Opportunities from Generative AI

Generative AI Use In The Classroom: Student Perception And Learning Outcomes

cdtlsd Lighting Talks , , , ,

Marissa K. L. E*  and Misty So-Sum WAI-COOK 

Centre for English Language Communication (CELC), NUS 

*elcmari@nus.edu.sg 

E, M. K., L., & Wai-Cook, M. S. S. (2024). Generative AI use in the classroom: Student perception and learning outcomes [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-marissakle-wai-cook/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Generative AI, academic literacies, higher education, academic writing 

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

The emergence of Generative AI has required educators to re-think their approaches to teaching and learning. In the context of the academic literacies classroom, the use of Generative AI has caused mixed reactions among educators. This is because of how the technology can potentially provide useful writing support to students, yet also be easily used to produce texts on demand that can meet assignment requirements. As such, calls have been made for writing instructors to teach students how to appropriately harness Generative AI as a tool to enhance their writing and research skills (Johinke et al., 2023).  

 

In response to this call, we describe here, two exploratory case studies from two different writing courses where Generative and semi-Generative AI tools have been used in the academic literacies classroom to teach students how to develop initial topic ideas for a research paper, assist them with the writing process and provide them with the opportunity to critically assess different platforms to search for scholarly sources for the research paper. The main difference between Generative and semi-Generative AI tools is that the latter is not as freely generative as the former. 

 

The focus in the first case is to examine if, and how, students use Generative and semi-Generative AI tools to conceptualise ideas and assist them in writing research papers, and their perceptions of the effectiveness of using such tools during their writing process. Querying student perception regarding their experience with such tools is important as it can be associated with their motivations underlying the use of the tools, as well as useful insights about the utility of such tools for future student cohorts (Chan & Hu, 2023). A survey was used in this first case.  

 

In the second case, the use of Consensus.ai, a semi-Generative AI tool employed for searching for research sources, is described. We show here how students not only used the tool to search for research sources, exhibiting useful research skills, but also used critical thinking (Huang et al., 2024) to discern whether these sources are useful in the context of a proposed small-scale research study assignment.  

 

Figure 1 shows an extracted portion from a graphic organizer used in class where students had to critically compare three platforms—Consensus.ai, Google Scholar and NUS Libraries databases like JSTOR. The first platform is the semi-Generative AI platform while the remaining two are more traditional platforms used for searching for research sources. Students used the organiser to pen down how the three platforms helped them with their literature search. In Figure 1, we can see how the student has formed a critical opinion about two of the platforms. He first points out how Consensus.ai has its strengths, like its ability to facilitate quick identification of the thesis of an article, in comparison to Google Scholar. He is also able to identify how Google Scholar has an advantage over Consensus.ai with its ability to cover a wider range of articles in comparison.  

 

Figure 1. Graphic Organiser

 

The fact that Generative AI is increasingly capable of producing writing that is grammatically correct and relevant to a given prompt, with only the characteristic of style lacking at this point, means that academic literacies courses need to go beyond text production and focus as well on how thinking skills like ideation that go into the process of producing text can be facilitated (Dai et al., 2023).  

REFERENCES

Chan, C. K. Y., & Hu, W. (2023). Students’ voices on generative AI: Perceptions, benefits, and challenges in higher education, International Journal of Educational Technology in Higher Education, 20(43), https://doi.org/10.1186/s41239-023-00411-8 

Dai, Y., Liu, A., & Lim, C. P. (2023). Reconceptualizing ChatGPT and generative AI as student-driven innovation in higher education. Procedia CIRP, 119, 84-90.  https://doi.org/10.1016/j.procir.2023.05.002

Huang, C. W., Coleman, M., Gachago, D., & Van Belle, J. P. (2024). Using ChatGPT to Encourage Critical AI Literacy Skills and for Assessment in Higher Education. In Van Rensburg, H.E., Snyman, D.P., Drevin, L., Drevin, G.R. (Eds.), ICT Education. SACLA 2023. Communications in Computer and Information Science, 1862. Springer, Cham. 

Johinke, R., Cummings, R., & Di Lauro, F. (2023). Reclaiming the technology of higher education for teaching digital writing in a post-pandemic world. Journal of University Teaching and Learning Practice, 20(2), https://doi.org/10.53761/1.20.02.01 

Opportunity, Implementation, Evaluation: Raising Awareness Of Generative AI Use In A Business Communication Course

cdtlsd Lighting Talks , , ,

Aileen Wanli LAM 

Centre for English Language Communication (CELC), NUS 

aileenlam@nus.edu.sg

Lam, A. W. (2024). Opportunity, implementation, evaluation: Raising awareness of generative AI use in a business communication course [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-awlam/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Generative AI, Awareness, Business Communication 

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

Generative AI adoption has surged over the past two years, and AI is now being used in many business functions and by more individuals than in previous years (Mckinsey, 2024b). A growing number of global CEOs and investors believe that AI adoption is crucial for productivity and value creation, so they seek out employees who are able to harness the use of AI (PWC, 2024) while employing higher cognitive skills such as critical thinking and decision-making (Mckinsey, 2024a). Since AI advancements have led to “significant implications for business communication and…pedagogy” (Sharma & Pandey, 2024, p.2), there is a need to relook business communication curriculum to remain relevant and up-to-date.   

  

This paper outlines an action research which started from a gap observed in a business communication course, an opportunity to gather information to develop a plan in Semester 1, the implementation of the plan in Semester 2, and the observation and analysis of the implementation post-semester. The findings were reflected upon and used to refine the approach in the next academic year. Though the findings are unique to the context of the course, the approach and tools used can be applied to other courses and disciplines.  

