detection tools by providing more nuanced and comprehensive analysis capabilities.

Pattern recognition, one of the key methodologies in AI-driven plagiarism detection, involves identifying recurring sequences of words or similar structures between texts. This technique has evolved significantly with the advent of machine learning and deep learning models, which facilitate more accurate and scalable recognition of patterns across large datasets. Altheneyan and Menai (2019) highlight how deep learning techniques, particularly support vector machines (SVMs), have achieved remarkable performance in identifying word overlaps and structural representations. Their research indicates that deep learning techniques are highly effective in both paraphrase identification and plagiarism detection, showing promise as the leading approach in this domain. By training on extensive corpora, AI tools can learn to recognize even subtle similarities that might be missed by human reviewers or simpler automated systems.

Semantic analysis builds on the capabilities of pattern recognition by delving deeper into the meaning of the text. This involves not just comparing the surface structure of sentences, but also understanding the underlying semantics. Lazemi and Ebrahimpour-Komleh (2020) provide an excellent example of this with their ParsiPayesh system, which employs deep learning models for semantic role labeling and structural analysis. Through these methods, ParsiPayesh can detect re-writing plagiarisms by analyzing the dependency trees of sentences and their semantic roles. This approach allows the system to identify significant modifications that maintain the original meaning of the text, thus uncovering more sophisticated forms of plagiarism. The success of ParsiPayesh underscores the importance of incorporating both structural and semantic analyses to enhance the efficacy of plagiarism detection systems, particularly in non-English languages where resources are scarce.

Another critical mechanism utilized by AI in plagiarism detection is synonym swapping detection. This technique involves identifying instances where plagiarizers attempt to disguise copied content by replacing words with their synonyms. Traditional tools often fail to catch such rephrasing. As mentioned by Nwohiri et al. (2021), AI-driven detectors leverage natural language processing (NLP) and forensic linguistics to detect these subtle changes. By utilizing advanced NLP techniques, AI tools can discern when the semantic content remains unchanged despite lexical alterations. Nwohiri et al. (2021) emphasize the need for AI systems that handle unintentional plagiarism due to inaccurate citation and intelligently navigate through the complexities of modern academic writing. Their proposed AI-driven detector is designed to train users on proper citation formats, thereby reducing instances of inadvertent plagiarism while enhancing detection accuracy.

In summary, AI’s application of pattern recognition, semantic analysis, and synonym swapping detection has significantly improved the landscape of plagiarism detection. The integration of deep learning models, as demonstrated in the works of Altheneyan and Menai (2019) and Lazemi and Ebrahimpour-Komleh (2020), has proven to be instrumental in achieving higher accuracy and efficiency. Moreover, the emphasis on forensic linguistics and NLP by Nwohiri et al. (2021) highlights AI’s capability to handle increasingly sophisticated plagiarism techniques and promote academic integrity through better citation practices. As AI technologies continue to advance, their role in detecting plagiarism and ensuring the quality of academic publications is likely to become even more pivotal.

 2.1 Accuracy and Efficiency in Text Analysis: Benchmark Studies, Real-time Processing, and Scalability

The landscape of plagiarism detection is undergoing a dynamic shift with the deployment of Artificial Intelligence (AI) technologies. Traditional plagiarism detection tools, while robust in their own right, predominantly rely on pattern matching and databases of known works to identify potential instances of academic dishonesty. In contrasting these with AI-based mechanisms, notable differences emerge in terms of accuracy, real-time processing capabilities, and scalability.

A comparative benchmark reveals that traditional tools often falter in instances requiring complex text interpretation and semantic understanding. According to Liu and Do (2024), AI models like Originlens capitalize on deep learning and natural language processing (NLP) to discern with a high degree of accuracy—approximately 96%—whether a text is AI-generated or human-authored. The integration of machine learning algorithms and augmented reality in Originlens highlights its ability to minimize human intervention while enhancing detection efficiency. This is a substantial leap compared to traditional tools which might rely excessively on exact text matching, thus limiting their effectiveness in cases of sophisticated paraphrasing or synonym swapping.

However, the pursuit of accuracy is not without challenges. Benedek and Nagy’s (2023) research on AI-based fraud detection in the automobile insurance sector underscores a counterintuitive observation. Their findings indicate that AI-driven systems, despite their theoretical advantages, may not always be more cost-effective than conventional statistical-econometric methodologies. This insight questionably translates to the field of plagiarism detection, where the cost of implementation and operation can outweigh the benefits if the AI systems are not meticulously calibrated for the purpose.

