Digital Innovations in Education Brief N°. 3 100 Student Voices on AI and Education © 2024 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, links/footnotes and other information shown in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The citation of works authored by others does not mean the World Bank endorses the views expressed by those authors or the content of their works. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Please cite the work as follows: Cobo, C., Munoz-Najar, A., Bertrand, M. (2024). 100 Student Voices on AI and Education. In Digital Innovations in Education. World Bank. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202- 522-2625; e-mail: pubrights@worldbank.org. Acknowledgements This brief is a product of the Education Global Practice for Latin America and the Caribbean at the World Bank as part of the Digital Innovations in Education series. This brief was prepared by Cris- tobal Cobo (Senior Education Specialist), Alberto Munoz-Najar (Consultant), and Bertrand Momo (Education Specialist), under the guidance of Alex Twinomugisha (Senior Education Specialist), Juan Baron (Senior Economist), and Robert Hawkins (Senior Education Economist). This work was conducted with the support of the MasterCard Foundation. 2 100 Student Voices on AI and Education 100 Student Voices on AI and Education 100 Student Voices on AI and Education 3 Table of Contents Section A: Background........................................................................... 5 Executive Summary............................................................................................... 5 1. Context .............................................................................................................. 6 2. AI and the labor market.................................................................................... 8 3. Youth and AI....................................................................................................... 9 Section B: Youth voices on AI................................................................10 4. Study design and Methodology....................................................................10 5. Findings and discussion..................................................................................12 Section C: Reflections for Enhancing AI Integration in Education........18 6. Reflections on AI fluency.................................................................................18 7. Reflections for governments and policy makers..........................................20 8. Reflections for higher education institutions and faculty............................24 9. Final reflections................................................................................................28 References............................................................................................................31 Annex....................................................................................................................32 4 100 Student Voices on AI and Education Section A: Background Executive Summary The rapid advancement of artifi- tions, uses, and concerns based on cial intelligence (AI) is disrupting focus group discussions conduct- the landscape of higher education ed in 10 countries. The research (HE), presenting both opportunities revealed that while students regu- and challenges. This paper discuss- larly use AI tools for academic pur- es the results of focus group dis- poses such as writing, coding, and cussions conducted in 10 countries creative projects, barriers such as (Cameroon, Colombia, Ethiopia, high internet costs and low connec- Georgia, Indonesia, Mali, Mexico, tivity still persist in some regions. Nigeria, Peru, and Rwanda), exam- Additionally, students recognized ining students’ perspectives, expe- AI’s potential to enhance learning riences, and concerns regarding through personalized feedback AI’s impact on education. and accelerated skill acquisition. However, many learners voiced This report is composed of three concerns about how overdepen- main sections. The first section dence on the technology may stifle provides a general background on critical thinking. Their awareness the impact of AI in HE, the labor of emerging AI career paths like market, and Youth and AI. Through prompt engineering varied, but a a thorough review of emerging evi- common thread was acknowledg- dence, this section explores the piv- ing insufficient preparedness from otal role that higher education insti- current higher education curricula, tutions (HEIs) should play in training signaling the need for specialized AI talent, equipping students for an AI training. Across institutions, AI AI-driven workforce, and shaping integration displayed disciplinary research and policies around AI’s disparities, with Science, Technolo- societal impacts. AI offers avenues gy, Engineering, and Mathematics to augment teaching, learning, ad- (STEM) fields being early adopters. ministration, and decision-making To bridge this gap, students advo- through tools like AI-assisted grad- cated for comprehensive AI educa- ing, adaptive learning systems, and tion spanning all disciplines, robust automated data analysis. Howev- ethical frameworks, hands-on skill er, integrating AI also necessitates development opportunities, and re-examining academic programs, academic-industry partnerships to enhancing technology infrastruc- equip graduates with AI fluency for ture, tackling ethical risks around the future workforce. privacy and bias, and cultivating uniquely human skills like critical The third section provides reflec- thinking that AI cannot easily rep- tions and suggestions for enhanc- licate. Achieving AI-readiness re- ing AI preparedness and fluency in quires institutional changes like HE. To harness AI’s transformative upskilling educators, developing potential responsibly, governments governance frameworks, ensuring must champion safe, responsi- equitable access to AI resources, ble, and human-centered policies and fostering a culture of innova- alongside public awareness initia- tion to harness AI’s full potential re- tives, research funding, and AI inte- sponsibly. gration into education accreditation frameworks. HEIs should proactive- Section two of the report discusses ly rethink teaching approaches, key findings on students’ AI percep- 100 Student Voices on AI and Education 5 curricula relevance to evolving job nitions vary, three key attributes markets, administrative governance typically characterize AI systems: models, and equitable technology (i) They are machine-based sys- access – facilitating faculty AI train- tems; (ii) They can infer outputs ing, personalized AI-assisted learn- (predictions, content, or decisions) ing, ethical usage guidelines, and based on human objectives; (iii) digital infrastructure. Prioritizing These outputs are often indistin- ethics and inclusion are paramount, guishable from those of humans2. as HEIs and faculty leverage AI to enhance accessibility, identify out- come disparities, and empower di- The relationship between AI and verse stakeholders. higher education systems is nei- ther linear nor unidirectional. 1. CONTEXT On the one hand, AI systems are significantly shaping teaching and In recent years, the nexus be- learning experiences in HE. On the tween education and artificial in- other hand, HEIs could play a role in telligence has evolved rapidly. AI training AI-ready workforce, equip- innovations, especially generative ping learners for an AI dominated AI, are already transforming teach- world, and shaping research and ing and learning. As machines be- the policy discourse on the role of come “smarter” and labor markets AI in society (see Illustration 1). The and economies get transformed, field of AI emerged on a university education systems must rethink campus, when an American univer- their operating models to ensure sity professor organized the Dart- that they can skill, upskill, and reskill mouth Summer Research Project individuals for the jobs of the fu- on Artificial Intelligence in 1956. ture. HEIs are at the forefront of this As the technology becomes more change, given their “feeder” role for pervasive, universities and TVET labor markets. institutions will play a key role in skilling and reskilling individuals for While there is no single defini- AI-dominated or -influenced labor tion for the term “Artificial Intel- markets. Additionally, HE research- ligence” (AI), it generally rep- ers will shape both the technical resents data-intensive systems dimensions (development, testing, that can perform tasks that are design, efficacy, and accuracy) and typically associated with hu- policy discourse around artificial in- man intelligence1. While defi- telligence systems. 1 WIPO (2024). What is Artificial Intelligence? 2 EU AI Act (2024). Article 3: Definitions Augment instruction. Automate simple admin tasks. Personalized & adaptive learning. Reshape demand for skills. Higher Artificial Education Intelligence Systems Systems Skill and reskill workforce for AI. Conduct research on AI’s use/impact. Influence policy discourse on AI. Illustration 1: Relationship between AI and HE systems 6 100 Student Voices on AI and Education The potential of AI in HE is vast, ing skills that are uniquely human, offering avenues to augment the such as critical thinking, creativity, experiences of instructors, learn- and emotional intelligence. HE sys- ers, administrators, and policy tems must adapt their programs to makers. AI-assisted grading and cultivate these skills, ensuring that feedback mechanisms can alleviate graduates are well-equipped to the burden on instructors, freeing navigate an AI-driven workforce.3 up time for more meaningful inter- Achieving this will necessitate re- actions with students. Learners also visiting planning, instruction, and stand to benefit from AI-powered assessments to impart and evaluate adaptive learning systems that tailor these novel skills for the 21st centu- educational content and delivery ry. In an age where machines can methods to their individual needs, think, education systems shouldn’t ensuring a more effective and effi- just teach individuals to have the cient learning experience. The use right answers, but also to ask the of AI in education transcends indi- right questions. Moreover, as the vidual institutions. Ministries of ed- chart below