 

As a course coordinator of an undergraduate business communication course, I had two questions on my mind: “What generative AI content should be included in a business communication course?” and “How effective would the introduction of generative AI content be?”   

  

In the first semester, an opportunity to collect information presented itself with the call from the academic community to acknowledge and cite the use of generative AI. This led to the creation and introduction of the AI Declaration forms (Lam, 2024b) for two assessments—the business proposal and the pitch. Students declared their use of generative AI and more specifically wrote about the tools they were using, and how they were using it (i.e. prompts and what they did to the outputs). The data was analysed and coded into themes (e.g. stages of writing or preparation) to inform the curriculum updates that focused on the principles and limitations of using AI tools in business communication. The key learning objective that was emphasised throughout was critical thinking through evaluation of generated outputs and through creation—their ability to combine parts to form a coherent whole or new product (Bloom’s 2001 revised taxonomy)  

 

The following topics (AlAfnan et al, 2024; Lam, 2024a; Riapina, 2023) were discussed over the semester 

  • Use of AI tools for research and idea generation with an emphasis on evaluation and creation (critical thinking)  
  • Issues of ownership of AI-generated outputs (e.g. terms of use, licenses, laws) and the implications of using AI in school   
  • Use of AI for the editing process (e.g. grammar, spelling, enhancing vocabulary, paraphrasing, style, tone) and implications of using AI at work   
  • Use of AI for presentations (i.e. creation of slides and visuals; writing of scripts)  
  • Use of AI to prepare for the Q&A (e.g. generating questions and practicing answers) with an understanding of the context  

  

In order to understand the effectiveness of the implementation, (1) a perception survey, and (2) an analysis of the AI declaration forms for semester 2 were carried out.  

 

In the perception survey, participants responded that their awareness of the principles and potential issues of using AI had increased from an average of 3.43 (pre-course) to 4.08 (post-course) on a scale of 5 (0= No Knowledge; 5= Very Knowledgeable) but they had varying responses to the usefulness of each topic. The analysis of responses in the AI declaration forms for Semester 2 found instances of critical thinking (i.e. student evaluation of generated outputs and the creation of coherent and/or original work). 

 

This presentation shares the findings of this action research and the significance of the approach. 

REFERENCES

AlAfnan, M. A., Dishari, S., & MohdZuki, S. F. (2024). Developing Soft Skills in the Artificial Intelligence Era: Communication, Business Writing, and Composition Skills. Journal of Artificial Intelligence and Technology. http://dx.doi.org/10.37965/jait.2024.0496

Anderson, L.W. (Ed.), Krathwohl, D.R. (Ed.), Airasian,P.W., Cruikshank, K.A., Mayer, R.E., Pintrich, P.R.,Raths, J., & Wittrock, M.C. (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom’s Taxonomy of Educational Objectives. Longman

Lam, A. W. (2024, May 1). Embracing AI in education. Faculty Focus. https://www.facultyfocus.com/articles/teaching-with-technology-articles/embracing-ai-in-education 

Lam, A. W. (2024, May 28). Learning from learners: Student use of AI. The Teaching Professor.  https://www.teachingprofessor.com/topics/teaching-strategies/teaching-with-technology/learning-from-learners-student-use-of-ai/ 

Riapina, N. (2023). Teaching AI-enabled business communication in higher education: A practical framework. Business and Professional Communication Quarterly, 23294906231199249. https://doi.org/10.1177/23294906231199249

Sharma, D., & Pandey, H. (2024). Pedagogical Impact of Text-Generative AI and ChatGPT on Business Communication. Business and Professional Communication Quarterly, 23294906241249113. https://doi.org/10.1177/23294906241249113 

Mckinsey. (2024a, March 18). The human side of generative AI: Creating a path to productivity    https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-human-side-of-generative-ai-creating-a-path-to-productivity 

Mckinsey. (2024b, May 20). The state of AI in early 2024: Gen AI adoption spikes and starts to generate value. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai 

PWC. (2024, May 21) AI Jobs Barometer. https://www.pwc.com/gx/en/issues/artificial-intelligence/ai-jobs-barometer.html 

Navigating Authentic Learning In The Age Of Generative AI

cdtlsd Lighting Talks , , , ,

Sharon LAU Pui Wan 

NUS-ISS 

sharon.lau@nus.edu.sg 

Lau, S. P. W. (2024). Navigating authentic learning in the age of generative AI [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-spwlau/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Generative AI, Authentic Learning, Critical Thinking, Educational Strategies 

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

In an era where generative AI is rapidly transforming educational landscapes, educators face both unprecedented opportunities and formidable challenges. As an adult educator in NUS, I propose to address the most pressing issue: how can we foster authentic learning experiences while leveraging the capabilities of generative AI? This lightning talk aims to provide a roadmap for educators to navigate this complex terrain, addressing the following aspects: 

Embrace AI as a co-educator 

Generative AI can serve as a dynamic partner in the educational process, providing personalised feedback, facilitating adaptive learning, and enabling interactive simulations. By integrating AI-driven tools, educators can create more engaging and tailored learning experiences. An example from my own practice involves using AI tools to simulate real-world business scenarios. Students interact with AI-generated market data, making strategic decisions that mirror professional challenges. This method not only deepens their understanding but also prepares them for the complexities of the professional world. The AI acts as a co-educator, providing immediate feedback and alternative perspectives, thereby enriching the learning experience. 

