PaperlayerView exact revision on arXiv

Source-linked AI summary

Students' Voices on Generative AI: Perceptions, Benefits, and Challenges in Higher Education

Cecilia Ka Yuk Chan, Wenjie Hu

arXiv:2305.00290v1cs.CYcs.AI

TL;DR

Students’ perceptions of generative AI in higher education remain underexamined despite its growing use and potential influence on learning. Using a survey of 399 Hong Kong university students, this study finds generally positive attitudes alongside concerns about reliability, privacy, ethics, and broader impacts.

  • Problem

    Student perceptions and experiences of GenAI in higher education remain underexamined, limiting evidence for understanding its effective integration into teaching and learning.

  • Method

    The study surveyed 399 undergraduate and postgraduate students in Hong Kong using closed- and open-ended online questions, analyzed descriptively and thematically.

  • Results

    Students generally understood GenAI’s capabilities and limitations and held positive attitudes, while expressing concerns about reliability, privacy, ethics, and impacts on development, careers, and societal values.

  • Takeaways & Limitations

    Considering student perceptions can help educators and policymakers tailor GenAI integration to students’ needs and concerns while promoting effective learning outcomes.

  • Takeaways & Limitations

    The relatively small sample, self-reported data, and cross-sectional design limit generalizability and prevent examining changes in perceptions over time.

Abstract

from arXiv · show

This study explores university students' perceptions of generative AI (GenAI) technologies, such as ChatGPT, in higher education, focusing on familiarity, their willingness to engage, potential benefits and challenges, and effective integration. A survey of 399 undergraduate and postgraduate students from various disciplines in Hong Kong revealed a generally positive attitude towards GenAI in teaching and learning. Students recognized the potential for personalized learning support, writing and brainstorming assistance, and research and analysis capabilities. However, concerns about accuracy, privacy, ethical issues, and the impact on personal development, career prospects, and societal values were also expressed. According to John Biggs' 3P model, student perceptions significantly influence learning approaches and outcomes. By understanding students' perceptions, educators and policymakers can tailor GenAI technologies to address needs and concerns while promoting effective learning outcomes. Insights from this study can inform policy development around the integration of GenAI technologies into higher education. By understanding students' perceptions and addressing their concerns, policymakers can create well-informed guidelines and strategies for the responsible and effective implementation of GenAI tools, ultimately enhancing teaching and learning experiences in higher education.

Generative Artifical Intelligence · Benefits and challenges of using generative AI in higher education

Generative AI comprises algorithms that generate data resembling existing datasets and includes tools such as ChatGPT, Bard, Stable Diffusion, and Dall-E. In higher education, these tools may enhance learning and writing support, but raise concerns about limitations, ethics, integrity, accuracy, harmful content, and the need for human oversight.

  • Generative Artifical Intelligence: Generative AI uses machine-learning algorithms to generate new data samples that mimic existing datasets.Variational autoencoders encode and decode data while maintaining essential features, while generative adversarial networks use competing neural networks.
  • Generative Artifical Intelligence: Examples of GenAI tools include ChatGPT, Bard, Stable Diffusion, and Dall-E.Their ability to handle complex prompts and produce human-like output has stimulated research across fields including education.
  • Generative Artifical Intelligence: ChatGPT is an autoregressive large language model with more than 175 billion parameters that generates human-like responses to text-based inputs.Its release in November 2022 prompted increased interest in GenAI use in higher education.
  • Benefits and challenges of using generative AI in higher education: GenAI can enhance students’ learning experiences by producing highly original outputs in response to user prompts.Text-to-text tools can support brainstorming and writing feedback, especially for non-native students, while text-to-image tools can provide additional educational support.
  • Benefits and challenges of using generative AI in higher education: GenAI outputs may be mostly original and relevant yet contain inappropriate references and lack personal perspectives.These limitations are linked to concerns about ethics, plagiarism, academic integrity, and the inability of AI to produce personal perspectives.
  • Benefits and challenges of using generative AI in higher education: GenAI-generated content may be biased, inaccurate, or harmful when training data contain such elements.Generated images may include nudity or obscenity and may support malicious uses such as deepfakes.
  • Benefits and challenges of using generative AI in higher education: GenAI tools cannot assess content validity or determine whether their outputs contain falsehoods or misinformation.Their use therefore requires human oversight.
  • Benefits and challenges of using generative AI in higher education: The study examines students’ familiarity with GenAI, perceived benefits and challenges in teaching and learning, and effective integration to enhance outcomes.These questions frame the higher-education focus of the section.