indicates, interest in AI ucation are increasingly leveraging skilling is rising (see Illustration 2). the tool to support curriculum de- This represents a unique opportuni- velopment and nation-wide con- ty for HEIs, as they can help upskill tent generation at scale. working professionals through for- mal and informal training programs However, the advent of AI raises in domains such as machine learn- significant risks and challenges ing, deep learning, and other relat- that HEIs must address. The inte- ed areas. gration of AI in education not only necessitates a re-examination of ac- ademic programs but also the exist- ing capacity building methodolo- 3 de Bem Machado, A., dos Santos, J. R., gies. In addition, there is a need to Sacavém, A., & Sousa, M. J. (2024). Digital enhance the data and technology Transformations: Artificial Intelligence in Higher infrastructure of HEIs. As AI increas- Education. In Digital Transformation in Higher Education Institutions (pp. 1-23). Cham: Springer ingly automates routine tasks, there Nature Switzerland. is a growing emphasis on develop- Global searches for “AI Course” on Google (2019-2024) 150 100 50 0 Week 9/1/2019 1/19/2020 6/7/2020 10/25/2020 3/14/2021 8/1/2021 12/19/2021 5/8/2022 9/25/2022 2/12/2023 1/2/2023 11/19/2023 Search Interest Illustration 2: Interest in AI skilling Note: Between 2022 and 2023, searches for “AI Course” have increased showing interest in AI skilling. This represents an opportunity for HE to contribute to workforce development. Source: Google Search Trends 100 Student Voices on AI and Education 7 2. AI AND THE LABOR MARKET According to the IMF,4 around 40 potential displacement of workers in percent of global employment is some occupations, others consider estimated to be exposed to AI, that AI could also lead to more inclu- with a notable divide between ad- sion and stronger economic mobility vanced, emerging markets, and de- by improving education quality and veloping economies. In advanced access, expanding credit availability, economies, the share of employment and lowering skill barriers.5 The im- exposed to AI is around 60 percent, pact of AI on workers will vary de- reflecting the preponderance of cog- pending on their education level nitive-intensive occupations in these and age. Young college-educated economies. In contrast, the share of workers are considered the most employment exposed to AI is lower adaptable but also the most vulner- in emerging markets (around 40 per- able, as they may need to frequent- cent) and substantially lower in low-in- ly switch job types. Historical pat- come countries (around 26 percent), terns suggest that high-exposure, where the employment structure is high-complementarity roles may of- tilted more toward manual and rou- fer wage premiums, while switching tine tasks. The expected changes in to low-exposure roles could lead the labor market due to AI remain to decreased wages. The ability to ambiguous, while some research- adjust to AI-induced changes will ers consider that AI adoption might be crucial for navigating the labor include shifts in job types and tasks, market. 4 International Monetary Fund. (2024). Gen-AI: 5 Filippucci, F. et al. (2024). Should AI stay or Artificial Intelligence and the Future of Work should AI go: The promises and perils of AI for (IMF Staff Discussion Note No. SDN/2024/001). productivity and growth Washington, D.C.: International Monetary Fund. 8 100 Student Voices on AI and Education Table 1: AI benefits and concerns Dimension AI Benefits AI Risks & Concerns Teaching and . AI-powered labor market insights . AI-generated curricular content may be research can improve course design. biased, inaccurate, and/or low quality . AI can help develop instructional . Over-reliance on AI may inhibit teach- materials (lesson plans, presenta- ers’ or curriculum experts’ agency tions, etc.) . Excessive use of AI may de-humanize . Automated grading and feedback teaching experience relieve burden of instructors . AI-assessment tools may amplify biases . AI supports teacher professional in their data development . Teachers may fear being replaced by AI . AI research tools can support liter- ature review, data collection, and processing Learning . AI can support self-directed learn- . Over-dependence on AI may limit stu- ing (AI personal tutors, AI quiz gen- dents’ critical thinking erators, etc.) . Individualized AI learning solutions may . AI can provide personalized and undermine social learning, self-regu- instantaneous, actionable feedback lation, and the community aspects of learning experiences learning Equity and . Automatic captioning benefits . AI may perpetuate historical bias inclusion learners with disabilities . Digital divides may become AI-divides, . AI can support the creation or ad- and key groups like women may be left aptation of content into local or behind languages Governance and . Automate early warning systems . AI raises privacy concerns data . Speed up data processing and anal- . Bad actors may use AI to cause harm ysis for decision making . Lack of guidance and support may leave teachers and/or students in a weak position Skills and labor . AI may create new job opportunities . AI may disrupt labor markets faster market linkages for graduates of HEIs. than institutions can respond, creating a . AI solutions can support employ- threat of obsolescence. ability efforts within HEIs (AI inter- view practice, AI CV review, etc.) 3. YOUTH AND AI adds to existing research, and aims to amplify the voices of students While there have been studies from diverse cultural, socioeconom- exploring student experiences ic, and academic fields and back- with AI in education, our paper grounds, providing insights into offers a unique global lens by fo- the challenges, opportunities, and cusing on 10 emerging countries. ethical considerations surrounding Previous research such as the “Stu- AI’s integration into HEIs across the dents’ Perceptions of Artificial Intel- Global South. By capturing the per- ligence in Higher Education”6 and spectives of students in countries the “AI and the Future of Learning” like Cameroon, Colombia, Ethiopia, report by the OECD,7 have shared Georgia, Indonesia, Mali, Mexico, how students in developed nations Nigeria, Peru, and Rwanda, we offer perceive and interact with AI tools a comprehensive understanding of in academic settings. This report the nuanced experiences and con- cerns that shape the adoption of 6 Timea & Veres (2023). Students’ Perceptions of AI in educational contexts outside Artificial Intelligence in Higher Education 7 OECD (2023). AI and the Future of Skills, High Income Countries (HICs). Volume 2 100 Student Voices on AI and Education 9 Section B: Youth voices on AI 4. STUDY DESIGN AND research sought to identify poten- tial blind spots, biases, or inequities METHODOLOGY that could arise from unchecked AI proliferation. To better understand the role that AI will play in education, the RESEARCH APPROACH World Bank EdTech team under- took a research initiative explor- As this research aims to answer ing how HE students are engag- exploratory questions of stu- ing with AI. As university and TVET dents’ perceptions, uses, and students will enter the workforce concerns about AI in HEIs, a qual- and drive innovation in the coming itative methodology is best suit- years, they can offer unique per- ed for this context. A qualitative spectives about the impact of AI approach allows to understand the in education. By closely examining ‘meanings’ that participants attach their current uses of AI tools, the to actions and how these under- challenges and opportunities they standings influence their behavior. identify, and their ethical consid- A qualitative approach also pro- erations surrounding AI adoption, vides in-depth understanding of the this research initiative aimed to in- context in which the research is tak- form policymakers and educators ing place and how students’ prac- in crafting an AI ecosystem that em- tices look like in the specific con- powers rather than hinders the next text. More specifically, the research generation. team has developed a case study for each of the 10 countries where The research spanned 10 coun- data has been collected, as case tries, diverse socioeconomic con- studies are appropriate to address texts, different fields of study, ‘how’ problems and research ques- and gender balance, ensuring a tions that require in-depth analysis rich diversity of perspectives was to understand complex social phe- captured. Through focus group nomena.8 After obtaining approval discussions, the World Bank’s Ed- from the World Bank’s country offic- Tech Team delved into the specific es in the mentioned countries, the AI applications students employed research team supported by local for academic pursuits, creative en- staff identified one HEI per each deavors, and personal use. The country, ranging from public to pri- focus group discussions were con- vate universities and technical voca- ducted remotely using Microsoft tional education and training (TVET) Teams, facilitating conversations in institutions. Letters were emailed to three languages: English, French, the selected HEIs and 100 students and Spanish. Additionally, the re- agreed to take part in this study. All search team probed the perceived participants that agreed to partici- accessibility and inclusivity of AI pate in the study signed a consent solutions, concerns around privacy form. The list of institutions is pre- and data rights, and the potential sented in table 2. displacement of human roles by increasingly capable AI systems. By 8 Yin, R. K. (2014). Case study research: Design giving voice to the youth at the van- and methods (5th ed.). Thousand Oaks, CA: Sage. guard of technological change, this 10 100 Student Voices on AI and Education Table 2: List of participating countries and institutions # Country Region Institution Type Mgmt Students Catholic University 1 Cameroon Central Africa University Private 10 Institute of Buea Universidad de Los 2 Colombia South America University Private 7 Andes Addis Ababa 3 Ethiopia East Africa University Public 8 University Georgian Technical 4 Georgia Eastern Europe University Public 13 University Universitas 5 Indonesia South East Asia University Public 12 Pendidikan Indonesia University Of Science 6 Mali West Africa Of Technical And University Public 10 Technology Bamako Universidad 7 Mexico North America University Public 15 Tecnológica de León 8 Nigeria West Africa University of Lagos University Public 12 Servicio Nacional de 9 Peru South America Adiestramiento en TVET Public 13 Trabajo Industrial African Leadership 10 Rwanda East Africa University Private 11 University METHODS themes covered include current access and use of AI tools, ethical The research team decided to use concerns, how AI is changing the focus group discussions as the learning process, potential benefits main method for data collection, and challenges of AI in education, as it is an adequate qualitative re- anticipated impacts on university search method to capture percep- life, awareness of AI career paths, tions of students. To guide the fo- how educational institutions are cus group discussions and capture preparing students for AI disrup- students’ perceptions, the research tions, and expectations for AI’s fu- team developed a semi-structured ture influence on learning and the questionnaire. This questionnaire job market.9 aims to explore students’ percep- tions and experiences with AI in education. It begins with introduc- tory questions about their overall 9 A description of the data analysis process thoughts on AI’s impact. The main and limitations of this study are provided in the Annex. 