Cultivate critical thinking  

One of the core components of authentic learning is the development of critical thinking skills. AI can play a crucial role in this area by presenting students with complex, open-ended problems that require creative solutions. Research has shown that AI can aid in the development of higher-order thinking skills by challenging students to analyse, interpret, and synthesise information. For example, in my classes, I utilise AI to design challenging case studies that require students to engage deeply with the material. These case studies present AI-generated data that students must critically evaluate and interpret. This approach fosters a deeper level of engagement and helps students develop the critical thinking skills necessary to navigate an increasingly complex world. The use of AI in presenting diverse scenarios and data sets encourages students to consider multiple perspectives and develop well-rounded solutions. 

 

During this talk, I will draw from the latest research and my practical experiences to highlight key strategies and insights. I will share examples of real-world practice to illustrate my key messages. 

 

Rethinking Assessment in the Age of GenAI

cdtlsd Lighting Talks , , ,

YANG Yi 

Department of Civil and Environmental Engineering,
College of Design & Engineering (CDE), NUS

yangyi@nus.edu.sg  

Yang, Y. (2024). Rethinking assessment in the age of GenAI [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-yangyi/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Generative AI, assessment methodology, academic integrity 

CATEGORY

Lightning Talk

INTRODUCTION

The advent of generative AI (GenAI) has profoundly transformed various sectors, with education being no exception. As GenAI technology advances, it presents both opportunities and challenges to traditional educational practices. While GenAI can automate many teaching tasks such as generating content (such as Figures 1, 2, 3 generated by AI for the subsequent paragraphs) and preparing lesson plans, one primary concern among educators is upholding academic integrity, given that GenAI can complete many assessments with minimal human effort. To address these challenges and leverage the potential of GenAI, there is a need to rethink and redesign our assessment models. This paper explores the strategies to adapt our assessment approaches in alignment with GenAI to enhance education outcomes. 

BE CLEAR OF THE OBJECTIVE 

First and foremost, educators should define clear objectives for their assessments. The format of assessments should align with specific learning goals in the context of GenAI. For example, if the aim is to gauge students’ retention of basic concepts, an in-class test that forbids GenAI use might be the most suitable. In contrast, if the goal is to create an authentic environment and enhance students’ ability to search for and organise information, a take-home assignment will be suitable. As the availability of GenAI has already become part of the authentic learning environment, the use of GenAI should be permitted or even encouraged in such assessments. It is critical that when AI tools are accessible for the assessments, all questions and tasks should be tested with the latest version of GenAI to ensure the assessments meet the intended learning objectives. 

Figure 1. Be clear of the objectives of the assessment (image generated by Microsoft Copilot) 

EMPHASISE SOFT SKILLS 

Secondly, assessments should place greater emphasis on soft skills. In the age of GenAI, competencies such as communication skills and teamwork become even more vital, as many tasks that traditionally require hard skills can now be automated by AI. These soft skills truly differentiate humans from machines and are becoming increasingly valuable in the professional sphere. To cultivate these soft skills, educators may consider designing project-based assessments with greater scoring weightage on oral presentations, Q&A sessions, and role-playing activities [1]. These assessment models not only enhance the development of soft skills, but also eliminate concerns regarding AI-related academic integrity issues by evaluating real-time interactions. 

Figure 2. Emphasise soft skills in assessment (image generated by Microsoft Copilot) 

INTEGRATE AI INTO ASSESSMENT MODELS 

Thirdly, integrating AI into assessments can foster innovative evaluation models. Instead of viewing AI merely as a challenge to academic integrity, educators and institutions can harness its capabilities to enrich and enhance learning experiences. For instance, an assessment could involve students reviewing and critiquing answers generated by GenAI. In an Urban Planning and Design course at the University of Hong Kong, students are tasked to: i) generate an essay using GenAI tools, ii) critically evaluate the AI-generated content, and iii) compose their own essay based on their critique [2]. This approach not only deepens students’ understanding of the subject but also helps them recognise the strengths and limitations of AI tools. By engaging with AI-generated content, students also develop critical thinking and analytical skills as they learn to evaluate and refine the information provided.  

Figure 3. Integrate genAI in assessment design (image generated by Microsoft Copilot)   

CONCLUDING REMARKS 

Educators should swiftly adapt to the evolving landscape of AI technology. By aligning assessments with clear learning objectives, emphasising the development of soft skills, and leveraging AI capabilities in assessment design, educators can create a more effective and relevant educational experience. This proactive approach ensures that students are not only knowledgeable in their subjects but also well-prepared to navigate and excel in a world increasingly shaped by AI.

REFERENCES​

Sharma, R. C., & Bozkurt, A. (2024). Transforming education with generative AI: Prompt engineering and synthetic content creation. IGI Global. 

Chan, C. K. Y., & Colloton, T. (2024). Generative AI in higher education: The ChatGPT effect (1st ed.). Taylor & Francis. 

Harnessing GPT to Develop a Personalised eLearning Web Application for Java Programming

cdtlsd Lighting Talks , , , ,

LIU Fan*, Esther TAN Meng Yoke, and CHIA Yuen Kwan 

NUS-ISS 

*isslf@nus.edu.sg  

Liu, F., Tan, E. M. Y., & Chia, Y. K. (2024). Harnessing GPT to develop a personalised elearning web application for Java programming [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-lfan-et-al/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Generative AI, Personalised eLearning, ChatGPT, Java Programming 

CATEGORY

Lightning Talk

BACKGROUND

E–learning systems are gaining popularity due to their extensive scalability, flexibility, self-paced capability as well as easy accessibility. Artificial Intelligence (AI) has been used in e-Learning systems (Murtaza et al., 2022; Sayed et. al., 2023) to make personalised learning recommendations to learners. Specifically, Generative Pre-trained Transformers (GPT) such as ChatGPT (OpenAI, 2023) has transformed education by offering intelligent assistance, customised content, personalised learning materials to cater to the individual learner’s needs (Rasul, 2023; Jin & Kim, 2023; Gharbi et al., 2024). AI-based eLearning can significantly enhance learning efficiency through personalised content delivery to learners (Bozkurt, 2021; Chen et al., 2023). However, these AI-based personalised eLearning techniques have not been effectively integrated to provide diverse and dynamic eLearning content for learners based on their personalities and learning preferences. 