In terms of real-time processing, AI tools have introduced a transformative approach. The processing capability of AI models outpaces traditional methods, which typically involve significant time lags due to extensive database queries and pattern checks. As highlighted by Liu and Do (2024), the integration of deep learning algorithms in real-time text analysis allows for immediate feedback, thereby accelerating the identification of potential plagiarism. This is particularly beneficial in fast-paced academic environments where the timely validation of text authenticity can prevent the submission of plagiarized content and maintain academic integrity.

Scalability, another crucial factor, sees AI-based tools taking the lead over traditional plagiarism detection systems. Traditional methods often struggle with large volumes of data, leading to performance bottlenecks. In contrast, as Asimopoulos et al. (2024) illustrate, AI models such as Transformers incorporate self-attention mechanisms and are highly scalable, adept at handling vast datasets without compromising on performance. Their study emphasizes the enhanced capabilities of AI in anonymizing text, which, though focused on privacy, also showcases the scalability aspect pertinent to plagiarism detection. Large-scale academic databases, extensive literary works, and voluminous student submissions are within the processing ambit of AI tools, ensuring comprehensive and swift plagiarism verification.

The synergy of AI models, especially when combined with other robust techniques, has resulted in significantly improved outcomes in text analysis. For instance, frameworks that integrate Conditional Random Fields (CRF), Long Short-Term Memory (LSTM), and Embeddings from Language Models (ELMo) exhibit superior performance in capturing long-term dependencies and contextual nuances of text compared to traditional methods (Asimopoulos et al., 2024). These sophisticated models collectively push the boundaries of plagiarism detection, providing a more nuanced and detailed scrutiny of textual evidence.

In conclusion, the advent of AI in plagiarism detection has revolutionized the accuracy, efficiency, and scalability of text analysis. While traditional tools have laid a solid groundwork, AI offers a futuristic approach by leveraging advanced algorithms and real-time processing capabilities. Nonetheless, it’s crucial to consider the financial and practical aspects, as exemplified by Benedek and Nagy (2023), to ensure that the integration of AI is both cost-effective and operationally viable. As research continues to evolve, striking a balance between traditional and AI-driven methods may offer the most comprehensive solution for maintaining academic integrity.

 2.2 Limitations and Challenges in Current Tools: False Positives, Ambiguity in Text Interpretation, and Resource Consumption

The application of Artificial Intelligence (AI) in plagiarism detection has certainly advanced over the years, yet it is not without its pitfalls. One significant limitation of current AI tools in plagiarism detection is the prevalence of false positives—situations where a text is erroneously flagged as plagiarized. According to Krishna et al. (2023), several detectors, such as watermarking and outlier detection, struggle when texts are subtly altered. For instance, their study showed that a paraphrase generation model, DIPPER, successfully reduced the detection accuracy of the DetectGPT tool from 70.3% to a mere 4.6% without altering the text’s fundamental semantics (Krishna et al., 2023). The false positive rate here highlights the vulnerability of current detection methodologies to intentional manipulation and paraphrasing techniques.

In addition to false positives, false negatives also pose a considerable challenge. Ni, Wu, and Su (2023) underscore this predicament in their examination of STARR-seq data, demonstrating that up to 87% of detected sequences might be located in repressive chromatin and may not function in the tested cells. This discrepancy not only exacerbates the problem of unrecognized plagiarism but also undermines the credibility of AI tools in academic integrity. While improvements have been proposed to mitigate systematic errors, the inherent limitations of the current models make it difficult to fully eliminate false negatives in text analysis.

Ambiguity in text interpretation is another significant hurdle. AI models often struggle to understand context-specific semantics, especially when dealing with technical jargon or nuanced language. Krishna et al. (2023) revealed that paraphrasing mechanisms can easily evade detection by subtly modifying the lexical diversity and content ordering of a text, thereby retaining its original meaning but presenting it in a different lexical form. Detecting such sophisticated alterations would require more advanced semantic analysis capabilities than many existing tools possess.

Resource consumption is yet another major challenge. Advanced AI models, particularly those utilizing deep learning and extensive databases, require considerable computational resources. Cappelen, Cappelen, and Tungodden’s (2023) study offers an economic lens to this issue, suggesting that the trade-off between developing robust AI tools and managing resource constraints is a significant decision-making factor. The balance between false positives and false negatives must be managed carefully, as overlooking either could lead to significant implications on both academic and economic fronts.