100 Student Voices on AI and Education 11 5. FINDINGS AND ty education access. As an Ethiopi- an student remarked, “AI in gener- DISCUSSION: al [is] an opportunity because I’ve HOW DO STUDENTS ACCESS been able to make my work easier.” AND USE AI? In contrast, European perspectives like those from Georgia revealed Across the countries represent- stronger reservations, citing con- ed, students demonstrated wide- cerns over AI perpetuating biases, spread familiarity and utilization providing inaccurate information of AI tools for diverse academ- from outdated training data, and ic purposes. In Cameroon, tools potentially hindering human roles. like Gemini were praised for “its A common thread, however, was efficiency in full AI data analysis, recognizing AI’s power necessitat- offering quick response times.” ed ethical guidelines as its academ- Colombian students highlighted ic presence grows. ChatGPT’s utility in “comparing the results [they] get by solving problems [themselves] with the re- Accessibility issues shaped ex- sults of AI tools.” Nigerian students periences in some nations. While found ChatGPT and Quillbot ac- tools like ChatGPT were freely avail- cessible, with one noting they are able, others required paid subscrip- “the easiest and fastest ways to get tions, excluding financially-con- solutions.” Rwandan students even strained students, unless university mentioned using AI detector tools funds provided access. Unreliable like Undetectable.ai, underscoring internet connectivity also ham- awareness beyond conventional pered adoption in parts of Cam- applications. However, accounting eroon and Rwanda. As a Georgian students in Cameroon expressed student highlighted some AI tools reservations about AI’s reliability in “required payment, potentially their field. hindering access for students with financial constraints.” Such dispar- ities underscore the importance of Regional differences emerged addressing digital divides. around perceptions of AI’s im- pact in education. Students in de- veloping nations tended to view AI Ethical concerns cut across regions, optimistically as “an opportunity” prompting calls for responsible AI and “potential equalizer” for quali- utilization. Art students in Indonesia felt threatened by AI generating full creative works, with one stating: “I condemn people who use AI 100% and then claim it as their job.” Mexi- can students warned about AI foster- ing dependency, providing misinfor- mation, and displacing human work- ers. However, many also recognized AI’s potential for streamlining work and empowering people, leading a Rwandan student to describe it as a “double-edged sword” whose im- Students in developing pact “depends on how responsibly we choose to utilize it.” This nuanced nations tended to view perspective captures the delicate AI optimistically as balance societies must strike in har- “an opportunity” and nessing AI’s benefits while upholding ethical principles and human agency. “potential equalizer” for quality education access. 12 100 Student Voices on AI and Education HOW AI IS CHANGING THE WAY sonalized guidance and “increasing STUDENTS LEARN? productivity by streamlining access to data and resources.” From saving Across the countries represented, time to sparking innovation, stu- students described AI enabling dents across regions recognized new modes of learning and skill AI’s transformative potential in aug- development. Participants high- menting learning experiences. lighted AI’s value in overcoming accessibility barriers, with an Ethi- However, ethical concerns sur- opian economics student sharing: rounding AI integration also sur- “I use AI to summarize and make faced globally. Cameroonian stu- articles more understandable from dents expressed “skepticism about paywalled and complex resources relying solely on AI for answers,” like the Harvard Business Review.” fearing impacts on critical thinking. Nigerian history majors leveraged An Indonesian architecture student AI for diverse perspectives beyond found “AI large language models ... traditional texts - “Different AI bots not always adequately knowledge- give you different perspectives on able” for specialized tools. Geor- the same historical facts.” For In- gian participants worried about AI’s donesia STEM students, AI trans- limitations “in tasks like brainstorm- formed coding with real-time de- ing, where it often yielded repeti- bugging assistance: “AI has 100% tive suggestions.” In Nigeria, some changed how we learn compared worried about “the temptation to to before advanced language mod- become overly dependent on AI for els.” Mexican programmers echoed generating academic content rath- using AI “as supplementary sup- er than thinking critically.” Peruvi- port when stuck on coding issues.” an TVET students cautioned about Meanwhile, Rwandan students not- “accepting AI-generated responses ed AI accelerating their learning unquestioningly” without cross-ref- curves, with one able to rapidly erencing other sources. However, launch a website without “months the notion that students exclusively to learn the necessary skills.” How- use AI tools for cheating purposes ever, this shift necessitated devel- is incomplete. As previously dis- oping new prompt engineering cussed, students are using AI for a abilities to harness AI effectively. range of academic tasks, including writing essays, coding, research, Students identified multiple generating ideas, improving con- potential benefits to further in- tent quality, text recognition, read- corporating AI into education. ing assistance, and even creative Across nations like Cameroon and projects like image generation and Colombia, many saw AI’s efficien- podcast editing. Across countries, cy providing more leisure time to students repeatedly emphasized improve mental well-being. Mexi- the need for human diligence to can participants appreciated how fact-check AI and not over-rely on it. AI “drastically reduce[s] research time by consolidating information.” Ethiopian engineers mentioned AI’s Looking 5 years into the future, value in “obtaining specific insights students envisioned AI signifi- by ... processing large data sets.” cantly reshaping university life For Nigerian economics majors, AI and learning paradigms. Partici- simplified complex theories - they pants’ perspectives highlight AI’s could prompt ChatGPT to “explain potential to democratize access to this concept to a 16-year-old.” Pe- knowledge while also raising con- ruvian video game designers cited cerns about exacerbating socio- AI aiding creativity in “creating char- economic inequalities. For instance, acters.” Looking ahead, Rwandan Ethiopian participants anticipated students foresaw AI enabling per- AI democratizing “accessibility to 100 Student Voices on AI and Education 13 information and knowledge re- ship in this field. Despite regional sources that were previously diffi- exposure differences, common cult to obtain.” However, they also threads emerged – not all students warned about risks of “exacerbat- were familiar with AI’s proliferation ing socioeconomic inequalities if across diverse career domains and digital disparities persist.” In Geor- there was a shared aspiration to gia, students saw AI streamlining equip themselves for the evolving administrative education tasks like landscape of AI-driven professions. “grading and data processing, free- ing up time for educators.” Rwan- When it came to readiness for pur- dan students expected AI to “dis- suing AI careers, students voiced rupt the traditional school system a mix of enthusiasm and concern. by revolutionizing how education On the one hand, participants ex- is accessed and delivered.” From hibited keen interest and motiva- equitable access to overhauling tion to develop the skills needed pedagogies, regional perspectives to thrive in an AI-driven workforce. converged on AI poised to reshape On the other hand, they acknowl- academia - prompting calls for pro- edged gaps in their current prepa- active, ethical implementation strat- ration and curricula that left them egies to harness AI’s transformative feeling ill-equipped for such roles. potential while mitigating risks and Many Ethiopian participants were disparities. eager about “prompt engineering” roles but admitted “not feeling fully HOW AI IS SHAPING STUDENTS’ prepared” yet. Rwandan students CAREER PROSPECTS? struggled with “high barriers to entry such as extensive experience Across the countries represented, and advanced degrees” for AI jobs. students demonstrated varying An Indonesian art student was open levels of awareness regarding AI to AI-driven creative careers “giv- career pathways. Many students en market demand,” though con- demonstrated an awareness of cerned about “imperfect nature of emerging AI roles, such as prompt current AI image outputs.” In Cam- engineering, AI data analysis, and eroon, some felt ill-equipped due AI entrepreneurship. For instance, to “lack of focused curricula cov- in Colombia, while some exhibit- ering computational thinking [and] ed limited knowledge, others