The role of ChatGPT in higher education: Benefits, challenges, and future research directions

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

Recent studies indicate GPT-3 can stimulate curiosity, improve students' question-asking skills, and generate programming code explanations.

“Furthermore, recent studies have shown that GPT-3 can stimulate curiosity, enhance students' questionasking skills, and generate programming code explanations (Abdelghani et al., 2022;MacNeil et al., 2022).”

• cites Abdelghani et al., 2022Not checked yet

  Not checked in the free run.

• cites MacNeil et al., 2022Not checked yet

  Not checked in the free run.

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

GPT uses deep learning and algorithms to perform language tasks such as text generation, question answering, and translation while using context to produce human-like responses.

“It leverages deep learning and powerful algorithms to perform various language-related tasks, such as text generation, question answering, and translation, while comprehending the context to produce responses that resemble human language (Lund et al., 2023).”

• cites Lund et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT's feedback can help students detect errors and guide them toward improvement, making it an effective More Knowledgeable Other in learning.

“This feedback can help students detect errors and guide them towards successful improvement, making ChatGPT an effective”

• cites Geng & Razali, 2020Not checked yet

  Not checked in the free run.

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

OpenAI introduced GPT in 2018 as an LLM designed to mimic human language processing.

“Generative Pre-trained Transformer (GPT), which was released by OpenAI (San Francisco, California) in 2018, is a type of Large Language Model (LLM) that aims to replicate human language processing capabilities (Cascella et al., 2023).”

• cites Cascella 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.

ChatGPT-3.5 appeared in November 2022, and ChatGPT-4 followed in March 2023.

“OpenAI released the ChatGPT-3.5language model family in November 2022 and, subsequently, the ChatGPT-4 family in March 2023 (Skavronskaya et al., 2023).”

• cites Skavronskaya et al., 2023Not checked yet

  Not checked in the free run.

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

Large language models have become of interest in higher education because of their wide range of applications.

“Large language models have become a subject of interest in higher education due to their extensive range of applications, especially as there are now other similar options, such as Bing Chat, Bard, and Ernie (Rudolph et al., 2023b).”

• cites Rudolph et al., 2023bNot checked yet

  Not checked in the free run.

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

ChatGPT can hold coherent, contextually relevant conversations by responding based on its understanding of the prompt's language and context.

“This chatbot can engage in coherent and contextually relevant conversations by responding based on its comprehension of the language and context of the prompts (Gilson et al., 2023;Pavlik, 2023).”

• cites Gilson et al., 2023Not checked yet

  Not checked in the free run.

• cites Pavlik, 2023Not checked yet

  Not checked in the free run.

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

Academics acknowledge that responsible AI integration into education is critical.

“Academics acknowledge that ensuring the responsible integration of AI into education is critical (Fadel et al., 2019;Polak et al., 2022).”

• cites Fadel et al., 2019Not checked yet

  Not checked in the free run.

• cites Polak et al., 2022Not checked yet

  Not checked in the free run.

May combine two claims.

Incorporating ChatGPT in learning enables educators and students to benefit from personalised learning experiences, efficient resource use, and adaptive learning approaches that enhance learning outcomes, consistent with constructivist principles.

“Overall, incorporating ChatGPT in the learning process enables educators and students to benefit from personalised learning experiences, efficient and effective use of resources, and adaptive learning approaches that enhance the learning outcomes for all, consistent with the principles of constructivist theory (Bada & Olusegun, 2015;Stapleton & Stefaniak, 2019).”

• cites Bada & Olusegun, 2015Not checked yet

  Not checked in the free run.

• cites Stapleton & Stefaniak, 2019Not checked yet

  Not checked in the free run.

May combine two claims.

LLMs can help conduct initial literature reviews, summarise research papers, generate draft versions of research papers, and assist authors from non-English-speaking backgrounds with language barriers.

