Generative Artificial Intelligence (AI) in Higher Education: A Comprehensive Review of Challenges, Opportunities, and Implications

May combine two claims.

Generative AI and LLMs in higher education present challenges related to educational assessment, academic integrity, privacy, misinformation, EDI, copyright, and inherent biases.

“In the context of higher education, the paper explores the challenges and opportunities posed by LLMs.These include issues related to educational assessment, potential threats to academic integrity, privacy concerns, the propagation of misinformation, Equity, Diversity, and Inclusion (EDI) aspects, copyright concerns and inherent biases within the models.”

• AssumptionRests on an assumption: The academic literature reviewed provides a comprehensive and representative account of the current state of generative AI in higher education.

Classified as probably not a substantial claim.

Strategies are available to address the challenges posed by LLMs and facilitate their successful adoption in educational settings.

“While these challenges are multifaceted and significant, the paper emphasises the availability of strategies to address them effectively and facilitate the successful adoption of LLMs in educational settings.”

• AssumptionRests on an assumption: The academic literature reviewed provides a comprehensive and representative account of the current state of generative AI in higher education.

Probably not a substantial claim: may not pass S3 (load-bearing).

Technology companies are in talks with leading media organisations about using their news content for AI training.

“It has been reported that technology companies are in talks with leading media organisations about using their news content for AI training (Criddle et al., 2023).”

• cites Criddle et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

ChatGPT reached an estimated 100 million monthly active users in just two months after its launch.

“The app has become the fastest-growing online application in history, reaching an estimated 100 million monthly active users in just two months after its launch (Reuters, 2023).”

• cites Reuters, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

ChatGPT can be fine-tuned for specific tasks or domains, increasing its versatility across various industries and applications.

“Furthermore, ChatGPT can be fine-tuned for specific tasks or domains, increasing its versatility across various industries and applications (Ray, 2023).”

• cites Ray, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The size of the data set used for OpenAI's most recent model, GPT-4, has not been disclosed.

“The size of the data set used for OpenAI's most recent model, GPT-4, has not been disclosed (Heikkilä, 2023).”

• cites Heikkilä, 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Bloomberg is developing a new generative AI model, trained on an extensive variety of financial data owed by the company.

“Finally, Bloomberg, the information provider for financial services that is also used by over 300 universities worldwide (Bloomberg, 2023a), is developing a new generative AI model, trained on an extensive variety of financial data owed by the company (Bloomberg, 2023b).”

• cites Bloomberg, 2023bNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Copyrights differ between jurisdictions.

“Further, it is important that copyrights differ between jurisdictions.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In AI development, the prevailing belief is that the larger the training data set, the better the results.

“In the realm of AI development, the prevailing belief is that the larger the training data set, the better the results.”

May combine two claims.

Traditional plagiarism-detection tools are rendered ineffective because AI is becoming increasingly proficient at imitating human-like text and AI-generated text is deemed original and thus undetectable.

“Traditional plagiarism-detection tools are rendered ineffective in this context, as AI is becoming increasingly proficient at imitating human-like text and AI-generated text is deemed original and thus undetectable (Dwivedi et al., 2023.)”

• cites Dwivedi et al., 2023Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

Current generative AI platforms have limitations such as inability to interpret visual media or understand up-to-date information beyond their training data.

“Gonsalves (2023) suggests that educators using multiple choice test for assessment may consider outsmarting the system by taking advantage of the current limitations of generative AI platforms, such as their inability to interpret visual media or understand up-to-date information beyond its training data.”

• cites Gonsalves (2023)Not checked yet

  Not checked in the free run.

May combine two claims.

Given TikTok's 3.7 million active UK users, with 26% aged 18-24, the potential for covert cheating is significant.

“Given TikTok's 3.7 million active UK users, with 26% aged 18-24 (Bendix, 2023), the potential for covert cheating is significant.”

• cites Bendix, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT is capable of generating human-like responses using natural language, essentially leading to an expert system that is available on demand.

“ChatGPT is capable of generating human-like responses using natural language, essentially leading to an expert system that is available on demand.”

• cites Haleem et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

OpenAI's GPT-2 model was trained using a data set that comprised 40 gigabytes of text.

“OpenAI's GPT-2 model was trained using a data set that comprised 40 gigabytes of text.”