Student perceptions of the use of GenAI in higher education · Methodology

The paper examines university students’ perceptions and experiences of GenAI because student perceptions can shape learning approaches and outcomes, while research on GenAI perceptions remains limited. Using an online survey of 399 Hong Kong undergraduate and postgraduate students with closed- and open-ended questions, the study investigates GenAI knowledge, higher-education integration, and related challenges.

  • Student perceptions of the use of GenAI in higher education: Student perceptions of their learning environment, abilities, and teaching strategies influence learning approaches and outcomes under Biggs’ 3P model.The model frames student perception as important to teaching and learning and links perceptions to learning outcomes.
  • Student perceptions of the use of GenAI in higher education: Research on AI tools has reported benefits for language acquisition, including grammar assistance, idea generation, and target-language communication.The cited studies examined tools such as chatbots, Plot Generator, and AI KAKU in language-learning contexts.
  • Student perceptions of the use of GenAI in higher education: Students’ AI knowledge can coincide with lower anxiety but also concerns about AI’s effects on human jobs.Prior findings therefore indicate that positive perceptions and apprehensions may coexist.
  • Student perceptions of the use of GenAI in higher education: 86 students in a Turkish university showed no correlation between chatbot-use frequency and visual design self-efficacy, course satisfaction, chatbot-use satisfaction, and learner autonomy.This finding suggests that frequency of use alone may not meaningfully explain student perceptions.
  • Student perceptions of the use of GenAI in higher education: Most research on student perceptions of AI or GenAI uses quantitative survey designs.Some studies additionally use open-ended survey questions or semi-structured interviews to collect free responses and probe students’ views.
  • Student perceptions of the use of GenAI in higher education: Research has substantially examined AI generally but has insufficiently investigated how students perceive GenAI and how it can be integrated into higher education.The study addresses this gap amid unprecedented interest in GenAI and its potential to enhance teaching and learning.
  • Methodology: The study used an online survey of university students in Hong Kong with closed- and open-ended questions about GenAI knowledge, higher-education integration, and potential challenges.The survey was designed to gather a large population of responses about students’ use and perceptions of GenAI in teaching and learning.
  • Methodology: 399 undergraduate and postgraduate students were recruited through convenience sampling from Hong Kong post-secondary institutions using an online platform and informed consent.Participants were selected based on availability and willingness to participate.

Results · Demographic information · Knowledge of Generative AI Technologies

Participants represented ten faculties, with near-balanced gender representation and more postgraduate than undergraduate students; 55.4% were enrolled in STEM fields. Students generally understood GenAI technologies, and knowledge was positively associated with usage frequency, including stronger recognition of potential factual inaccuracies among more frequent users.

  • Demographic information: Ten faculties were represented, spanning Architecture, Arts, Business, Dentistry, Education, Engineering, Law, Medicine, Science, and Social Sciences.
  • Demographic information: 204 males (51.1%) and 195 females (48.9%) participated in the study.
  • Demographic information: 55.6% (n = 222) were postgraduate students, compared with 44.4% (n = 177) undergraduates.
  • Demographic information: 55.4% (n = 221) were enrolled in STEM fields, mainly Engineering (33.1%) and Science (14.5%).
  • Knowledge of Generative AI Technologies: Knowledge scores ranged from 3.89 to 4.15, indicating generally good understanding of GenAI technologies.
  • Knowledge of Generative AI Technologies: The highest-rated understanding concerned GenAI limitations in complex tasks (Mean=4.15, SD=0.82), while emotional intelligence and empathy considerations scored lowest (Mean=3.89, SD=0.97).
  • Knowledge of Generative AI Technologies: Knowledge of GenAI technologies showed a moderate positive correlation with frequency of use (r=0.1, p<0.05).
  • Knowledge of Generative AI Technologies: Students using GenAI at least sometimes rated factual-inaccuracy awareness higher (Mean=4.22 SD=0.829) than never or rarely users (Mean=3.99, SD=0.847), t=2.695, p<0.01.