“Such LLMs can help conduct initial literature reviews, summarise research papers, generate draft versions of research papers (Rahman et al., 2023;Rudolph et al., 2023a), and even assist authors from non-English speaking backgrounds in overcoming language barriers (Gao et al., 2022).”

• cites Rahman et al., 2023Not checked yet

  Not checked in the free run.

• cites Rudolph et al., 2023aNot checked yet

  Not checked in the free run.

• cites Gao et al., 2022Not checked yet

  Not checked in the free run.

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

Innovative assessment activities in higher education help students engage with learning resources, think critically, and have authentic learning experiences.

“In the higher education sector, innovative assessment activities have gained a lot of attention because they assist students in getting involved with learning resources to think critically and have real learning experiences (Boud & Soler, 2016).”

• cites Boud & Soler, 2016Not checked yet

  Not checked in the free run.

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

Because ChatGPT can give individualised feedback grounded in constructivist learning theory, it may substantially improve learning experiences and help students succeed.

“Overall, ChatGPT's ability to provide individualised feedback based on the constructivist theory of learning has significant potential to enhance the learning experience and promote successful learning outcomes.”

May combine two claims.

ChatGPT's research and analysis support, though currently basic, has potential to fundamentally impact research and higher education, depending on prompt quality.

“ChatGPT's research and analysis support, though currently at a basic level, has the potential to fundamentally impact research and higher education, depending on the quality of the prompts (Dwivedi et al., 2023).”

• cites Dwivedi et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT may be prone to hallucinations and may fabricate references and quotes, so users should exercise caution.

“Nevertheless, users should exercise caution as ChatGPT may be prone to hallucinations (Alkaissi & McFarlane, 2023) and fabricate references and quotes (Sallam, 2023;Shen et al., 2023).”

• cites Alkaissi & McFarlane, 2023Not checked yet

  Not checked in the free run.

• cites Sallam, 2023Not checked yet

  Not checked in the free run.

• cites Shen et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT can act as a research assistant by answering users' questions based on related literature it has learned and by analysing data.

“ChatGPT can act as a research assistant, answering users' questions based on the related literature it has learned (Lin, 2023), and analysing data (Goel, 2020).”

• cites Lin, 2023Not checked yet

  Not checked in the free run.

• cites Goel, 2020Not checked yet

  Not checked in the free run.

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

Automated administrative support through ChatGPT can use data and analytics to provide timely, personalised non-academic feedback to students, such as deadline notifications, task reminders, and progress reports.

“Automated administrative support through ChatGPT can use data and analytics to provide timely and personalised non-academic feedback to students, such as notifications about upcoming deadlines, reminders about incomplete tasks, and progress reports on completed tasks, based on individual needs and preferences.”

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

The constructivist theory of learning supports using ChatGPT for automated administrative support because it can facilitate active participation, provide personalised feedback, and create a supportive learning environment.

“While further research is needed to fully understand the potential of ChatGPT in the tertiary education sector, the constructivist theory of learning supports the use of ChatGPT for automated administrative support as it can facilitate active participation in learning, provide personalised feedback to students, and create a supportive learning environment.”

May combine two claims.

For academic staff, ChatGPT may summarise and clarify student emails for administrative processing and generate personalised response templates for student queries.

“For academic staff, ChatGPT may be able to summarise and clarify student emails for administrative members to process more efficiently and generate personalised response templates for staff to address students' queries (Dwivedi et al., 2023).”

• cites Dwivedi et al., 2023Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT can provide individualised feedback based on students' prompts, making learning more rewarding.

“As indicated earlier, ChatGPT can provide individualised feedback based on the students' prompts, making learning a more rewarding experience (Bridges, 2009;Weldy & Turnipseed, 2010).”

• cites Bridges, 2009Not checked yet

  Not checked in the free run.

• cites Weldy & Turnipseed, 2010Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT can use big data and learning analytics to monitor student performance, predict success, and respond to students' emotional states in real time, resulting in personalised adaptive learning consistent with constructivist theory.