ex- algorithm design.” Across countries pressed curiosity about roles like like Mexico and Nigeria, the prevail- “training AI tools to describe im- ing view was that additional spe- ages.” Georgian participants cited cialized training would be required startups searching for AI for roles to confidently transition into cut- like junior AI developer. Indone- ting-edge AI professions. However, sian, Ethiopian, and Nigerian stu- students displayed an overarching dents mentioned emerging fields willingness to upskill and be “life- like prompt engineering that allows long learners to thrive in AI-driven harnessing the power of AI for craft- economies”, signaling their aspi- ing effective prompts. An Indone- ration to bridge the preparedness sian journalism student highlight- gap through continuous learning. ed “the potential for utilizing AI to efficiently generate text from news The potential threat of AI dis- reports.” Entrepreneurial prospects rupting traditional career paths offered by AI tools were also rec- emerged as a significant concern ognized, with a law student in Co- for students across regions. Across lombia citing a professor’s creation countries, students expressed con- of an AI tool to streamline legal cern about the potential of AI to processes and in Georgia students disrupt their career paths. Students mentioned startup accelerators and in IT fields such as software devel- hackathons fostering the develop- opment, data science, and robotics ment of AI skills and entrepreneur- worried about losing job oppor- 14 100 Student Voices on AI and Education tunities to AI as advanced models take a more systematic approach increasingly automate computing to robustly equip graduates with AI and robotic tasks. In Ethiopia, soft- literacy for emerging opportunities. ware developers worried about “losing job opportunities to AI” as DO STUDENTS THINK advanced models increasingly au- EDUCATION SYSTEMS AND tomate coding tasks. Georgian data INSTITUTIONS ARE READY? science students similarly feared AI Institutions across the repre- swiftly processing datasets could sented countries exhibit varying “displace” analysts. Peruvian stu- levels of readiness to address dents raised worries regarding “ro- AI disruptions within education. botics and AI integration displacing Concerns raised by students high- human workers” as AI augments light disparities between technical efficiency across sectors like man- and non-technical faculties in inte- ufacturing. However, some Indo- grating AI concepts into curricula. nesians in human-centric fields like Engineering and IT faculties are psychology believed AI currently often more proactive, incorporat- lacks the emotional intelligence ing AI-related courses and practical required - “AI has no feelings...but skills training. However, non-tech- in my field we use emotions for nical fields lag behind, indicating counseling.” An Ethiopian junior a need for comprehensive AI edu- developer mentioned that if AI is cation across all disciplines. Ethio- integrated thoughtfully in differ- pia, students lamented the lack of ent fields, markets could avoid hu- AI education in universities, with man displacement. These concerns one student stating, “We don’t raised by students regarding job have a single department offering displacement due to AI automation AI courses; our curriculum needs a underscore the need for thoughtful major update.” The importance of integration and proactive strategies faculty training emerges as a critical to mitigate potential impacts while factor, with calls for institutions to harnessing the transformative po- invest in programs that familiarize tential of AI technologies across di- educators with AI tools and meth- verse professional domains. odologies. Additionally, students Students’ experiences with uni- stress the necessity of updating versity efforts to prepare them curricula to include foundational for AI-driven professions varied AI courses and ethical frameworks significantly by institution and discipline. In Colombia, partici- pants reported institutional work- shops promoting “responsible AI use” while others faced limited ac- cess. In Ethiopia, students lamented their universities providing “inade- quate” AI career guidance - “I think it’s not enough and they’re not pro- viding us with enough knowledge.” Indonesian computer science de- Concerns raised by partments offered substantial AI students highlight career exposure through “seminars and industry events,” contrasting disparities between psychology where information was technical and non- lacking. While ALU in Rwanda intro- technical faculties duced targeted AI courses, other nations saw only pockets of AI inte- in integrating gration by forward-thinking faculty, AI concepts into creating inconsistencies. Overall, students advocated universities curricula 100 Student Voices on AI and Education 15 that guide responsible AI adoption uniform training for all faculty mem- within academia. Policies and pro- bers.” Training initiatives tailored to tocols governing AI use in educa- instructors’ needs are recommend- tion are deemed essential to navi- ed to enhance their proficiency in gate the transformative impact of AI technologies and foster a culture AI on traditional teaching methods of innovation within academic envi- effectively. ronments. The preparedness of professors Students offer insightful reflec- and lecturers to integrate AI tech- tions for institutions to enhance nologies into teaching method- readiness for AI disruptions. They ologies varies notably, often re- advocate for comprehensive AI ed- flecting generational differences. ucation, emphasizing the need for Younger professors tend to be more foundational AI courses and practi- open to adopting AI tools com- cal skills training across disciplines. pared to their older counterparts, In Indonesia, students advocated who may exhibit resistance or lim- for increased AI course offerings ited knowledge of AI capabilities. and workshops, emphasizing prac- In Georgia, a student mentioned, tical skills development. A student “Younger professors are experi- mentioned, “We need more AI ex- menting with AI assistants, while perts … and frequent seminars to older ones prefer traditional meth- keep pace with AI advancements.” ods.” Teaching methods also differ, Additionally, students stress the im- with progressive educators leverag- portance of fostering a culture of ing AI for assignments and projects openness to innovation, encourag- while others maintain conventional ing collaboration between faculty approaches. The challenge lies in and AI experts, and establishing bridging this divide and ensuring clear policies on AI use in educa- consistent guidance on responsible tion. In Rwanda, students stressed AI deployment across all academic the importance of developing AI programs. In Mexico, students em- policies and guidelines, as one stu- phasized the need for consistent dent suggested, “Institutions should guidance on responsible AI de- collaborate with ministries to estab- ployment, with one student stating, lish ethical AI frameworks and edu- “Some instructors embrace AI, but cate students about responsible AI others lack knowledge; we need usage.” Students also highlight the importance of experiential learning opportunities and partnerships with industry stakeholders to prepare students for AI-driven career paths. Moreover, students from Georgia called for universities to promote responsible AI usage among stu- dents, emphasizing the need for human oversight to correct imper- The challenge lies fect or “hallucinated” outputs gen- in bridging this erated by AI tools. Overall, pro- active measures such as compre- divide and ensuring hensive AI education and policy consistent guidance development are recommended to on responsible AI ensure institutions are equipped to navigate the challenges and oppor- deployment across all tunities presented by AI disruptions academic programs. in education. 16 100 Student Voices on AI and Education The results from this cross-country study provide valuable insights into how AI is shaping the HE landscape and preparing students for future careers. Here is a compendium of 10 main remarks: 1. Widespread AI adoption: Students across regions demonstrated familiarity and utilization of AI tools for various academic purposes, including writing, research, analysis, and creative projects. 2. Accessibility challenges: While some AI tools were freely available, others required paid subscriptions or reliable internet connectivi- ty, creating accessibility barriers for students in developing nations and those with financial constraints. 3. AI as an educational equalizer: Students viewed AI optimistically as an opportunity for democratizing access to quality education and resources, but some students also expressed concerns about po- tential biases and inaccuracies. 4. Transforming learning experiences: AI enabled new modes of learn- ing, from real-time coding assistance, personalized guidance, and accelerated skill development. However, ethical concerns about over-reliance and critical thinking erosion were also raised. 5. Need for human oversight and fact-checking: While recognizing AI’s capabilities, students across regions emphasized the importance of not blindly accepting AI-generated outputs as fact. They stressed the necessity of human diligence in fact-checking AI responses and not over-relying on AI at the expense of critical thinking skills. 6. Career prospects and readiness: Students exhibited varying levels of awareness about emerging AI careers like prompt engineering and AI entrepreneurship. While eager to develop relevant skills, many felt ill-equipped due to limited curricula and high barriers to entry. 7. Job displacement concerns: Students across IT, data science, and manufacturing fields feared AI automation could displace human workers, underscoring the need for thoughtful integration strate- gies. 8. Institutional readiness disparities: Engineering and IT faculties were more proactive in integrating AI education, while non-technical fields lagged behind. Consistent guidance, faculty training, com- prehensive AI education across disciplines, industry partnerships, and clear frameworks on responsible AI usage were deemed nec- essary. 