Willingness to use Generative AI Technologies

Students expressed strong willingness to integrate GenAI into their learning practices and future careers, valuing its usefulness, accessibility, and personalized support. Willingness was positively associated with students’ knowledge and frequency of GenAI use.

  • Willingness to use Generative AI Technologies: Students reported willingness to integrate GenAI into future learning practices (Mean=3.85, SD=1.02) and careers (Mean=4.05; SD=0.96).Students also believed GenAI could improve digital competence (3.70, 0.96).
  • Willingness to use Generative AI Technologies: Willingness to use GenAI was positively correlated with knowledge of GenAI (r=0.189; p<0.001) and frequency of use (r=0.326; p<0.001).Students with greater knowledge and more frequent use were more likely to intend future use.
  • Willingness to use Generative AI Technologies: Students valued GenAI for saving time (4.20, 0.82), providing unique insights (Mean=3.74; SD=1.08), and offering personalized feedback (Mean=3.61; SD=1.06).These perceived benefits supported positive attitudes toward integrating GenAI into learning.
  • Willingness to use Generative AI Technologies: Students regarded GenAI as accessible around the clock (4.12, 0.83) and useful for anonymous student support services (3.77, 0.99).Availability and anonymity were identified as practical features of GenAI technologies such as ChatGPT.

Concerns about Generative AI Technologies · The Benefits and Challenges for Students’ Willingness and Concerns · (1) Personalized and immediate learning support

Students viewed GenAI as a source of personalized, immediate, and tailored learning support, while students’ concerns varied by usage experience but were not significantly correlated with GenAI knowledge. The findings also highlight applications in language learning and future teaching.

  • Concerns about Generative AI Technologies: Students who never or rarely used GenAI differed significantly from other participants in their concerns.The difference was statistically significant (t=3.873, p<0.01).
  • Concerns about Generative AI Technologies: Students’ concerns were not significantly correlated with their knowledge about GenAI technologies.The reported correlation was r=0.096; p>0.05.
  • The Benefits and Challenges for Students’ Willingness and Concerns: When students struggle with assignments, GenAI can function as a virtual tutor by providing personalized learning support.Students described ChatGPT as useful when they had doubts and could not find others to help.
  • (1) Personalized and immediate learning support: Students valued GenAI’s ability to answer questions immediately and provide customized recommendations and feedback.These features were described as useful alongside immediate answers.
  • (1) Personalized and immediate learning support: GenAI can provide learning resources tailored to students’ specific needs.Students proposed using ChatGPT to generate short texts from entered words to support second-language vocabulary memorization.
  • The Benefits and Challenges for Students’ Willingness and Concerns: Some education students believed GenAI could assist them in their future teaching.Their comments specifically concerned the use of ChatGPT in future educational practice.

(2) Writing and brainstorming support · (3) Research and analysis support · (4) Visual and audio multi-media support

Students viewed GenAI as support for generating and refining writing, conducting research and analysis, and creating visual content. They also saw value in automating repetitive tasks to free time for study and research.