“In contributing to a smart learning environment, ChatGPT can utilise big data and learning analytics to monitor student performance, predict success, and respond to students, including their emotional states, in real-time, resulting in personalised adaptive learning (PAL) that is consistent with constructivist theory (Peng et al., 2019;Rudolph et al., 2023a).”

• cites Peng et al., 2019Not checked yet

  Not checked in the free run.

• cites Rudolph et al., 2023aNot checked yet

  Not checked in the free run.

May combine two claims.

The integration of ChatGPT can help create a supportive learning environment by providing timely and accurate information, reducing administrative burdens, and saving costs for higher education institutions.

“The integration of ChatGPT can assist in creating a supportive learning environment for students by providing timely and accurate information, reducing administrative burdens, and presenting a cost-saving measure for higher education institutions.”

May combine two claims.

Automated administrative support provides efficiency and effectiveness, allows synchronous interaction for students, and presents a cost-savings measure for higher education institutions.

“This provides a degree of efficiency and effectiveness, allowing for a synchronous interaction for students (Howlett, 2017;Okonkwo & Ade-Ibijola, 2021) and presenting a cost-savings measure for higher education institutions (Merelo et al., 2022).”

• cites Howlett, 2017Not checked yet

  Not checked in the free run.

• cites Okonkwo & Ade-Ibijola, 2021Not checked yet

  Not checked in the free run.

• cites Merelo et al., 2022Not checked yet

  Not checked in the free run.

May combine two claims.

Implementing advanced LLMs like ChatGPT has the potential to revolutionise tertiary education by automating some administrative support and providing efficiency and effectiveness.

“It is obvious that the implementation of advanced LLMs like ChatGPT has the potential to revolutionise the tertiary education sector by automating some elements of administrative support and providing a degree of efficiency and effectiveness.”

May combine two claims.

ChatGPT's conversational nature facilitates active construction of students' knowledge through continuous engagement, pattern finding, and learning by experimentation.

“The conversational nature of LLMs such as ChatGPT facilitates the active construction of students' knowledge as they are continuously engaged with the task, encouraged to find patterns through a scaffolded approach (Stapleton & Stefaniak, 2019), and learn through experimentation and experience, which is an important part of knowledge generation (Rudolph et al., 2023a).”

• cites Stapleton & Stefaniak, 2019Not checked yet

  Not checked in the free run.

• cites Rudolph et al., 2023aNot checked yet

  Not checked in the free run.

May combine two claims.

AI applications like ChatGPT can provide accurate and efficient individualised feedback and automated grading, but users must carefully check outputs because they depend on prompts.

“AI applications like ChatGPT can provide accurate and efficient individualised feedback and automated grading, but users need to carefully check the outputs as they depend on the prompts (Rudolph et al., 2023a).”

• cites Rudolph et al., 2023aNot checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT-generated text may contain factual biases due to biased training data, which could perpetuate learners' misconceptions.

“Furthermore, ChatGPT-generated text may contain factual biases due to biased training data, which could perpetuate misconceptions held by learners (Karim, 2023).”

• cites Karim, 2023Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

According to constructivism, learning is best when instructor-learner interactions are good.

“According to the constructivism learning theory, learning happens best when there are good interactions between the instructor and learner (Schuh, 2003).”

• cites Schuh, 2003Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT challenges ethical and equity practices in higher education and potentially contradicts constructivist learning theory.

“It is no secret that ChatGPT challenges ethical and equity practices in the higher education sector, as it potentially contradicts the constructivist theory of learning that emphasises active student participation and the construction of knowledge.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

May combine two claims.

Using ChatGPT as a tool for exploration and inquiry allows students to actively construct their own knowledge and meaning, reducing the likelihood of academic dishonesty such as plagiarism or cheating.