9. Faculty adoption divide: Younger professors were more open to adopting AI tools in teaching, while older faculty members exhibit- ed resistance or limited knowledge, highlighting the need for com- prehensive training initiatives. 10. Envisioning AI’s future impact: Looking ahead five years, students anticipated AI reshaping university life. Proactive implementation strategies were seen as crucial to harness AI’s transformative poten- tial while mitigating risks and disparities. 100 Student Voices on AI and Education 17 Section C: Reflections Governments, policy makers, HEIs, tems. Developing this multidimen- and faculty have a central role in sional skill set is pivotal to harness shaping how AI is integrated into generative AI’s potential in aug- education systems, ensuring this menting learning while mitigating process is done responsibly and associated risks of bias, privacy vi- fairly for everyone involved. olations, and intellectual property infringement.12 6. REFLECTIONS ON AI fluency entails the knowledge AI FLUENCY and skills needed for individuals to critically comprehend, utilize, NAVIGATING THE INTEGRATION and assess AI systems within an OF GENERATIVE AI IN HE increasingly digital environment. The rapid advancement of new It comprises three core competen- technologies, especially genera- cy areas: (a). Understanding: This tive AI, often outpaces the devel- area involves technical comprehen- opment of frameworks and insti- sion of AI systems, including skills tutional guidance. Education insti- such as data utilization, automa- tutions, regulators, and multilateral tion, algorithmic comprehension, organizations10 are progressing at and pattern recognition. (b). Use: different rates in their endeavors to This competency area focuses on adopt and adapt to new AI frame- effectively utilizing AI to enhance works that define the necessary or execute diverse tasks and work- skills and capabilities for HE stu- flows. Competency to create digital dents and educators. It is likely that content might become less relevant these frameworks and guidance in times of readily available digital will evolve over time.11 content developed by GenAI tools. (c). Contributions and Evaluation: As artificial intelligence (AI) pro- This area entails assessing and con- liferates across sectors, its pro- tributing to improving AI systems found implications for education across various dimensions, such as necessitate a concerted effort to data privacy, ethics, bias, credibility, foster “AI fluency” among learn- accessibility, and societal impacts. ers, workers, and communities Competence in this area enables at large. AI fluency encompasses the scrutiny of AI inputs, methodol- the robust competencies required ogies, outcomes, and identification to critically comprehend, ethically of potential risks or limitations. See assess, and judiciously apply AI sys- table 3 for AI fluency competencies. 10 Mills, K., Ruiz, P., & Lee, K.-w. (February 21, 2024). Revealing an AI Literacy Framework for Learners and Educators. Digital Promise logo. [Blog post]. 12 Gimpel, H., Gutheil, N., Mayer, V., Bandtel, M., 11 Bekiaridis, G., & Attwell, G. (2024). Supplement Büttgen, M., Decker, S., & Urbach, N. (2024). to the DigCompEDU Framework: Introduction to (Generative) AI Competencies for Future-Proof AI in Education1. Active Citizens Partnership & Graduates: Inspiration for Higher Education Pontydysgu. Institutions. Hohenheim Discussion Papers in Business, Economics and Social Sciences. 18 100 Student Voices on AI and Education Table 3: AI fluency competencies Competencies Dimension Evaluate and Understand Use Contribute Apply AI tools AI tools and Foundational Create, analyze or evaluate and systems to systems understanding of AI AI systems problem-solve Understand of AI risks and Incorporate AI Support responsible and AI ethics and societal impacts, including ethics in daily work safe AI, including safety by society human agency and life design Understand AI’s impact Research AI trends and AI, careers, and Use AI for on careers and lifelong shape education and work learning professional growth learning with AI Note: Partially Adapted from UNESCO’s Draft AI Competency Framework and Digicom 2.2 Creating the conditions for the ef- plications across disciplines. Guide- fective development and promo- lines must be developed to address tion of AI fluency will necessitate ethical considerations, data privacy, several institutional changes, such and the responsible use of AI tech- as revisiting and expanding career nologies within academic settings. plans, upskilling HE staff, adopting Moreover, fostering a culture of governance frameworks, ensuring innovation is crucial to harness the access to critical infrastructure, and potential of AI in research, teaching, allocating resources. Stakeholders, and administration. An AI-ready in- ranging from policymakers to edu- stitution is proactive in integrating cators, must implement measures AI into its curriculum, research ini- to foster beneficial deployments tiatives, and administrative process- while mitigating risks. Failure to do es, ultimately preparing students so could exacerbate existing dis- and faculty to thrive. parities, creating a dichotomy be- REFLECTION QUESTIONS ON STU- tween those benefiting from gener- DENT AND TEACHER AI FLUENCY ative AI’s augmentative capabilities and those further marginalized by . How will the ability to learn the disruption13. At the same time, programming or writing achieving AI readiness at HEIs en- evolve or change with the ad- vent of AI? compasses more than just equip- ping students with AI skills and flu- . How should we expand and ency; it also involves ensuring that rethink the definition of data the institution is prepared to lever- literacy skills in an era where age AI effectively. Faculty training AI can quickly process and plays a pivotal role in this readiness, analyze large datasets? Will the focus shift from basic data as educators need to grasp the manipulation to advanced in- technical aspects of AI and its im- terpretation of AI-generated insights? 13 Southworth, J., et al. (2023). Developing a model for AI Across the curriculum: Transform- . How can we refine our under- ing the higher education landscape via inno- standing and recognition of vation in AI literacy. Computers and Education: Artificial Intelligence, 4, 100127. content and outputs to discern 100 Student Voices on AI and Education 19 human versus machine knowl- lic goods on AI, and implementing edge production? other system-level actions within . How can we effectively distin- the education sector. guish between tasks and roles The safe, responsible, and equi- that can be easily automated table use of AI hinges upon clear or displaced and those that require intensive human in- and fit-for-purpose policies, reg- volvement? ulation, and guidance by gov- ernments. Countries have opted . As AI systems become more for different approaches to regu- proficient at information re- lating AI, with risk-based and prin- trieval and synthesis, how will ciples-based being the two most research skills be taught? Will the focus shift from finding in- common approaches (See illustra- formation to critically evaluat- tion 3). Risk-based approaches fo- ing AI-generated content and cus on categorizing AI systems and crafting precise queries? their uses based on possible risks to individuals, institutions, and com- 7. REFLECTIONS FOR munities. For instance, the EU’s AI Act, considered the foremost leg- GOVERNMENTS AND islation on AI, categorizes AI risks POLICY MAKERS based on four levels: unaccept- able risk, high risk, limited risk, and AI is a pressing issue for govern- minimal risk.15 On the other hand, ments’ skills agendas. As indicat- principles-based approaches such ed in the UN AI Act, member states as the UK’s AI framework, set core should encourage safe, secure, and principles that should guide the de- trustworthy AI systems in an “inclu- velopment, deployment, and use of sive and equitable manner, and for AI.16 the benefit of all.”14 Governments play a particularly important role at Illustration 3: Common approaches the nexus of AI education, includ- ing policy making, developing pub- 15 European Parliament (2024). Artificial Intelli- gence Act 14 United Nations (2024). Seizing the opportuni- 16 UK Department for Science, Innovation & ties of safe, secure and trustworthy artificial intel- Technology (2024). A pro-innovation approach ligence systems for sustainable development. to AI regulation: government response. Common approaches to regulating AI Risk-Based Approach Principles-Based Approach Regulating AI applications based on Focusing on broad ethical principles the level of risk they pose. such as fairness and equity. Example: Example: EU AI Act. UK AI Framework. Illustration 3: Common approaches to regulating AI 20 100 Student Voices on AI and Education While there are different ap- proaches to AI policymaking for stitutions, governments can education, human centeredness issue directives to underpin should be a common thread. Poli- safe AI adoption and use. Ad- cies should integrate and prioritize ditionally, government quality the views, opinions, interests, and assurance mechanisms such as accreditation bodies, can also safety of teachers, students, and the integrate AI-related metrics in public. Importantly, beyond AI-spe- their monitoring process and cific policies, governments should procedures. ensure that other related policies are reviewed or revised to consid- The rapid advancement of AI dis- er new technologies. For instance, ruption requires thorough discus- existing policies on disability inclu- sions between governments and sion, gender equality, and even na- HEIs to understand and take ac- tional curricula frameworks should tion on AI’s potential and implica- all be reviewed in light of AI and its tions. The table below provides AI implications for education and soci- government actions around five key ety. areas: teaching and assessment, learning, equity and inclusion, data Beyond setting regulations, gov- and governance, and labor market ernment programs, projects, and linkages. The table proposes quick other-directed interventions can wins, as well as long-term actions support the ethical mainstream- that may take more time. While rec- ing of AI in education. These can ommendations are presented as include the following: distinct actions, it is important to . Public goods, such as infor- note that AI in education requires mation