• cites Heikkilä, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT can understand the context of a conversation and generate relevant responses, making it more effective at mimicking human-like interactions.

“It can understand the context of a conversation and generate relevant responses, making it more effective at mimicking human-like interactions.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT has brought advancements to generative AI including improved contextual understanding, language generation, task adaptability, and multilingual proficiency.

“While chatbots are not a new invention, ChatGPT has brought several advancements to generative AI, including improved contextual understanding, language generation, task adaptability, and multilingual proficiency.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Two authors have filed a lawsuit against OpenAI, alleging that their copyrighted books were used to train the AI model, ChatGPT, without their permission.

“For instance, it has been reported that two authors have filed a lawsuit against OpenAI, alleging that their copyrighted books were used to train the AI model, ChatGPT, without their permission.”

• cites Creamer, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

GPT-3, the model upon which ChatGPT is built, utilised a significantly larger data set of 570 GB.

“GPT-3, the model upon which ChatGPT is built, utilised a significantly larger data set of 570 GB.”

• cites Heikkilä, 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Adobe's 'Firefly', a generative AI image tool, guarantees commercial usage safety, offering IP indemnification as it is trained on company-owned images.

“For instance, Adobe's 'Firefly', a generative AI image tool, guarantees commercial usage safety, offering IP indemnification as it is trained on company-owned images (Gold, 2023).”

• cites Gold, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT is trained on extensive data and is competent across disciplines.

“Trained on extensive data, it is competent across disciplines.”

Classified as probably not a substantial claim.

There are potential opportunities to transform higher education through LLMs.

“Furthermore, the paper recognises the potential opportunities to transform higher education.”

• AssumptionRests on an assumption: The academic literature reviewed provides a comprehensive and representative account of the current state of generative AI in higher education.

Probably not a substantial claim: may not pass S3 (load-bearing).

ChatGPT's advanced language generation capabilities allow it to produce coherent, contextually accurate, and grammatically correct text.

“Its advanced language generation capabilities allow it to produce coherent, contextually accurate, and grammatically correct text.”

Classified as probably not a substantial claim. May combine two claims.

Designing tests to outsmart generative AI may be more time consuming, difficult in certain subjects, and attempts to out-design chatbots might prove fruitless in the long term as technology improves.

“However, designing such tests may be more time consuming, difficult to do in certain subjects, and as technology improves, attempts to out-design chatbots might prove fruitless in the long term (Mills, 2023).”

• cites Mills, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Google, along with its parent company Alphabet and AI subsidiary DeepMind, has been accused in a lawsuit of scraping user data without consent and violating copyright laws to train its AI products.

“In a separate case, Google, along with its parent company Alphabet and AI subsidiary DeepMind, has been accused in a lawsuit of scraping user data without consent and violating copyright laws to train its AI products.”

• cites Dixit, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

ChatGPT uses its training data to produce an answer.

“This model uses its training data to produce an answer (Haleem et al., 2023).”

• cites Haleem et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Both Tlili et al. and Haensch et al. emphasise very little discourse related to negative aspects of using generative AI in education, meaning systems such as ChatGPT may be used without much reflection and evaluation.

“Moreover, both Tlili et al. (2023) and Haensch et al. (2023) emphasise very little discourse related to negative aspects of using generative AI in education, meaning that systems such as ChatGPT may be used without much reflection and evaluation.”

• cites Tlili et al. (2023)Not checked yet

  Not checked in the free run.

• cites Haensch et al. (2023)Not checked yet

  Not checked in the free run.

May combine two claims.

Using generative AI leads to substantial risks of accidental plagiarism and copyright infringements.

“Therefore, using generative AI leads to substantial risks of accidental plagiarism and copyright infringements (Gimpel et al., 2023).”

• cites Gimpel et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

While ChatGPT 3.5 does not have internet access, OpenAI allows its newer model GPT 4.0 internet access through the Bing search engine or third-party plug-ins.

“For instance, while ChatGPT 3.5 does not have internet access, OpenAI allows its newer model GPT 4.0 internet access through the Bing search engine or third-party plug-ins (Wiggers, 2023).”

• cites Wiggers, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Google has updated privacy terms which now clearly state their right to use publicly available data for AI model training.

“Google has updated privacy terms which now clearly state their right to use publicly available data for AI model training.”