“Using ChatGPT as a tool for exploration and inquiry, students can actively construct their own knowledge and meaning, reducing the likelihood of academic dishonesty, such as plagiarism or cheating (Keith, 2022).”

• cites Keith, 2022Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT can generate initial ideas for assessment design, create multiple-choice or short-answer questions, and produce drafts of case studies or other assessments for editing.

“ChatGPT can generate initial ideas for assessment design, create multiple-choice or short-answer questions for academics, and produce drafts of case studies or other assessments for further editing (Bridgeman et al., 2023;Liu & Bridgeman, 2023).”

• cites Bridgeman et al., 2023Not checked yet

  Not checked in the free run.

• cites Liu & Bridgeman, 2023Not checked yet

  Not checked in the free run.

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

Creating authentic learning experiences requires collaborative and consultative learning experiences.

“Creating authentic learning experiences would require collaborative and consultative learning experiences (Muhajirah, 2020).”

• cites Muhajirah, 2020Not checked yet

  Not checked in the free run.

May combine two claims.

Passive shortcuts that may breach academic integrity hinder learners' active involvement and impede learning.

“Passive shortcuts, potentially resulting in academic integrity breaches, hinder the active involvement of learners and hence impede learning.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

May combine two claims.

A challenge of ChatGPT is acceptance of AI-provided feedback over human instructor feedback, which conflicts with constructivist theory's emphasis on interactions and social collaboration.

“Another challenge of ChatGPT is the acceptance of feedback provided by AI rather than human instructors, which is against the constructivist theory of learning that emphasises interactions and social collaboration in learning (Hein, 1991).”

• cites Hein, 1991Not checked yet

  Not checked in the free run.

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

May combine two claims.

ChatGPT could impact students' ability to actively construct their own knowledge because some students have greater access than others, resulting in a knowledge gap.

“However, ChatGPT could impact students' ability to actively construct their own knowledge, as some students may have greater access to ChatGPT than others, resulting in a knowledge gap between students (Hein, 1991).”

• cites Hein, 1991Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Misusing ChatGPT to instantly create content as a shortcut conflicts with constructivism and learning theory.

“The misuse of ChatGPT to create content instantly as a shortcut goes against the philosophy of constructivism and any learning theory, for that matter.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

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

ChatGPT's ability to develop innovative and authentic student assessments depends on its focus on the work context and the students' existing knowledge.

“Proposition 10 (P10): ChatGPT's ability to develop innovative and authentic student assessments depends on its focus on the work context and the students' existing knowledge.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

May combine two claims.

ChatGPT can be integrated into assessment tasks where students critique generated text or essays and build high-quality articles from drafts, provoking existing mental models and developing critical thinking.

“It can also be integrated into assessment tasks, where students critique generated text or essays and build high-quality articles based on generated drafts, provoking students' existing mental models and developing critical thinking skills (Dennick, 2016).”

• cites Dennick, 2016Not checked yet

  Not checked in the free run.

May combine two claims.

ChatGPT can be integrated into innovative assessment activities to facilitate collaborative learning, scaffolding, real-time feedback, personalised learning, scalability, interactivity, and knowledge creation and dissipation.

“This technology can be integrated into innovative assessment activities, facilitating collaborative learning, scaffolding, real-time feedback, personalised learning, scalability, interactivity, and fostering knowledge creation and dissipation effectively (Kumar, 2021).”

• cites Kumar, 2021Not checked yet

  Not checked in the free run.

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

Integrating ChatGPT into innovative assessment activities can promote critical thinking, problem-solving, and collaboration skills among students.

“Proposition 12 (P12): The integration of ChatGPT in innovative assessment activities can promote critical thinking, problemsolving, and collaboration skills among students.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.

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

The learning environment includes social groups, instructional strategies, and a motivational atmosphere.

“Furthermore, the learning environment includes social groups, instructional strategies, and a motivational atmosphere, to name a few (Zajda, 2021).”

• cites Zajda, 2021Not checked yet

  Not checked in the free run.