websites, to gen- a systemic and holistic approach. erate awareness on AI: For They must be driven by strong vi- instance, the government of sion to support the responsible Singapore has set up ‘Learn use of AI with humans at the core. with AI’, a public awareness This, in turn, will require dedicat- program to raise awareness of ed resources: budget, clear roles AI among teachers, students, for teams at national and regional and even parents. levels, dedicated time for existing . Government education in- teams, performance targets (KPIs), stitutions and organizations: and constant monitoring and eval- Public HEIs enroll about two uation. thirds of tertiary students globally. In these public in- Table 4: Government 100 Student Voices on AI and Education 21 actions on AI Dimension Quick Wins Long-term Actions Teaching and as- Conduct a consultation of teachers Offer national and regional training on sessment to assess AI use and awareness AI and education for educators Develop cross-curricular capacities Allocate funding and resources to AI beyond specific disciplines research and development (R&D) Learning Organize a national essay or video Implement national program on AI for contest on AI and education for learning students and teachers Equity and inclusion Consult key groups, (including per- Ensure that AI policy emphasizes data sons with disabilities, women, etc.), privacy and security on use, access and perceptions of AI in education Governance and Set AI KPIs in governance and Integrate AI tools in governance and data management management (e.g. early warning sys- tems) Allocate funds to open research institutes and observatories for evi- dence-based decision-making Skills and labor Leverage AI tools for skills map- Conduct analysis of AI’s impact on jobs market linkages ping and to update skills frame- nationally, and reflect major changes in works education curricula and policy 5 COMMON MISTAKES TO AVOID addressing the ethical challeng- WHEN TACKLING AI IN EDUCA- es of AI within their institutions TION or departments. Likewise, inno- vation and technology related . Not investing in a culture KPIs could be included in per- of agility: The right culture formance management mecha- is essential to support both nisms within ministries. safe access and safe use of AI in education systems. Hence, . Not connecting AI to oth- equipping students to thrive in er education priorities and in an AI-enabled world requires initiatives: Within education governments and ministries of systems, the use of AI doesn’t education to promote experi- happen in a vacuum. To galva- mentation, speed up processes, nize momentum, it is critical to and take calculated and reason- connect AI policy and actions able risks. Education systems to broader issues in HE. For are not always the most agile – instance, governments may but the speed of evolution of AI explore how AI can contrib- requires rapid actions. ute to reducing dropout rates through early warning systems; . Ignoring incentives: To sup- or connect AI access to existing port the safe and responsible digitalization or internet access adoption and use of AI, gov- programs/projects. ernments must provide the right incentives to education . Outsourcing all AI interven- systems. For instance, ministry tions to Information Technol- or public university officials can ogy (IT) departments: While AI be rewarded or recognized for is certainly a computer science implementing AI guidelines or topic, AI in education is not 22 100 Student Voices on AI and Education just an IT issue. Therefore, the AI Governments and agenda should not be the sole responsibility of IT or computer ministries of education departments within ministries of should engage in a education. A whole-of-ministry ap- balanced approach proach is needed to truly address the risks and reap the benefits of that encourages AI in education. experimentation but . Extreme responses: Following also establishes clear the release of generative AI tools standards such as ChatGPT, several insti- tutions took extreme decisions: some school systems outrightly banned the tool, while others fully integrated the tool with limited guardrails.17 Governments and ministries of education should en- gage in a balanced approach that encourages experimentation but also establishes clear standards and guidelines to mitigate adverse consequences of AI on students, teachers, and school systems. Castillo, E. (2023). These Schools and Colleges 17 Have Banned Chat GPT and Similar AI Tools. Table 5: Reflection questions for governments and policymakers Dimension Reflection questions for governments and policymakers Teaching and assess- How can the government integrate AI-related competencies into teacher ment training programs and professional development initiatives? How can the government involve educators in developing policies and guide- lines for the appropriate use of AI in teaching and assessment? Learning What government initiatives can be undertaken to promote the responsible use of AI in learning among students? Equity and inclusion How can the government ensure that the integration of AI in education does not exacerbate existing inequalities? Are key education stakeholders aware of AI use and its risks and potential? Do they integrate AI in education planning, delivery, and assessment? Governance and While waiting to develop or update an AI policy, what quick actions can our data government take to raise awareness and provide guidelines for safe AI use? Skills and labor mar- How can we engage the private sector, including startups and firms, to co-de- ket linkages velop, iterate or customize AI for our local realities? What strategies can be implemented to align HE curricula with the skills and competencies required for AI-driven professions? 100 Student Voices on AI and Education 23 8. REFLECTIONS FOR versities’ AI-readiness and actions. While many universities provided HEIS AND FACULTY support and guidance on AI use, HEIs and faculty can play a vital others were mute on the subject, role in preparing for an AI-en- creating a void that left professors abled world, but for that, they to themselves, deciding often ar- too will have to evolve. The risks bitrarily on whether to use AI. HEIs linked to inaction are significant. should define clear guidelines and Universities and TVET institutions priority actions to support the safe that adapt to AI will be more like- and responsible use of AI in edu- ly to attract students, stay relevant, cation. The most appropriate AI ac- and grow. Those who fail to change tions for universities should depend may become less relevant and po- on the current level of AI maturity, tentially face disruption. As the as demonstrated in the decision findings above have demonstrated, tree below. there is great heterogeneity in uni- Are faculty skilled to teach AI concepts and manage AI implications? Implement faculty No Yes Are there development guidelines on the programs on AI ethical use of AI? Develop AI guidelines No Yes Is AI fluency prioritizing ethics and integrated across safety all disciplines? No Illustration 4: AI readiness decision tree for HEIs Mainstream AI fluency across disciplines 24 100 Student Voices on AI and Education HEIs should rethink teaching and emotional skills such as critical think- assessment in the light of AI. HEIs ing, communication, collaboration, should consider the use of AI tools and problem-solving. Importantly, for teaching and provide faculty universities should provide cours- development opportunities in all es on AI-related fields and support dimensions of teaching, from les- students to join these emerging son planning to assessment. Addi- areas. Additionally, universities can tionally, protocols should be imple- also generate revenues by support- mented to identify and prevent bias ing workforce development efforts in automated assessment tools. of firms and industry. Different faculties may have varying levels of AI aptitudes and use – it HEIs should rethink governance is critical to conduct consultations/ and data for the age of AI. HEIs audits to identify current levels of AI should urgently establish clear use and aptitude, identify concerns, guidelines for staff, students, and and create appropriate strategies. the university community on the HEI’s can leverage peer learning as use of AI in education. While guide- a powerful tool to encourage the lines would depend on a specific safe adoption of AI. Through col- context of HEIs, the following key laborative sessions among faculty, components should be considered: early adopters can share concerns, (i) Ethics and responsible use; (ii) views, and ideas based on their own Data privacy and security; (iii) Aca- experience. Faculty are the back- demic integrity; (iv) Inclusion; (v) AI bone of university systems. Faculty and curriculum development. In ad- training can encompass several di- dition to policy, universities should mensions including: teaching with explore the use of AI to support or AI, teaching for an AI world, and automate administrative tasks. For contributing to AI research. example, AI solutions can automate early warning systems and identify AI requires novel approaches students at risk of dropping out. to learning. As the FGD findings demonstrated, AI is already reshap- ing students’ approaches to self-di- rected learning. In this context, HEIs should explore the use of AI to sup- port personalized, adaptive, and interactive learning for students. HEIs can acquire, develop, or tailor AI tutors to guide students through concepts. Beyond using AI to learn, universities have a critical role to play in instilling the notion of life- long learning in students. HEIs and faculty should urgently revisit labor market linkages in HEIs should consider the light of AI. Universities should the use of AI tools for monitor employment trends and teaching and provide reflect these in program develop- faculty development ment, curricula, and assessments. This process requires investing in opportunities in key transferable skills and socio- all dimensions of teaching 100 Student Voices on AI and Education 25 Infrastructure access for AI should Ethics and inclusion implications be prioritized. There can be no re- should be at the top of the AI sponsible adoption of AI without agenda. Universities should ensure robust infrastructure. Universities AI benefits all students equitably. should prioritize equitable access Universities should also leverage AI to internet and devices for both to support inclusion and non-dis- staff and students. Universities crimination. For instance, AI-pow- should implement strong cyber- ered solutions can automate tran- security measures to protect sen- scription and conversion of text to sitive data and prevent unautho- speech for learners with disabilities. rized access to AI systems. Further- Additionally, AI can empower di- more, implementing a “Sovereign verse faculty, for instance, empow- AI” policy, as defined by Nvidia’s ering lecturers with disabilities to CEO Jensen Huang, where each create content, conduct instruction, country owns and controls the and assess students’ work. AI can production of its own AI infra- also conduct analyses to identify structure, would allow countries disparities in access and outcomes to protect their cultural values and and inform remedial actions. Exist- ensure data sovereignty.18 ing gaps in access to devices or in- ternet necessary for AI use should be identified and addressed. 