  • (2) Writing and brainstorming support: GenAI can help students generate ideas, find inspiration, and develop writing topics by providing question-based outputs as starting points.Students described ChatGPT as a convenient source of general ideas and inspiration when they struggle to begin writing.
  • (2) Writing and brainstorming support: GenAI can improve writing through grammar correction, paraphrasing, article polishing, consultation, and personalized feedback, especially for non-native or struggling writers.Students specifically identified post-writing assistance and feedback as useful for developing writing skills.
  • (3) Research and analysis support: In research, GenAI can facilitate literature searching, summarize readings, generate hypotheses from data analysis, and help researchers stay current with research trends.These capabilities were linked to AI’s ability to acquire, compile, and consolidate large amounts of information and knowledge.
  • (4) Visual and audio multi-media support: Students used GenAI to create artworks and support visualization, with AI-generated images attracting particular attention from STEM students.Examples included DALL-E, Stable Diffusion, and other AI art technologies for generating visual samples.
  • (4) Visual and audio multi-media support: GenAI can support visual and multimedia content creation when students lack ideas about how to visualize concepts.Participants described AI-generated samples and insights as useful resources for content creation.
  • (4) Visual and audio multi-media support: AI can efficiently handle repetitive or non-creative administrative tasks, giving students more time to focus on their studies and research.Participants emphasized that accelerating routine work could reduce the burden of tedious administrative activities.

What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies? · (1) Challenges concerning accuracy and transparency

Students expressed optimism about GenAI because they viewed it as part of technological evolution and believed humans would retain control and expertise. Nevertheless, more than half remained concerned about GenAI’s reliability, impact, accuracy, and transparency.

  • What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies?: Some students were optimistic about integrating GenAI because they viewed it as part of technological evolution and current technology trends.One participant compared its development with the public adoption of computers 40 years ago.
  • What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies?: Students’ willingness to use GenAI also contributed to their optimism about its future integration.
  • What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies?: Some participants were less concerned because they believed humans would maintain control and oversight while using GenAI efficiently.
  • What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies?: Students also believed GenAI would not replace human skills and expertise.
  • What are the reasons behind students’ concerns or lack of concerns regarding generative AI technologies?: More than half of participants remained concerned about GenAI’s integration, particularly its reliability and broader impact.
  • (1) Challenges concerning accuracy and transparency: GenAI can produce fluent, human-sounding responses, but students noted that the accuracy and validity of its information cannot always be verified.They feared that false information could mislead people.
  • (1) Challenges concerning accuracy and transparency: Transparency was another significant concern because the complexity and opacity of AI systems make their decisions difficult to understand.

(2) Challenges concerning privacy and ethical issues … (5) Challenges concerning human values

Students identified privacy, plagiarism, and overreliance risks alongside concerns about personal development, job replacement, societal inequality, and misalignment with human values. These challenges highlight risks to information security, academic integrity, holistic competencies, career prospects, and fairness.

  • (2) Challenges concerning privacy and ethical issues: Students, especially those in arts and social science, worried that GenAI could collect personal information from messages, creating privacy and security risks if unprotected.Students noted that messages may be used to improve the system, increasing concerns about protection of personal information.
  • (2) Challenges concerning privacy and ethical issues: Students repeatedly raised plagiarism concerns because GenAI’s development may make AI-generated or plagiarized information increasingly difficult to identify.An art student questioned whether users can distinguish an AI bot or AI-generated content as the technology improves.
  • (3) Challenges concerning holistic competencies: Students warned that overreliance on GenAI could hinder personal growth, skills, intellectual development, critical thinking, decision-making, and creativity over time.Participants specifically linked reliance on AI-provided information with reduced critical thinking and noted possible negative effects on creativity.
  • (4) Challenges concerning career prospects: Students most frequently identified job replacement as a societal risk, fearing that GenAI could eliminate jobs they are preparing for or interested in.Examples included concerns about losing future employment and AI replacing a desired GIS analyst position.
  • (5) Challenges concerning human values: Some students feared that GenAI could misalign with human values and become dangerous, while contributing to social injustice and inequality.Participants specifically mentioned widening the gap between rich and poor.
  • (5) Challenges concerning human values: Students also worried that widespread AI use could be unfair to students who do not use it and might affect academic institutions and education.The passage identifies fairness toward nonusers and possible effects on educational settings as additional societal concerns.