May combine two claims.

ChatGPT's content generation capabilities excel in aspects people previously used to filter information, potentially creating a false sense of reliability.

“However, ChatGPT's content generation capabilities excel in these aspects, potentially creating a false sense of reliability.”

Classified as probably not a substantial claim.

Embedding AI literacy in graduate skills can enhance their employability in a rapidly evolving job market.

“Thus, embedding AI literacy in graduate skills can enhance their employability in a rapidly evolving job market.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Developing use cases for generative AI is a significant challenge for educational providers.

“Developing use cases for generative AI is a significant challenge for educational providers.”

May combine two claims.

Non-native English speakers can use ChatGPT for grammar feedback and as a semi-translator for complex terms, aiding comprehension.

“Non-native English speakers can use ChatGPT for grammar feedback and as a semi-translator for complex terms, aiding comprehension (Sullivan et al., 2023).”

• cites Sullivan et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Generative AI may support research, writing, and data analysis, opening new possibilities for educators and students and potentially unleashing their digital creativity.

“Overall, generative AI may support research, writing, and data analysis which effectively opens new possibilities for educators and students, potentially unleashing their digital creativity.”

May combine two claims.

Tlili et al. examined OpenAI's policies and found that while conversations with ChatGPT are stored and used to improve its performance, the specifics of storage and use are not entirely clear.

“Tlili et al. (2023) examined OpenAI's policies and found that while conversations with ChatGPT are stored and used to improve its performance, the specifics of storage and use are not entirely clear.”

• cites Tlili et al. (2023)Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT has been widely criticised for producing biased and discriminatory content, which has been caused by using training data that reflects the biases of society.

“ChatGPT has been widely criticised for producing biased and discriminatory content, which has been caused by using training data that reflects the biases of society (Weinberger, 2019;Dwivedi et al., 2023).”

• cites Weinberger, 2019Not checked yet

  Not checked in the free run.

• cites Dwivedi et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Generative AI may assist in the creation of inventive assessments that could promote creativity and critical thinking skills, contributing to comprehensive and meaningful evaluation of learning outcomes.

“This technology may assist in creation of inventive assessments which could promote creativity and critical thinking skills, contributing to comprehensive and meaningful evaluation of learning outcomes (Rasul et al., 2023).”

• cites Rasul et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Generative AI offers immediate feedback and advice, assisting students in rectifying mistakes and enhancing their study techniques.

“In addition, it also offers immediate feedback and advice, assisting students in rectifying mistakes and enhancing their study techniques (Javaid at al., 2023;Yu and Guo, 2023).”

• cites Javaid at al., 2023Not checked yet

  Not checked in the free run.

• cites Yu and Guo, 2023Not checked yet

  Not checked in the free run.

May combine two claims.

Given the vast data processed by ChatGPT's machine learning algorithms, it is vulnerable to cyberattacks, risking unauthorised access or misuse of sensitive information.

“Given the vast data processed by ChatGPT's machine learning algorithms, it is vulnerable to cyberattacks, risking unauthorised access or misuse of sensitive information (Dwivedi et al., 2023).”

• cites Dwivedi et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

AI can support adaptive writing and highlight essential information in various formats.

“AI can also support adaptive writing and highlight essential information in various formats (Kasneci et al., 2023).”

• cites Kasneci et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

LLMs possess notable capabilities for education, but these are yet to be fully explored.

“LLMs possess notable capabilities for education, but these are yet to be fully explored.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

While privacy issues are genuine, they should not hinder AI's integration into higher education.

“In summary, while privacy issues are genuine, they should not hinder AI's integration into higher education.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The US Federal Trade Commission is scrutinising OpenAI for potential false or harmful statements about real individuals and its data privacy methods.

“The US Federal Trade Commission is also scrutinising OpenAI for potential false or harmful statements about real individuals and its data privacy methods (BBC News, 2023).”

• cites BBC News, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Young learners could unintentionally share personal details with ChatGPT, underscoring the need to protect the privacy of users, particularly the younger demographic.

“Young learners could unintentionally share personal details with ChatGPT, underscoring the need to protect the privacy of users, particularly the younger demographic.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

The widespread integration of LLMs in higher education can offer substantial advantages for students and faculty.

“The widespread integration of LLMs in higher education can offer substantial advantages for students and faculty.”