Furthermore, the necessity of maintaining vast databases to optimize detection accuracy adds to the resource burden. Krishna et al. (2023) highlighted a method involving a large database of 15 million sequences to detect paraphrased text successfully. While this enhanced the robustness of the detection mechanism, it also spelled out the need for significant storage and processing power. This scenario poses an inherent scalability issue, limiting the accessibility of such high-precision tools to well-funded institutions.

Ultimately, addressing these limitations requires a multifaceted approach. Integrating more advanced semantic engines capable of subtle meaning differentiation, establishing better algorithms for real-time processing, and investing in computational resources could form part of the solution. However, the balance between technological advancements and practical feasibility remains a compelling question. As AI continues to evolve, resolving these challenges will be crucial for enhancing its reliability and accuracy in academic settings.

In conclusion, while AI tools have revolutionized plagiarism detection, they are not infallible. False positives, false negatives, interpretive ambiguity, and high resource consumption underline the nuanced and complex landscape of AI-aided plagiarism detection. As the technology matures, addressing these limitations will not only enhance its efficacy but also fortify the academic integrity it seeks to protect.

 3.1 Integrating AI for Improved Reliability: Automated Peer Review, Cross-Referencing, and Content Validation

Artificial Intelligence (AI) has made tangible strides in enhancing the reliability of academic publications. One of the most promising developments is the integration of AI in the peer review process, which traditionally has been labor-intensive and fraught with inconsistencies. AI-powered tools offer a more efficient, unbiased, and thorough examination of manuscripts, thereby significantly improving the reliability and quality of academic outputs.

A notable application of AI in academic settings is automating the peer review process. Oliveira, Rios, and Jiang’s (2023) study reveals the effectiveness of generative AI (genAI) in streamlining the code review process among computer science students. The study found that AI could provide timely, consistent, and objective feedback, significantly enhancing student engagement. The use of genAI led to the identification of a larger number of issues within shorter timelines, resulting in more efficient problem-solving. This suggests that AI-powered peer review could be effectively applied beyond software projects, potentially addressing broader issues of reliability, bias, and fairness in academic publishing.

Moreover, the adoption of AI tools can significantly alleviate the workload of human reviewers. Checco et al. (2021) highlight the increasing strain on the peer review workflow due to the surge in academic submissions. They argue that AI-assisted initial screenings can save millions of working hours by flagging potential issues like plagiarism and format discrepancies. More advanced AI tools can even assess the quality of studies and summarize content, reducing the reviewer workload. Their research demonstrates that AI can predict review scores based on textual content alone, revealing correlations between review decisions and quality measures. This predictive capability can help uncover biases in the review process and ensure a more objective assessment of academic work.

Importantly, AI does not merely promise efficiency but also aims to enhance the precision of the review process. AI algorithms can cross-reference large datasets to identify similarities and discrepancies that human reviewers might overlook. For example, automated systems can instantly compare manuscripts against vast databases of existing literature, identifying subtle forms of plagiarism and ensuring that the citations and references are consistent and accurate. This cross-referencing capability is crucial in maintaining the integrity of academic publications, ensuring that all sources are accurately represented, and preventing intellectual property violations.

Yet, the integration of AI in academic publishing isn’t without challenges. Munz, Hennick, and Stewart (2023) discuss the need for anticipatory thinking and model risk audits to maximize the reliability of AI applications. They argue that while AI holds the potential for substantial improvements, its deployment must be carefully managed to mitigate risks such as model robustness failures, data security issues, and biases. By adopting a flexible model risk audit framework, organizations can better prepare for emerging regulations and ensure responsible AI deployments. This approach involves characterizing risks at the model level and applying rigorous monitoring to ensure that AI tools function as intended without introducing new ethical concerns.

Ethical considerations are paramount in deploying AI for academic purposes. Checco et al., (2021) raise concerns about algorithmic bias and the ethical implications of AI-assisted peer review systems. AI algorithms might inadvertently replicate existing biases present in training data, potentially leading to unfair evaluations. Therefore, developing transparent, explainable AI models is essential to ensure trust and accountability in academic publishing. This entails rigorous testing and validation procedures to uncover and rectify any biases, ensuring that AI tools reliably contribute to the academic review process.