Discussion

Students generally understand and positively view GenAI’s potential in higher education, while retaining concerns about reliance, accuracy, transparency, privacy, ethics, and education’s value. Their perceptions support user-informed integration strategies for enhancing teaching and learning outcomes.

  • Student perceptions: Students are generally familiar with GenAI and understand its capabilities, limitations, benefits, and risks.Perceptions of benefits and risks differ according to students’ experiences with GenAI technologies.
  • Student perceptions: Students’ knowledge of GenAI and frequency of use are positively correlated, suggesting that exposure and hands-on experience may enhance understanding and acceptance.Despite GenAI’s relative novelty for public use, students appear to understand its benefits and risks well.
  • Challenges and concerns: Students recognize GenAI’s potential while expressing reservations about over-reliance, university education’s value, accuracy, transparency, privacy, and ethics.Plagiarism concerns arise because students may struggle to determine whether GenAI-generated work is original.
  • Challenges and concerns: Students’ concerns are not significantly correlated with their GenAI knowledge, indicating that understanding the technology does not eliminate reservations.GenAI tools may be unable to assess validity or identify falsehoods, creating a need for human oversight.
  • Implications for integration: Understanding students’ perceptions can help educators and policymakers integrate GenAI into higher education to enhance teaching and learning outcomes.Student acceptance is presented as key to the successful uptake of educational technological innovations.

Conclusion · Implications

The study emphasizes that students’ perceptions of GenAI shape their learning approaches and outcomes, making their willingness and concerns important for effective integration. It recommends educational, technical, curricular, and policy measures for responsible implementation that enhances learning while preparing students for an AI-era workforce.

  • Conclusion: Student perceptions of the learning environment, abilities, and teaching strategies influence learning approaches and outcomes.Positive perceptions are associated with deep learning, whereas negative perceptions are associated with surface learning.
  • Conclusion: Understanding students’ willingness and concerns can help educators integrate GenAI so it complements and enhances traditional teaching methods.Perceiving GenAI as valuable and supportive may encourage students to adopt a deep approach to learning.
  • Implications: Institutions should provide educational resources and workshops to familiarize students with GenAI and its ethical and societal implications.Such preparation can support informed academic decisions when students use these technologies.
  • Implications: GenAI development and implementation should prioritize transparency, accuracy, and privacy to foster trust and mitigate risks.Explainable AI models and robust data protection policies are proposed measures for clearer decisions and safeguarded privacy.
  • Implications: Higher education institutions should rethink policies, curricula, and teaching approaches for a future in which GenAI technologies are prevalent.Suggested priorities include interdisciplinary learning, critical thinking, creativity, digital literacy, and AI ethics education.
  • Implications: The study calls for a balanced approach that addresses students’ concerns while maximizing GenAI’s potential benefits in higher education.This approach is intended to enhance teaching and learning outcomes while preparing students for the future workforce in the AI-era.

Limitations and Future Research

The study’s small sample, self-reported data, and cross-sectional design limit interpretation and generalizability. Future research should use broader and longitudinal approaches to examine GenAI integration, learning outcomes, student differences, and responsible implementation.

  • Limitations: The relatively small sample may limit generalizability to Hong Kong’s broader student population, while self-reported data may introduce social-desirability and recall biases.The study also identifies its cross-sectional design as a limitation, although the supplied passage is truncated before explaining its implications.
  • Future Research: Future research should employ larger, more diverse samples and longitudinal designs to track changing student perceptions and GenAI integration in higher education.Studies should also examine the relationship between GenAI use and learning outcomes.
  • Future Research: Future studies could compare students across disciplines, academic backgrounds, age groups, and cultural contexts to examine differing experiences with AI literacy.The passage specifies investigation of particular student groups, but its final wording is truncated.
  • Future Research: Further research should identify ways to integrate GenAI into higher education while minimizing privacy and security risks and supporting responsible, effective use.The intended contexts are teaching and learning.
Loading 2305.00290v1…