PURPOSE

Propose an approach to provide dynamic and personalised content for learners through integrated Generative AI techniques. We developed an eLearning web application that connects to ChatGPT API, which will automatically generate dynamic course contents/exercises/quizzes based on the learners’ personality such as level of language proficiency and personal learning preferences for Java foundations. 

RESEARCH QUESTIONS

  1. How to automatically generate dynamic and personalised Java foundational content for each learner on the fly?
  2. How to automatically generate dynamic and personalised coding exercise questions for each learner on the fly?
  3. How to provide a more efficient learning experience for each learner based on the learner’s personality and learning preferences? 

METHOD

Design and develop an eLearning web application that seamlessly integrates the following technologies:

  1. Automatically assesses learner’s learning preferences using the Felder-Silverman learning style model (FSLSM) (Felder & Silverman, 2012). The FSLSM categorises learners into specific dimensions based on their preferred ways of acquiring and processing information. This model is used in our eLearning application to customise the course delivery format to suit the learners’ learning preferences.
  2. Harness the GPT model to generate dynamic course contents
  3. Provide diverse and personalised eLearning content based on learners’ personality and learning preferences.

 

The general architecture of the eLearning application consists of three modules: web browser client, eLearning server, and Microservices (ChatGPT server). The web browser client enables learners to interact with the eLearning application. The eLearning server delivers the core functionality of the system which includes the generation of dynamic topics, personalised exercises, and quizzes while the Microservice represents the external ChatGPT server.  

KEY FINDINGS 

This study explores how ChatGPT with FSLSM can be harnessed to create a customised Java programming course and provide learners with dynamic and personalised course materials. This method promotes a customised learning journey catering to the different learner’s needs in terms of their levels of Java proficiency and learning preferences.  

FUTURE WORK

The future work of the proposed eLearning application is summarised as follows: (1) using ChatGPT model to generate comprehensive course content including advanced topics of Java, (2) extend this research work to generate dynamic and personalised course content for other object-oriented programming languages as C++, C#, and Python, (3) introduce foundational topics, exercises and quizzes with support of different learning methods, such as hands-on workshops, auto-grading and learning feedback and report., (4) extend the application of the personalised eLearning application to include other academic disciplines. This future work should consider the challenges of the use of ChatGPT in education with regards to the accuracy and reliability of content.

REFERENCES​

Bozkurt, A., Karadeniz, A., Baneres, D., Guerrero-Roldán, A. E., & Rodríguez, M. E. (2021). Artificial intelligence and reflections from educational landscape: A review of AI Studies in half a century. Sustainability, 13(2), 800. https://doi.org/10.3390/su13020800

Chen, E., Huang, R., Chen, H. S., Tseng, Y. H., & Li, L. Y. (2023, June). GPTutor: a ChatGPT-powered programming tool for code explanation. In International Conference on Artificial Intelligence in Education (pp. 321-327). Springer Nature Switzerland. 

Felder, R., & Silverman, L. (2012). Index of learning styles questionnaire. North Carolina State University. Retrieved from http://www.engr.ncsu.edu/learn ingst yles/ilswe b.html. 

Gharbi, M., Taib Mohtadi, M., & Fal Merkazi, A. (2024). Revolutionizing Moocs with fine-tuned Chatgpt: Personalized Learning At Scale. International Journal of Computing and Digital Systems, 16(1), 1-11. 

Jin, J., & Kim, M. (2023). GPT-Empowered Personalized eLearning System for Programming Languages. Applied Sciences, 13(23), 12773. https://doi.org/10.3390/app132312773

Murtaza, M., Ahmed, Y., Shamsi, J. A., Sherwani, F., & Usman, M. (2022). AI-based personalized e-learning systems: Issues, challenges, and solutions. IEEE Access, 10, 81323–81342. https://dx.doi.org/10.1109/ACCESS.2022.3193938

OpenAI (2023). ChatGPT: Optimizing Language Models for Dialogue. https://openai.com/index/chatgpt/

Rasul, T., Nair, S., Kalendra, D., Robin, M., de Oliveira Santini, F., Ladeira, W. J., & Heathcote, L. (2023). The role of ChatGPT in higher education: Benefits, challenges, and future research directions. Journal of Applied Learning and Teaching, 6(1), 41-56. http://dx.doi.org/10.37074/jalt.2023.6.1.29

Sayed, W. S., Noeman, A. M., Abdellatif, A., Abdelrazek, M., Badawy, M. G., Hamed, A., & El-Tantawy, S. (2023). AI-based adaptive personalized content presentation and exercises navigation for an effective and engaging E-learning platform. Multimedia Tools and Applications, 82(3), 3303-3333. https://doi.org/10.1007/s11042-022-13076-8

Leveraging Multi-Agent Generative AI for Next-Generation Education and Career Development

cdtlsd Lighting Talks , , , , ,

WANG Hongtai* and YEO Wee Kiang 

Department of Information Systems and Analytics,
School of Computing (SOC), NUS

*e1132287@nus.edu.sg

Wang, H., & Yeo, W. K. (2024). Leveraging multi-agent generative AI for next-generation education and career development [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-hwang-wkyeo/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Educational Support, Career Advisory, Generative AI, Multi-Agent System, Large Language Model 

CATEGORY

Lightning Talk

INTRODUCTION

In today’s complex education and career planning landscape, students face challenges in navigating vast resources and making informed decisions. Traditional career advisory services often lack the necessary personalisation. This project aims to leverage AI technologies, particularly large language models (LLMs) and multi-agent systems, to develop a novel solution that provides tailored guidance and support. 