18 Edwards, B. (2024). Nvidia CEO calls for “Sovereign AI” as his firm overtakes Amazon in market value. Table 4: AI interventions to consider for HEIs Dimension Examples of interventions Teaching Equip faculty to teach with AI, teach for an AI world, contribute to AI education, and foster the lifelong learning. Experiment with various tools for instruction and assessment and implement guardrails for inclusive and non-discriminative use. Learning Experiment with the use of AI for personalized and interactive learning. Inclusion and ethics Implement measures to ensure fairness, non-discrimination, and inclusive use of AI in education. Ensure that AI solutions are explainable and transparent. Governance and man- Develop AI guidelines for the safe and responsible use of AI. agement Explore the use of AI to automate administrative tasks. Skills and labor mar- Strengthen linkages with industry to support AI skills, enhance ket linkages teaching relevance, and develop new models of work-based learn- ing Help upskill and reskill workers for an AI world. Re-imagine career services for the age of AI and ensure that stu- dents develop socio-emotional skills. 26 100 Student Voices on AI and Education The integration of generative AI REFLECTION QUESTIONS FOR in HEIs offers a unique oppor- HEIS tunity to bridge disciplines and . How should HE approaches foster innovation. Rather than ho- evolve in an age where ma- mogenizing academic fields, this chines think? integration requires a balanced ap- . How might faculty foster crit- proach that leverages the strengths ical thinking and indepen- of diverse academic disciplines. dence in the age of AI? Humanities and social sciences can . What strategies can HEIs im- provide critical perspectives on the plement to ensure equitable ethical and societal implications of access to AI tools for all stu- AI, while STEM fields can drive tech- dents and faculty? . How should HEIs redesign cur- nical advancements. This interdisci- ricula and assessments to ac- plinary synergy not only enhances count for AI’s impact on learn- the academic landscape but also ing and skill development? prepares students for an evolving . How can institutions leverage job market where AI literacy and AI to improve personalized human-centric skills are increasing- and adaptive learning experi- ly valued. As institutions adapt to ences without compromising this technological shift, maintaining human interaction? disciplinary diversity will be key to developing well-rounded grad- uates capable of navigating and shaping the AI-augmented future of work and society. 100 Student Voices on AI and Education 27 9. FINAL REFLECTIONS The report has highlighted key de- icies to ensure that AI integration velopments in artificial intelligence aligns with educational goals and that necessitate immediate action ethical standards. and response. Given the rapid de- velopment of digital technologies, Governments and policymakers it is vital for HE systems to monitor stand at the forefront of shaping AI emerging themes and trends the responsible integration of AI frontiers closely. in education. Their role is crucial in addressing the disparities in AI ac- The integration of AI in HE pres- cess and readiness observed across ents a complex landscape of op- different regions. As a Georgian stu- portunities and challenges that dent remarked that some AI tools demand thoughtful consider- “required payment, potentially hin- ation and proactive strategies. As dering access for students with fi- AI tools become increasingly prev- nancial constraints.” Policymakers alent in academic settings, students must work to ensure equitable ac- are already experiencing their im- cess to AI resources and align HE pact. As a Colombian student not- curricula with evolving labor market ed, they use AI for “comparing the needs. This is particularly important results [they] get by solving prob- given the concerns raised by stu- lems [themselves] with the results dents about job displacement, as of AI tools.” This highlights the po- exemplified by Ethiopian software tential for AI to enhance learning developers worrying about “los- experiences. However, concerns ing job opportunities to AI” as ad- about over-reliance persist, with a vanced models increasingly auto- Cameroonian student expressing mate coding tasks. This alignment “skepticism about relying solely on requires ongoing collaboration be- AI for answers,” fearing impacts on tween academia, industry, and gov- critical thinking. Thus, HEIs face the ernment to identify emerging skills delicate task of leveraging these gaps and develop responsive edu- technologies to enhance learning cational programs. Public initiatives experiences while addressing val- such as Singapore’s ‘Learn with AI’ id concerns about over-reliance program can potentially generate and the potential erosion of crit- awareness, demystify AI, address ical thinking skills. The develop- concerns about potential job dis- ment of comprehensive AI fluency placement, and prepare society for among both students and faculty AI’s impacts. has emerged as a crucial impera- tive, encompassing not just tech- HEIs face an urgent imperative nical proficiency but also ethical to adapt to the AI-enabled world. assessment and critical evaluation This adaptation requires a funda- of AI’s societal impacts. HEIs must mental rethinking of core educa- foster an environment where AI is tional practices and structures. seen not as a replacement for hu- Students are already envisioning man intellect, but as a powerful tool significant changes, with Rwandan to augment learning, research, and students expecting AI to “disrupt innovation. This requires a shift in the traditional school system by rev- pedagogical approaches, curricu- olutionizing how education is ac- lum design, and institutional pol- cessed and delivered.” Institutions 28 100 Student Voices on AI and Education must respond to students’ needs these concerns while leveraging for AI-relevant skills, addressing AI’s potential to support inclusivity. concerns like those expressed by As an Ethiopian economics student Cameroonian students who felt ill- shared, AI can help in “summariz- equipped due to “lack of focused ing and making articles more un- curricula covering computational derstandable from paywalled and thinking [and] algorithm design.” complex resources,” potentially de- Faculty development programs are mocratizing access to knowledge. crucial, equipping educators not Supporting inclusivity goes beyond only to teach with AI tools but also surface-level considerations to ad- to prepare students for an AI-domi- dress deep-seated biases that may nated workforce. Clear institutional be embedded in AI algorithms or guidelines for AI use must be es- training data. Robust data privacy tablished, addressing issues of ac- and security measures are essential, ademic integrity, data privacy, and particularly given the sensitive na- ethical considerations. Leveraging ture of student information. Foster- AI for administrative efficiency, such ing interdisciplinary collaboration is as automating early warning sys- crucial in navigating these complex tems for at-risk students, can free issues, ensuring that the integration up resources for more impactful of AI in education considers not just educational initiatives. However, technical proficiency but also soci- as institutions embrace these tech- etal impacts and ethical consider- nologies, they must be vigilant in ations. addressing infrastructure dispari- ties. Ensuring equitable access to Looking to the future, the HE AI tools, high-speed internet, and landscape will continue to evolve necessary hardware is essential to rapidly in response to AI advance- prevent the exacerbation of exist- ments, requiring institutions to ing educational inequalities. This remain agile and forward-think- may require targeted investments ing. Students are already anticipat- in digital infrastructure and support ing significant changes, with Ethi- systems, particularly for students opian participants foreseeing AI from disadvantaged backgrounds. The ethical implications of AI in education are profound and mul- tifaceted, demanding prioritiza- tion at every level of implemen- tation. Students across regions have raised important ethical con- cerns. An Indonesian art student condemned “people who use AI 100% and then claim it as their job,” Ensuring equitable highlighting issues of authentici- access to AI tools, high- ty and attribution in creative fields. speed internet, and Nigerian students worried about necessary hardware “the temptation to become overly dependent on AI for generating ac- is essential to prevent ademic content rather than thinking the exacerbation of critically.” Institutions must address existing educational inequalities. 