May combine two claims.

Many users still fall victim to misinformation despite the need to verify AI-provided information, due to cognitive biases, reliance on headlines, and the persistence of false information.

“While users should verify AIprovided information, many still fall victim to misinformation (Lim et al., 2023), due to cognitive biases, reliance on headlines, and the persistence of false information (Centre for Information Technology and Society, 2023).”

• cites Lim et al., 2023Not checked yet

  Not checked in the free run.

• cites Centre for Information Technology and Society, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The fact that AI may produce biased responses represents a valuable opportunity to raise awareness and engage in a discussion about inherent biases which are present in society and therefore reflected in the LLM's training data.

“Moreover, the fact that AI may produce biased responses represents a valuable opportunity to raise awareness and engage in a discussion about inherent biases which are present in society and therefore reflected in the LLM's training data (Heaven, 2023).”

• cites Heaven, 2023Not checked yet

  Not checked in the free run.

May combine two claims.

AI poses challenges but also offers solutions to bolster equality, diversity, and inclusion.

“While AI poses challenges, it also offers solutions to bolster equality, diversity, and inclusion.”

May combine two claims.

Chatbots can handle course queries, direct students to resources, provide materials suited to different learning styles, and simplify complex ideas, benefiting those with communication challenges.

“They can handle course queries, direct students to resources, provide materials suited to different learning styles (Gupta and Chen, 2022), and simplify complex ideas, benefiting those with communication challenges (Hemsley et al., 2023).”

• cites Gupta and Chen, 2022Not checked yet

  Not checked in the free run.

• cites Hemsley et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

While the ability of AI to bolster EDI is clear, the risks associated with AI use can be managed.

“In conclusion, while the ability of AI to bolster EDI is clear, the risks associated with AI use can be managed.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

As tools like ChatGPT become more common at work, graduates must be equipped with the right knowledge and skills to use them adeptly.

“As tools like ChatGPT become more common at work, graduates must be equipped with the right knowledge and skills to use them adeptly.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT's initial answer could prompt further questioning and encourage students to apply their knowledge and reasoning skills.

“Gilson et al. (2023), note that ChatGPT's initial answer could prompt further questioning and encourage students to apply their knowledge and reasoning skills”

• cites Gilson et al. (2023)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Once students have the necessary knowledge and skills, they can evaluate the outputs generated by AI, weeding out inaccuracies and leveraging the beneficial and helpful components of the responses offered by the technology.

“Once students have the necessary knowledge and skills, they can evaluate the outputs generated by AI, weeding out inaccuracies and leveraging the beneficial and helpful components of the responses offered by the technology.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Fictional references, which are often markers of AI-generated content, can help educators pinpoint students falsely claiming AI work as their own.

“Moreover, fictional references, which are often markers of AI-generated content, can help educators pinpoint students falsely claiming AI work as their own.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

ChatGPT is capable of generating computer code snippets in different programming languages, crafting tables and lists, and creating Excel formulas.

“It is capable of generating computer code snippets in different programming languages, crafting tables and lists, and creating Excel formulas.”

May combine two claims.

Japanese authorities have voiced privacy concerns, emphasising the need to weigh these against the benefits of generative AI.

“Similarly, Japanese authorities have voiced privacy concerns, emphasising the need to weigh these against the benefits of generative AI (Kaur, 2023).”

• cites Kaur, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

While these outputs may require some editing and verification, ChatGPT serves as a valuable time-saving resource.

“While these outputs may require some editing and verification, ChatGPT undeniably serves as a valuable time-saving resource (Centre for Teaching and Learning, 2023).”

• cites Centre for Teaching and Learning, 2023Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

Generative AI, when used appropriately in education, offers significant potential to enhance EDI and promote inclusive learning.

“Generative AI, when used appropriately in education, offers significant potential to enhance EDI and promote inclusive learning.”

May combine two claims.

The Covid-19 pandemic illustrated how AI can mass-produce questionable content and amplify misinformation through social media bots.

“The Covid-19 pandemic illustrated how AI can mass-produce questionable content and amplify misinformation through social media bots (Gisondi, et al., 2023).”

• cites Gisondi, et al., 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The privacy implications of ChatGPT are particularly relevant for learners and educators who may lack in-depth knowledge of technology and privacy.