The system’s functionality is divided into two core functionalities: Tutor and Career Advisory. Together, these features create a cohesive system that guides students through both academic and career development with personalised solutions. 

KEY CAPABILITIES 

Tutor Section 

  • Tutorial Teaching: This function introduces subject matter concepts progressively, from simple to complex, with clear examples for better understanding. 
  • Question-and-Answer: This function allows users to ask questions about the course materials and receive contextually relevant answers. 
  • Providing Review Questions: This function generates questions based on provided materials, testing students’ mastery of concepts through customised review questions. 

 

Career Advisory Section 

  • Generate Interview Questions. Generates tailored interview questions based on job descriptions by analysing required skills and knowledge. 
  • Analyse Shortcomings. Compare the student’s resume and the job description to highlight skill and knowledge gaps. 
  • Generate Cover Letters. Automatically examine the job description and user’s resume, extract the most relevant resume highlights, and generate a cover letter emphasizing the user’s advantages for the role. 
  • Assist Users in Gaining Essential Skills. Creates technology tutorials tailored to job descriptions, providing links to online courses and materials to enhance learning. 

METHODOLOGY 

Architecture 

The architecture of this project is designed as a dialogue system based on a generative AI model. The core component is a LLM agent that acts as a traffic controller and interacts with the user. Based on the user’s input prompt, this agent will discern whether the user’s queries are related to Tutor or Career Advisory and thereafter provide appropriate responses accordingly. 

If the user’s request is straightforward, the LLM agent handles it directly. For more complex requirements, the agent collects detailed information about the user’s needs and activates specialised agent groups (Cheng et al., 2024) designed to address specific issues. An agent group consists of multiple LLM agents created using Autogen (Wu et al., 2023) that is responsible for one specific task. These specialised agent groups include: 

  1. Education Agent Groups: Responsible for tasks such as coaching users with course materials, answering their questions, and providing review questions. 
  2. Career Agent Groups: Responsible for generating interview questions, analysing the user’s shortcomings, generating cover letters, and providing learning resources. 

Figure 1. Diagram of our system with LLM agent and role-based groups.

KEY FINDINGS 

Interface Design 

In the Tutor section, users engage directly with agent tutors through a system that allows seamless switching between instructional content and question-and-answer sessions. 

Figure 2. Our AI platform assists users in learning Operations Management through dialogue. 

 

The Career section offers unrestricted communication with AI-powered LLM agents, providing essential career advisory. This includes generating personalised tables that identify skill gaps relative to job requirements and assisting in crafting tailored cover letters to improve job applications. 

Figure 3. User interface for generating a comparison table between job requirements and applicant qualifications. 

SIGNIFICANCE OF THE PROJECT 

  • Enhanced Personalisation: Tailored learning experiences with progressive subject matter introduction, customized Q&A support, and personalised review questions. Career guidance includes tailored interview questions, skill gap analysis, and personalized cover letters. 
  • Improved Accessibility: 24/7 support with easy access to relevant online courses and materials. 
  • Efficiency: Streamlined assistance through specialized agent groups, providing quick and precise help. Identifies skill gaps and offers targeted learning recommendations. 
  • Career Preparation: Generates mock interview questions and skill gap analyses for better preparation. Automatically creates personalised cover letters and offers resume advice. 

REFERENCES​

Cheng, Y., Zhang, C., Zhang, Z., Meng, X., Hong, S., Li, W., Wang, Z., Wang, Z., Yin, F., Zhao, J., & He, X. (2024). Exploring large language model-based intelligent agents: Definitions, methods, and prospects.  ArXiv. abs/2401.03428. 

Wu, Q., Bansal, G., Zhang, J., Wu, Y., Li, B., Zhu, E., Jiang, L., Zhang, X., Zhang, S., Liu, J., Awadallah, A.H., White, R.W., Burger, D., & Wang, C. (2023). AutoGen: Enabling next-gen LLM applications via multi-agent conversation. ArXiv. abs/2308.08155v2.

The Usage of Generative Artificial Intelligence (AI) Avatars in Psychiatric Medical Education

cdtlsd Lighting Talks , , ,

Charlene GOH Xing Le1, Eugene WANG1, Judy C. G. SNG2,*

1Yong Loo Lin School of Medicine (YLLSOM), NUS
2Department of Pharmacology, YLLSOM, NUS 

*phcsngj@nus.edu.sg

Sng, J. C. G., Goh, C. X. L., & Wang, E. (2024). The usage of generative artificial intelligence (AI) avatars for psychiatric medical education [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-jcgsng-et-al/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Opportunities from Generative AI, video avatars, patient simulation, medical education

CATEGORY

Lightning Talk

BACKGROUND

The integration of generative artificial intelligence (AI) into our daily lives has significantly increased in recent years, with numerous healthcare institutions leveraging AI and medical technology to enhance diagnostic and treatment accuracy. However, its utilisation in medical education remains limited. This project aims to bridge this gap by promoting the incorporation of AI into medical education. This is done by supplementing existing resources with AI-generated videos of simulated patients. Currently, the range of available avatars online is limited and not entirely relevant to the context of Singapore. By providing virtual scenarios featuring patients with psychiatric medical conditions, we aim to develop essential skills in medical students, such as empathy and sensitivity, within a controlled environment. This approach seeks to enhance students’ clinical skills and decision-making, ensuring they are well-prepared for real-world patient interactions as practitioners. 