100 Student Voices on AI and Education 29 democratizing “accessibility to in- ing new programs and pedagog- formation and knowledge resourc- ical approaches to address these es that were previously difficult to emerging areas. As an Indonesian obtain.” However, they also warned journalism student highlighted, about risks of “exacerbating socio- there’s “potential for utilizing AI to economic inequalities if digital dis- efficiently generate text from news parities persist.” The intersection of reports,” suggesting the need for AI with other emerging technolo- curricula that prepare students for gies promises to create new fields AI-augmented professional prac- of study and career paths. As a tices. Cultivating transferable skills Colombian student noted, there’s like critical thinking, creativity, emo- growing interest in roles like “train- tional intelligence, and adaptability ing AI tools to describe images,” becomes even more crucial in this indicating the emergence of new rapidly changing landscape. The AI-related careers. In fact, the inter- concept of lifelong learning must section of AI with other emerging be deeply embedded in education- technologies, such as virtual and al philosophies, preparing students augmented reality, biotechnology, for careers that may require contin- and quantum computing, promises uous reskilling and upskilling. As to create entirely new fields of study labor markets undergo AI-driven and career paths. HEIs must be pre- transformations, HEIs must forge pared to adapt quickly, develop- stronger partnerships with indus- try, government, and civil society to anticipate future skills needs and develop responsive educational models. Ultimately, the successful integration of AI in education will depend on a collaborative, ethical, and human-centered approach that harnesses the technology’s poten- tial while addressing the concerns and aspirations voiced by students across diverse global contexts. Ultimately, the successful integration of AI in education will depend on a collaborative, ethical, and human- centered approach 30 100 Student Voices on AI and Education References Abbas, M., Jam, F. A., & Khan, T. I. (2024). International Monetary Fund. (2024). Is it harmful or helpful? Examining Gen-AI: Artificial Intelligence the causes and consequences and the Future of Work (IMF Staff of generative AI usage among Discussion Note No. SDN/2024/001). university students. International Washington, D.C.: International Journal of Educational Technology in Monetary Fund. Higher Education, 21(1), 10. Mills, K., Ruiz, P., & Lee, K.-w. (February Bekiaridis, G., & Attwell, G. (2024). 21, 2024). Revealing an AI Literacy Supplement to the DigCompEDU Framework for Learners and Framework: Introduction to AI Educators. Digital Promise logo. in Education1. Active Citizens [Blog post]. Partnership & Pontydysgu. OECD (2023). AI and the Future of Skills, Castillo, E. (2023). These Schools and Volume 2 Colleges Have Banned Chat GPT and Similar AI Tools. Southworth, J., et al. (2023). Developing a model for AI Across the curriculum: Darvishi, A., Khosravi, H., Sadiq, S., Transforming the higher education Gašević, D., & Siemens, G. (2024). landscape via innovation in AI Impact of AI assistance on student literacy. Computers and Education: agency. Computers & Education, Artificial Intelligence, 4, 100127. 210, 104967. Timea & Veres (2023). Students’ de Bem Machado, A., dos Santos, J. R., Perceptions of Artificial Intelligence Sacavém, A., & Sousa, M. J. (2024). in Higher Education Digital Transformations: Artificial Intelligence in Higher Education. UK Department for Science, Innovation In Digital Transformation in Higher & Technology (2024). A pro- Education Institutions (pp. 1-23). innovation approach to AI regulation: Cham: Springer Nature Switzerland. government response. Edwards, B. (2024). Nvidia CEO calls for United Nations (2024). Seizing the “Sovereign AI” as his firm overtakes opportunities of safe, secure and Amazon in market value. trustworthy artificial intelligence systems for sustainable European Parliament (2024). Artificial development. Intelligence Act WIPO (2024). What is Artificial EU AI Act (2024). Article 3: Definitions Intelligence? Filippucci, F. et al. (2024). Should AI Yin, R. K. (2014). Case study research: stay or should AI go: The promises Design and methods (5th ed.). and perils of AI for productivity and Thousand Oaks, CA: Sage. growth Gimpel, H., Gutheil, N., Mayer, V., Bandtel, M., Büttgen, M., Decker, S., & Urbach, N. (2024). (Generative) AI Competencies for Future-Proof Graduates: Inspiration for Higher Education Institutions. Hohenheim Discussion Papers in Business, Economics and Social Sciences. 100 Student Voices on AI and Education 31 Annex 1: AI Benefits and risks Dimension AI AI Risks Policy Benefits Recommendations for AI in Education Curriculum devel- AI-powered labor AI-generated -Implement guidelines and standards opment (knowl- market-insights curricular con- for AI content quality assurance to miti- edge/ skill) can improve tent may be gate biases and inaccuracies. course design. biased, inaccu- -Encourage diversity and inclusion in AI AI can help gen- rate or sexist (or development teams to ensure balanced erate lesson plans low quality) perspectives. and creative visu- Over-reliance on -Establish mechanisms for curriculum al content AI may inhibit external experts to provide oversight curriculum ex- and validation of AI-generated content. perts’ agency. Instruction and AI-powered Excessive use of -Develop policies to ensure the respon- Assessment (ped- adaptive learning AI may de-hu- sible integration of AI in teaching and agogy/applying tools provide manize teaching assessment practices, balancing auto- knowledge) personalized experience. mation with human interaction. learning. AI-assessment -Implement bias detection and mitiga- Automated grad- tools may ampli- tion protocols in AI-assessment tools. ing and feedback fy biases -Provide training and support for edu- reliefs burden of cators to understand and address the instructors implications of AI in pedagogy. Learner Agency AI can support Over-depen- -Foster policies that promote a bal- and Independence self-directed dence on AI may anced approach to AI integration, em- (formal/informal learning, e.g. AI limit students’ powering learners while emphasizing learning) personal tutors critical thinking the importance of critical thinking skills. -Establish guidelines for AI usage in education that encourage learner au- tonomy and independence. -Invest in initiatives to bridge the digital divide and ensure equitable access to AI-enabled resources for all learners. Inclusion AI-powered auto- AI may perpetu- -Enact policies to address bias in AI matic captioning ate bias algorithms and promote diversity in AI benefits learners Digital divide development teams. with disabilities may become an -Implement measures to mitigate the Creation (or ad- AI divide digital divide and ensure equitable aptation) of con- access to AI technologies. Women and oth- tents into minori- -Support initiatives that promote inclu- er underserved ty languages sivity and diversity in AI research, devel- groups may be opment, and deployment. left behind Administration, Automate early AI raises privacy -Establish comprehensive policies and Governance, and warning systems concerns regulations for AI governance, empha- Policy Speed up data Bad actors may sizing data privacy and security. processing and use AI to cause -Strengthen enforcement mechanisms analysis for deci- harm to prevent misuse of AI technologies sion making Ef- and address potential harms. Lack of guid- fective guidance -Provide guidance and support for ed- ance and sup- and regulation ucational institutions to navigate ethical port my leave on AI can better and legal implications of AI adoption. students in a prepare the com- -Ensure transparency and accountability weak position munity in AI-related decision-making process- es. 32 100 Student Voices on AI and Education Annex 2: Data analysis process followed for focus group discussions The data collection and analysis tion placed on them”.19 In general, process involved several steps the research team used data reduc- leveraging both human exper- tion techniques at an early stage of tise and AI tools. During the focus data gathering and continued with groups, facilitators took notes orga- this approach until this research nized around the predetermined was finalized to avoid making data research themes, while also allow- analysis an endless process. ing for an open-ended approach. After completing the discussions, LIMITATIONS OF STUDY the full notes were structured into This research initiative aims to the main themes of access/use of AI, new skills/ways of learning, career provide rich descriptions of stu- opportunities, and education read- dents’ experiences with AI in iness. The voice recordings were higher education to have a broad- transcribed using the AI speech- er understanding of the impact to-text tool Rev.ai to enrich the fa- of these technologies. This study cilitator notes. For focus groups does not take the prescriptive role conducted in other languages, the of offering solutions that can be notes were translated into English later ‘generalized’ to a larger pop- using ChatGPT. The translated notes were then input into Claude. ulation. This is not to say that the ai to systematically structure the in- results will only apply to the initial formation and generate draft coun- study context; in fact, findings from try reports following the research the context in which this research themes. This AI-assisted process is based can be ‘transferrable’ to allowed the team to efficiently ana- other contexts that are congruent lyze the qualitative data across mul- with the context of the present case tiple countries. Finally, the research- study.20 ers conducted a transversal analysis synthesizing the key findings that cut across the individual country re- ports. The data obtained from focus group discussions provided rich information for further analysis, but, at the same time, generated an enormous amount of informa- tion, as generally happens with qualitative research, especially with methods such as focus groups. The research team selected the infor- mation considered more relevant to 19 Cohen, L., Manion, L., & Morrison, K. (2011a). this study, avoiding being “over-se- Approaches to qualitative data analysis. In Re- lective, unrepresentative, and un- search methods in education (7th ed., pp. 559– 573). London, England: Routledge. fair to the situation in hand in the 20 Lincoln, Y., & Guba, E. (1985). Naturalistic choice of data and the interpreta- inquiry. London: SAGE Publications. 100 Student Voices on AI and Education 33 34 100 Student Voices on AI and Education 100 Student Voices on AI and Education 35 36 100 Student Voices on AI and Education