“The privacy implications of ChatGPT are particularly relevant for learners and educators who may lack in-depth knowledge of technology and privacy.”

Probably not a substantial claim: may not pass S3 (load-bearing).

OpenAI responded to Italy's restriction by addressing and clarifying privacy matters.

“OpenAI responded by addressing and clarifying these privacy matters ( McCallum, 2023).”

• cites McCallum, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Generative AI could be used as an aid to improve writing and research skills.

“while Rudolph et al. (2023) state that generative AI could be used as an aid to improve writing and research skills.”

• cites Rudolph et al. (2023)Not checked yet

  Not checked in the free run.

May combine two claims.

For neurodivergent students, AI can assist in time management, information processing, and thought organisation.

“For neurodivergent students, AI can assist in time management, information processing, and thought organisation (McMurtrie, 2023).”

• cites McMurtrie, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Companies developing LLMs are aware of these issues and are working to address them.

“Companies developing LLMs are aware of these issues and are working to address them (O'Brien, 2023).”

• cites O'Brien, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Integrating AI models into curricula can nurture creativity, promote critical thinking, and prepare students for a future where AI collaboration is commonplace in professional life.

“Integrating these AI models into curricula can nurture creativity, promote critical thinking, and prepare students for a future where AI collaboration is commonplace in professional life.”

May combine two claims.

ChatGPT does not operate based on ethical principles, nor can it discern between right and wrong, or truth and falsehood.

“It is crucial to understand that ChatGPT does not operate based on ethical principles, nor can it discern between right and wrong, or truth and falsehood.”

• cites Sabzalieva and Valentini, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Generative AI is set to enhance the educational landscape through aiding research, boosting writing skills, and enabling innovative problem-solving.

“Through aiding research, boosting writing skills, and enabling innovative problem-solving, this technology is set to enhance the educational landscape”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

The multifaceted attributes of generative AI and LLMs underscore heightened efficiency and democratise AI, making it an accessible tool for diverse audiences.

“Their multifaceted attributes not only underscore heightened efficiency but also democratise AI, making it an accessible tool for diverse audiences.”

Probably not a substantial claim: may not pass S3 (load-bearing).

OpenAI claims to have fine-tuned ChatGPT 4.0, achieving an 82% reduction in the model's tendency to produce disallowed content compared to its predecessor.

“OpenAI has recognised this issue and claims to have fine-tuned ChatGPT 4.0, achieving an 82% reduction in the model's tendency to produce disallowed content compared to its predecessor.”

• cites OpenAI, 2023bNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

ChatGPT has a pro-environmental, left-libertarian bias.

“Furthermore, Hartmann et al. (2023) highlight ChatGPT's pro-environmental, left-libertarian bias.”

• cites Hartmann et al. (2023)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Generative AI and LLMs, with their capacity for personalised lessons and engaging with complex subjects, hold the potential to transform traditional classroom settings.

“Generative AI and LLMs, with their capacity for personalised lessons and engaging with complex subjects, hold the potential to transform traditional classroom settings.”

May combine two claims.

ChatGPT merely gathers data from the databases and texts it processes online, thereby inheriting any cognitive biases present in that information.

“This tool merely gathers data from the databases and texts it processes online, thereby inheriting any cognitive biases present in that information(Sabzalieva and Valentini, 2023).”

• cites Sabzalieva and Valentini, 2023Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Generative AI is still a relatively novel technology, and the question of how to successfully integrate it within higher education remains unanswered.

“Currently, generative AI is still a relatively novel technology, therefore the question of how to successfully integrate it within higher education remains unanswered.”

Probably not a substantial claim: may not pass S3 (load-bearing).

GPT-4 is 29% more aligned with company safety policies on sensitive data handling.

“Additionally, GPT-4 is 29% more aligned with company safety policies on sensitive data handling(OpenAI, 2023b).”

• cites OpenAI, 2023bNot checked yet

  Not checked in the free run.

May combine two claims.

LLMs can perpetuate and amplify existing biases and unfairness in society, which can negatively impact teaching and learning processes and outcomes.

“Overall, LLMs can perpetuate and amplify existing biases and unfairness in society, which can negatively impact teaching and learning processes and outcomes(Abdelghani et al., 2023).”

• cites Abdelghani et al., 2023Not checked yet

  Not checked in the free run.