METHODS

AI technology was employed at multiple stages of building the video avatars. Initially, a doctor- patient conversation is generated using ChatGPT-4, after inputting a prompt that corresponds to a specific psychiatric patient profile. The patient profile includes the positive and negative signs presented by the patient, their socio-economic background, past medical history, and the suggested treatment plan. Medical experts in the psychiatric field meticulously crafted the simulated patient’s information to ensure an accurate representation of the condition. Two main symptoms were chosen: low mood (including adjustment disorder, dysthymia, and major depressive disorder) and chest tightness (not amounting to chest pain, including adjustment disorder, panic disorder, and somatic symptom disorder). After fine-tuning the script to ensure the scenario was localised and culturally accurate to Singapore’s context, the D-iD software was utilised to generate an AI video of a patient, emulating sentiments and synchronising speech with facial expressions. 

 

The video is then shown to students, whose objective is to accurately diagnose the patient. The patient avatar video aims to increase interactivity with students in a classroom setting, enhancing their diagnostic skills and understanding of psychiatric conditions. 

RESULTS

Preliminary feedback from medical professionals suggests its great potential to become a staple resource for healthcare students. In the coming months, we plan to pilot this with a small group of healthcare students to gather feedback and suggestions to refine the ensemble of avatars we currently have. The data collection will include metrics such as diagnostic accuracy and effectiveness in improving the students’ knowledge. Further studies will be conducted to determine the usage and effectiveness in nurturing the holistic development of medical students, encompassing both academic and psychosocial skills. 

CONCLUSION

As AI integrates more seamlessly into the medical curriculum, it is crucial to cultivate an engaging, interactive, and safe space for students, bridging the gap between theoretical knowledge and clinical practice. We believe that incorporating AI as early as possible in medical education is vital for developing holistic practitioners for a sustainable future, ultimately improving patient care in healthcare institutions. 

 

Infusing Contextual Elements With Generative AI Tools To Reinforce Learning For Students

cdtlsd Lighting Talks , , , ,

TAN Chun Liang

Department of Architecture,
College of Design and Engineering (CDE), NUS

tcl@nus.edu.sg 

Tan, C. L. (2024). Infusing contextual elements with generative AI tools to reinforce learning for students [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-cltan/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Urban Greening, Site context, Video assignment, Peer review 

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

The recent rise of AI content generators has had significant impact on both learning and teaching in educational institutions. Although the university encourages the responsible use of AI content-generation tools, a point of concern is on the authenticity of student report submissions. How do we ensure that usage of such platforms can help us augment the learning experience and not become a tool for students to conveniently churn out AI-generated reports at the eleventh hour to submit as their own?  

 

This is an important question to address for the course LA5303 “Urban Greening: Technologies and Techniques”, where students are taught ways to utilise urban greenery to improve the environment. Without a real understanding of the course content, students may not learn about the proper ways of urban greening and fall back to more superficial and cosmetic treatments of the landscape, leading to greenwashing. 

 

Drawing on the Significant Learning pedagogical framework (Fink, 2013), I try to tap on the Human Dimension, in learning about oneself in others to reinforce learning. I explored peer learning and the theory of Distributed Cognition (Hutchins, 2020) to acquire knowledge through an individual’s social and physical environment. In this manner, cognitive resources can be shared socially, and students can achieve more as a group than with just individual effort.  

 

Strategy 1: Experiential learning, peer learning, and industry talks 

Singapore is adorned with many examples of urban greenery projects, ranging from ground-level parks to sky-rise greenery projects. Since this course is about urban greenery, I began to introduce more activities to encourage more experiential and peer learning: where students go for site visits, assess real projects in Singapore and learn from each other instead of having me give them second- or third-hand information via lectures.

 

Figure 1. Learning journey to SproutHub @ Henderson  

 

Learning journeys to prominent urban greening spaces such as SproutHub@Henderson (Figure 1) and talks by industry partners (Figure 2) were organised to let students gain first-hand experience of the components of urban greenery covered in this course.  

Figure 2. Talks by Mr Christopher Leow, prominent urban farmer (Top row) and Representatives from Elmich Pte Ltd, sharing green wall and green roof products (Bottom row) 

 

In addition, students were tasked to visit urban greenery projects of their choosing and to create bite-sized (1-minute) videos documenting their learning (Figure 3). The videos were subsequently uploaded onto an online bulletin board (Miro) for others to review and comment on (Figure 4). Students were given the opportunity to provide comments on the videos produced by their peers. This directly increased their learning of urban greenery projects by 50 (total number of students in the course), as knowledge is now crowdsourced and contextual.  

 

Figure 3. Screenshots of a video done by a student, documenting urban greenery learning  

 

Figure 4. Comments from peers on the Miro board 

 

Strategy 2: Pre-emptive strike to nullify the impact of A.I. generated content  

Instead of warning students not to use AI-generated content, I insisted that it was the first thing they did for their assignment. Students had to append their prompts and AI-generated results in their report and show how they built on the AI content to come up with their final report. The AI-generated content thus became another layer of educational scaffolding for the students. More importantly, students were instructed to include specific examples of how to improve on urban greening using examples of videos done by their peers from Strategy 1. In this case, it became less likely for students to cheat with AI content generators such as ChatGPT because the examples are unique. 

REFERENCES​

Fink, L. D. (2013). Creating significant learning experiences: An integrated approach to designing college courses. John Wiley & Sons.