In conclusion, AI’s integration into academic publishing holds significant promise for improving the reliability and quality of scholarly work. By automating the peer review process, facilitating thorough cross-referencing, and ensuring precise content validation, AI can address many of the inefficiencies and inconsistencies that currently plague academic publishing. However, realizing these benefits requires vigilant monitoring, ethical considerations, and transparent practices to mitigate potential risks and ensure the responsible use of AI technologies.

 3.2 Future Prospects for AI in Academic Publishing: Enhanced Algorithms, Ethical Considerations, and Policy Implementation

Artificial Intelligence (AI) in academic publishing is entering a transformative phase, characterized by enhanced algorithms, ethical considerations, and policy implementation. As AI technologies continue to evolve, their capacity to assist in the comprehensive review and publication of academic materials becomes increasingly profound. The future prospects for AI in this domain promise to reshape not only the operational mechanics but also the ethical and regulatory frameworks within which academic publishing operates.

Enhanced algorithms are at the forefront of AI advancements in academic publishing. These algorithms, powered by state-of-the-art machine learning and deep learning techniques, are designed to process vast datasets with remarkable accuracy. For instance, algorithms can now effortlessly cross-reference citations, detect inconsistencies, and identify potential plagiarism with a level of precision that traditional tools cannot match. Lund, Lamba, and Oh (2024) highlight that generative AI technologies like ChatGPT have already achieved content generation capabilities that rival or exceed those of human writers. This suggests that future AI algorithms could further elevate the quality and consistency of academic publications, ensuring they meet the highest standards of scholarly rigor.

However, the integration of advanced AI in academic publishing also brings forth significant ethical considerations. The potential for AI to be misused for scholarly misconduct, such as fabricating data or generating misleading research, necessitates stringent oversight and ethical guidelines. According to Lund, Lamba, and Oh (2024), a collaborative approach involving publishers, editors, reviewers, and authors is essential to leverage AI productively and ethically. This includes establishing clear protocols for AI’s role in the writing, editing, and peer-review processes to prevent any misuse that could compromise academic integrity. Transparency in AI usage—such as disclosing when and how AI tools are used in the research and publication process—is critical for maintaining trust within the academic community.

Moreover, policy implementation is a vital aspect of responsibly integrating AI into academic publishing. The Heredia Declaration, as discussed by Penabad-Camacho et al. (2024), proposes principles for the responsible use of AI in scientific publishing. These principles advocate for transparency, traceability, and respect for intellectual property, emphasizing that AI should be viewed as a supportive tool rather than a replacement for human judgment. The Declaration also underscores the importance of reporting which AI models were used, what data was consulted, and the period of consultation, ensuring that AI’s contributions are fully accountable and verifiable.

Furthermore, AI’s role in academic publishing must be considered within the broader context of educational and academic policies. Tanveer, Hassan, and Bhaumik (2020) discuss how AI can be integrated into sustainable development education, showcasing AI’s potential to improve educational quality and accessibility, particularly in developing nations. This integration requires thoughtful implementation strategies, as well as policies that accommodate AI’s dual role in enhancing educational outcomes and maintaining ethical standards. Policymakers must consider the implications of AI’s growing presence in academia, ensuring that the technology supports rather than undermines educational equity and integrity.

Looking forward, the prospects for AI in academic publishing are both promising and complex. The continuous enhancement of AI algorithms promises to boost the efficiency and reliability of academic reviews and publications. At the same time, addressing ethical concerns and implementing robust policies will be crucial to prevent potential misuse and ensure that AI’s integration upholds the core values of academic research and publication. Through collaborative efforts and transparent practices, AI can significantly contribute to the quality and integrity of academic literature, ultimately fostering a more rigorous and trustworthy scholarly environment.

References:

Altheneyan, A., & Menai, M. E. B.. (2019). Evaluation of State-of-the-Art Paraphrase Identification and Its Application to Automatic Plagiarism Detection. International Journal of Pattern Recognition and Artificial Intelligence, 34(4), 2053004. https://doi.org/10.1142/s0218001420530043

Asimopoulos, D., Siniosoglou, I., Argyriou, V., Goudos, S. K., Psannis, K. E., Karditsioti, N., … Sarigiannidis, P. G.. (2024). Evaluating the Efficacy of AI Techniques in Textual Anonymization: A Comparative Study. 2024 7th International Balkan Conference on Communications and Networking (balkancom), 242–246. https://doi.org/10.1109/BalkanCom61808.2024.10557182