May combine two claims.

Chatbot technology has shown positive influences on explicit reasoning, learning achievement, knowledge retention, and learning interest.

“Although chatbot technology has shown positive influences on explicit reasoning, learning achievement, knowledge retention, and learning interest, studies have not yet demonstrated significant improvements in critical thinking, learning engagement, and motivation (Deng & Yu, 2023).”

• cites Deng & Yu, 2023Not checked yet

  Not checked in the free run.

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

Using ChatGPT to facilitate learning could lead to unethical and inequitable practices and destroy the spirit of learning.

“Using ChatGPT to facilitate learning could lead to unethical and inequitable practices, destroying the spirit of learning (Hein, 1991).”

• cites Hein, 1991Not checked yet

  Not checked in the free run.

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

Constructivist theory suggests learners actively develop knowledge through experiences and interactions rather than passively acquiring it from external tools like ChatGPT.

“However, the constructivist theory of learning suggests that learners actively develop knowledge for themselves through experiences and interactions with others rather than passively acquiring it through external tools like ChatGPT (Geary, 1995).”

• cites Geary, 1995Not checked yet

  Not checked in the free run.

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

Incorporating AI literacy as part of graduate skills could enhance graduates' employability and preparedness for the rapidly evolving job market.

“Therefore, incorporating AI literacy as part of graduate skills could enhance graduates' employability and preparedness for the rapidly evolving job market.”

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

The use of ChatGPT in learning and assessment can impact the development of graduate skills, such as critical thinking and problem-solving.

“Proposition 18 (P18): The use of ChatGPT in learning and assessment can impact the development of graduate skills, such as critical thinking and problem-solving.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Graduate skills including critical thinking, problem-solving, communication, collaboration, teamwork, leadership, adaptability, digital literacy, global and cultural awareness, ethics, and professionalism are essential for future professional and personal success.

“Graduate skills, including critical thinking and problemsolving, communication, collaboration and teamwork, leadership, adaptability, digital literacy, global and cultural awareness, ethics, and professionalism, are essential for future professional and personal success (Abelha et al., 2020;Osmani et al., 2019;Oliver & de St Jorre, 2018; University of Adelaide, 2022; University of Sydney, 2022).”

• cites Abelha et al., 2020Not checked yet

  Not checked in the free run.

• cites Osmani et al., 2019Not checked yet

  Not checked in the free run.

• cites Oliver & de St Jorre, 2018Not checked yet

  Not checked in the free run.

May combine two claims.

Reliance on ChatGPT could make it challenging to evaluate the effectiveness of group learning activities and assess students' ability to work collaboratively.

“This could also make it challenging to evaluate the effectiveness of group learning activities and assess students' ability to work collaboratively.”

May combine two claims.

Using ChatGPT for assessment is a passive process that does not allow social interaction, hindering students' ability to construct meaning through reflection.

“However, the use of ChatGPT for assessment is a passive process and does not allow for social interaction, hindering students' ability to construct meaning through reflection on their experiences (Biggs, 2014).”

• cites Biggs, 2014Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Using ChatGPT presents an opportunity to incorporate AI literacy as part of graduate skills, preparing graduates for workplace application of LLMs that may replace some jobs and create new ones.

“The use of ChatGPT presents an opportunity to incorporate artificial intelligence literacy as part of graduate skills, preparing graduates for effective workplace application of large language models that may replace some existing jobs and create new ones (Cradle, 2023).”

• cites Cradle, 2023Not checked yet

  Not checked in the free run.

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

Incorporating discovery-based and experiential learning methods into curriculum and assessment could promote graduate skill development beyond ChatGPT alone.

“Therefore, the incorporation of these methods into the curriculum and assessment could promote the development of graduate skills beyond the use of ChatGPT alone.”

• AssumptionRests on an assumption: Constructivist learning theory is the appropriate framework for evaluating ChatGPT's role in higher education.