Hutchins, E. (2020). The distributed cognition perspective on human interaction. In Roots of human sociality (pp. 375-398). Routledge. 

Harnessing Generative AI As A Personalised Tutor: Enhancing Interdisciplinary Learning Outcomes For Biotechnology Graduate Students

cdtlsd Lighting Talks , , , , ,

Xin Xiang LIM 

Department of Biological Sciences,
Faculty of Science, NUS 

xinxiang@nus.edu.sg 

Lim, X. X. (2024). Harnessing generative AI as a personalised tutor: Enhancing interdisciplinary learning outcomes for biotechnology graduate students [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-xxlim/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

ChatGPT, Student’s Prompt Analysis, Interdisciplinary learning, Generative AI, Personalised Tutor

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

The escalating complexity of societal issues necessitates that graduates from higher educational institutions engage in problem-solving endeavours that transcend singular disciplines (Mansilla & Duraising, 2007; Repko., 2007). When students grasp and interconnect a diverse array of knowledge and skills, their educational experiences become more fulfilling, and their employment prospects broaden (Ivanitskaya & Montgomery., 2002). This phenomenon is particularly pertinent in biotechnology, where innovation is paramount in transforming laboratory discoveries into marketable products that address societal challenges. The innovation process extends beyond biology, requiring an understanding of market needs, funding sources, business models, risk management, and competitor analysis—critical real-world considerations. Thus, fostering innovation demands collaborative, interdisciplinary approaches to facilitate the cross-pollination of ideas and the integration of multiple perspectives. Equipping students with the skills to identify problems, develop prototypes, and conduct market research can provide a robust framework for innovation (Boms et al., 2022). 

 

Pharmaceutical sciences, chemistry, and biotechnology, typically have limited exposure to essential business considerations necessary for evaluating product market viability. This gap underscores the need for substantial scaffolding in business and innovation concepts to enable students to integrate their biotechnological expertise with business innovation, ultimately facilitating product creation. 

 

Generative AI offers significant potential to enrich learning experiences, fostering creativity, critical thinking, and motivation among students. ChatGPT, for instance, has demonstrated efficacy in enhancing interactive learning and personalised tutoring (Baidoo-Anu & Ansah., 2023). Leveraging ChatGPT promotes inquiry-based learning and student-centric approaches, both of which are effective in enhancing learning outcomes. Previous studies indicate that generative AI tools increase intrinsic motivation, conversational engagement, and continuous idea expression among students (Ryan & Deci., 2020) This context presents two educational opportunities: 1) inquiry-based active learning through prompt generation (prompt engineering), and 2) learning from generative AI responses. Developing writing prompts and prompting strategies has become a critical skill in handling generative AI. Prompt datasets collected from generative AI, including log data, can capture students’ learning processes in non-invasive ways. Prompt analysis thus provides an opportunity for educators to gain insights into students’ perceptions, motivations, and behaviors concerning interdisciplinary learning. 

 

Evaluation of students’ interdisciplinary learning outcomes will be conducted through three primary methods: 1) pre- and post-course surveys to capture students’ self-perceptions of their interdisciplinary skills and knowledge, providing insights into their metacognition and epistemology (Lattuca et al., 2012); 2) Assessment of student assignments and presentations using published, peer-reviewed rubrics for interdisciplinarity; and 3) Analysis of student prompts generated with ChatGPT. The student prompts will be analysed using natural language processing techniques, as outlined in recent studies, to gain insights into students’ cognitive processes and modes of thinking (Lee et al., 2023). Through this triangulated approach, the overall aim of the study is to comprehensively and rigorously evaluate the cognitive processes underpinning interdisciplinary learning. 

 

This lightning talk will specifically describe the insights gleaned from the analysis of students’ prompts so as to identify and evaluate possible learning barriers faced by students in the process of interdisciplinary learning. 

 

Figure 1. Interdisciplinary approach to biotechnology innovation. 

REFERENCES

Baidoo-Anu, D., & Ansah, L. O. (2023). Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI, 7(1), 52-62. psychology, 61, 101860. https://doi.org/10.1038/s41587-022-01253-x

Boms, O., Shi, Z., Mallipeddi, N., Chung, J. J., Marks, W. H., Whitehead, D. C., & Succi, M. D. (2022). Integrating innovation as a core objective in medical training. Nature Biotechnology, 40(3), 434-437. 

Boix Mansilla, V., & Dawes Duraising, E. (2007). Targeted assessment of students’ interdisciplinary work: An empirically grounded framework proposed. Journal of Higher Education, 78(2), 215-237. https://doi.org/10.1080/00221546.2007.11780874

Ivanitskaya, L., Clark, D., Montgomery, G. et al. (2002). Interdisciplinary Learning: Process and Outcomes. Innovative Higher Education 27, 95–111. https://doi.org/10.1023/A:1021105309984 

Lattuca, L. R., & Knight, D. B., & Bergom, I. M. (2012, June), Developing a Measure of Interdisciplinary Competence for Engineers. Paper presented at 2012 ASEE Annual Conference & Exposition, San Antonio, Texas. https://dx.doi.org/10.18260/1-2—21173 

Lee, U., Han, A., Lee, J., Lee, E., Kim, J., Kim, H., & Lim, C. (2023). Prompt Aloud!: Incorporating image-generative AI into STEAM class with learning analytics using prompt data. Education and Information Technologies, 1-31. https://doi.org/10.1007/s10639-023-12150-4

Repko, A. F. (2007). Integrating interdisciplinarity: How the theories of common ground and cognitive interdisciplinarity are informing the debate on interdisciplinary integration. Issues in Integrative Studies, 25, 1-31. http://hdl.handle.net/10323/4501