Benedek, B., & Nagy, B.. (2023). Traditional versus AI-Based Fraud Detection: Cost Efficiency in the Field of Automobile Insurance. Financial and Economic Review. https://doi.org/10.33893/fer.22.2.77

Cappelen, A. W., Cappelen, C., & Tungodden, B.. (2023). Second-Best Fairness: The Trade-Off between False Positives and False Negatives. American Economic Review, 113(9), 2458–2485. https://doi.org/10.1257/aer.20211015

Checco, A., Bracciale, L., Loreti, P., Pinfield, S., & Bianchi, G.. (2021). AI-assisted peer review. Humanities and Social Sciences Communications, 8(1). https://doi.org/10.1057/s41599-020-00703-8

Krishna, K., Song, Y., Karpinska, M., Wieting, J., & Iyyer, M.. (2023). Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense. Arxiv, abs/2303.13408. https://doi.org/10.48550/arXiv.2303.13408

Lazemi, S., & Ebrahimpour-Komleh, H.. (2020). ParsiPayesh: Persian Plagiarism Detection based on Semantic and Structural Analysis. 2020 10th International Conference on Computer and Knowledge Engineering (ICCKE), 525–533. https://doi.org/10.1109/ICCKE50421.2020.9303672

Liu, H. M., & Do, J.. (2024). Originlens: A Real-Time AI-Generated Text and Plagiarism Detection using Deep Learning and Augmented Reality. Artificial Intelligence and Big Data. https://doi.org/10.5121/csit.2024.140405

Lund, B., Lamba, M., & Oh, S. H.. (2024). The Impact of AI on Academic Research and Publishing. Arxiv, abs/2406.06009. https://doi.org/10.48550/arXiv.2406.06009

Manzoor, M. F., Farooq, M., Haseeb, M., Farooq, U., Khalid, S., & Abid, A.. (2023). Exploring the Landscape of Intrinsic Plagiarism Detection: Benchmarks, Techniques, Evolution, and Challenges. IEEE Access, 11, 140519–140545. https://doi.org/10.1109/ACCESS.2023.3338855

Moravvej, S. V., Mousavirad, S. J., Oliva, D., & Mohammadi, F.. (2023). A Novel Plagiarism Detection Approach Combining BERT-based Word Embedding, Attention-based LSTMs and an Improved Differential Evolution Algorithm. Arxiv, abs/2305.02374. https://doi.org/10.48550/arXiv.2305.02374

Munz, P., Hennick, M., & Stewart, J.. (2023). Maximizing AI reliability through anticipatory thinking and model risk audits. AI Magazine, 44(2), 173–184. https://doi.org/10.1002/aaai.12099

Ni, P., Wu, S., & Su, Z.. (2023). Underlying causes for prevalent false positives and false negatives in STARR-seq data. NAR Genomics and Bioinformatics, 5(3). https://doi.org/10.1093/nargab/lqad085

Nwohiri, A. M., Joda, O., & Ajayi, O. O.. (2021). AI-POWERED PLAGIARISM DETECTION: LEVERAGING FORENSIC LINGUISTICS AND NATURAL LANGUAGE PROCESSING. FUDMA JOURNAL OF SCIENCES. https://doi.org/10.33003/fjs-2021-0503-700

Oliveira, E. A., Rios, S., & Jiang, Z.. (2023). AI-powered peer review process. ASCILITE Publications. https://doi.org/10.14742/apubs.2023.482

Penabad-Camacho, L., Penabad-Camacho, M. A., Mora-Campos, A., Cerdas-Vega, G., Morales-López, Y., Ulate-Segura, M., … Castro-Solano, M. M.. (2024). Heredia Declaration: Principles on the use of Artificial Intelligence in scientific publishing. Revista Electrónica Educare. https://doi.org/10.15359/ree.28-s.19967

Quidwai, M., Li, C. X., & Dube, P.. (2023). Beyond Black Box AI generated Plagiarism Detection: From Sentence to Document Level. Workshop on Innovative Use of NLP for Building Educational Applications, 727–735. https://doi.org/10.48550/arXiv.2306.08122

Tanveer, M., Hassan, S., & Bhaumik, A.. (2020). Academic Policy Regarding Sustainability and Artificial Intelligence (AI). Sustainability, 12, 9435. https://doi.org/10.3390/su12229435

Photo by Amanda Jones on Unsplash