Reimagining Data Storytelling with Generative AI

cdtlsd Lighting Talks , , , , ,

Evelyn ANG 

Data Literacy Programme
Office of the President, NUS

eve.ang@nus.edu.sg

Ang, E. (2024). Reimagining data storytelling with generative AI [Lightning talk]. In Higher Education Conference in Singapore (HECS) 2024, 3 December, National University of Singapore. https://blog.nus.edu.sg/hecs/hecs2024-eang/

SUB-THEME

Opportunities from Generative AI 

KEYWORDS

Data storytelling, Generative AIs, adult learners, ChatGPT-4o, custom GPTs 

CATEGORY

Lightning Talk

EXTENDED ABSTRACT

Data storytelling is a new superpower for making complex data accessible and engaging (Loewen, 2024a). Schwabish (2014) as well as Green and Brock (2000) highlight how visual and narrative elements enhance comprehension and persuasion, essential for effective data communication. Dykes (2020) demonstrates through real-world examples how compelling data stories can lead to more informed business decisions. Loewen (2024b) describes data storytelling as the art behind the science—the art of making sense out of a deluge of data, shaping it into something that sticks. The integration of generative AIs in storytelling creates more engaging narratives, akin to how bards once used music to enliven stories. Despite myths about data storytelling being just simplistic visualisation, it can be said to be a misconception. Dykes explained that effective data storytelling uses coherent narratives supported by meaningful visualisations to engage audiences deeply. Moreover, Generative AIs democratise the ability to analyse vast datasets, allowing humans to focus on creativity and emotional intelligence (Dykes, 2024). By combining AI capabilities with human adaptability, data storytellers can make data insights more compelling and actionable. Li (2024) has done a detailed scan into data storytelling tools available, and most are prototypes for research purposes. McKinsey & Company (2024) published an article reporting a surge in AI adoption in at least one business function in early 2024. Generative AI adoption is moving beyond professional setting and is much more likely to be used in both work and personal settings. 

 

Generative AI is here to stay and beckons the question how we can purpose generative AIs in data storytelling. 

 

In this lightning talk, I will highlight broadly what is good data storytelling as suggested by Knaflic (2015) in her book Storytelling with Data in areas (1) understanding the context, (2) choose appropriate visual display, (3) eliminate clutter, (4) focus attention where you want it, (5) think like a designer, and lastly (6) tell a story. Now to address the elephant in the room—how will Gen AI fit into this picture? Recent work by Li (2024) proposed four distinct levels of AI involvement in working with data from the data workers’ perspectives, based on the levels of human agency versus AI automation. However, today’s advancement of AI has yet to be able to only perform a singular role with simple prompt inputs effectively. Kesari (2024) proposed a matrix of how different tools with GenAI are suited for different kinds of decisions to be made. 

 

How do we put all these together towards better data storytelling? I will broadly show how we can position fit-for-purpose use of GenAIs into the data storytelling preparatory work based on customGPTs. I will also weave in how GenAIs can be purposefully deployed so leaving us humans to do what we do best—creativity and connecting with our audience (Dyke, 2024). We will also visit how the most popular generative AI—ChatGPT—will be able to become your new companion in data storytelling through my CustomGPT—NarratEve. I will also touch on using Custom GPTs (Loewen, 2024) that can make your data storytelling and preparation even more effective. 

REFERENCES

Dykes, B. (2020). Effective data storytelling: How to drive change with data, narrative, and visuals. John Wiley and Sons, Inc. 

Dykes, B. (2024). The Future of Data Storytelling is Augmented, not Automated. Forbes. https://www.forbes.com/sites/brentdykes/2024/02/27/the-future-of-data-storytelling-is-augmented-not-automated

Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701-721. https://doi.org/10.1037//0022-3514.79.5.701

Kesari, G. (2024, 17 January). The Enduring Power of Data Storytelling in the Generative AI Era. MIT.edu. https://sloanreview.mit.edu/article/the-enduring-power-of-data-storytelling-in-the-generative-ai-era/

Knaflic, C. N. (2015).  Storytelling with Data: a data visualization guide for business professionals. Wiley. 

Li, H. (2024). Why is AI not a Panacea for Data Workers? An Interview Study on Human- AI Collaboration in Data Storytelling. arXiv 

Li, H. (2024). Where are we so far? Understanding Data Storytelling Tools from the perspective of Human-AI collaboration. arXiv 

Loewen, J. (2024). Custom GPT Creation For Data Visualization: A Step-by-Step Guide. Towardsai.net. https://towardsai.net/p/data-analysis/custom-gpt-creation-for-data-visualization-a-step-by-step-guide

Loewen, J. (2024a). Why Data Storytelling is Your New Superpower. Medium https://medium.com/data-storytelling-corner/why-data-storytelling-is-your-new-superpower-9f76e62762ce 

Loewen, J. (2024b). What the Heck is Data Storytelling Anyways? Here Are The Basics. Medium. https://medium.com/data-storytelling-corner/what-the-heck-is-data-storytelling-anyways-here-are-the-basics-c47c72cba44b

McKinsey & Company (2024). The state of AI in early 2024: Gen AI adoption spikes and starts to generate value. McKinsey https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai 

Schwabish, J. A. (2014). An economist’s guide to visualizing data. Journal of Economic Perspectives, 28(1), 209-234. https://dx.DOI.org/10.1257/jep.28.1.209

Viewing Message: 1 of 1.
 Success

Blog.nus login is now via SSO. Don't see or cannot edit your blogs after logging in? Please get in touch with us, and we will fix that. (More information.)