EdMedia 2024: Au Revoir Brussels, ¡Hola Barcelona

AACE EdMedia has been held annually since 1988, fostering innovation and collaboration among educators, researchers, and practitioners from around the globe. The 36^th conference from July 1-5 2024 in Brussels brought together 250 attendees from 48 different countries, representing over 90 different institutions. The conference program featured 150 presentations, four excellent keynotes, several hands-on workshops and a lively poster session, all organized via the online platform Academic Experts. EdMedia conference proceedings are available on LearnTechLib.

The first conference day was marked by excellent workshops that set a high standard for the event. Sarah Prestridge and Aideen Hunter delivered an insightful and highly practical session on scholarly writing patterns, providing techniques, strategies and structures for academic writing. Their article ‘Mastering the writing game: practical insights for early career researchers and supervisors’ is worth a read, no matter how many publications you have under your belt. Another workshop was conducted by Erfan Jalili Jalal, Nico van der Wiel, and Maartje Henderikx on large language models and prompting techniques.

The overarching theme of this year’s EdMedia conference was the role of generative AI in education. In this context, the conference highlight was the excellent keynote by Mike Sharples, emphasizing AI’s potential as a social learning agent. The talk was followed by an additional one-hour keynote conversation that forged connections and emerging understanding.

Sharples’ presentation underscored the social component of new technology, which is often underestimated, and explored how AI can simulate social intelligence, even if it lacks true understanding. Sharples began by discussing the essential elements of social AI, including persistent memory, autonomous agents capable of interaction, integration into social media, and intercommunicating AI tools. Despite these advancements, he posed the critical question: does AI possess social intelligence? His conclusion was that while AI can simulate social intelligence effectively, it does not genuinely understand or possess it.

He highlighted the inevitable integration of AI into social interactions, moving from simple prompt-response systems to continuous, persistent conversations with AI agents. This shift is already evident in applications like gaming and translation services. Sharples referenced Gordon Pask’s 1975 idea that all human learning involves conversation, suggesting that AI can participate in this process, simulating effective social interactions to facilitate learning.

Sharples discussed the ambiguous implications of AI in education through may pathbreaking examples. Using Google translate in conversation mode allows participants to communicate in their languages, mediated by AI. While the technology offers fascinating opportunities for translanguaging and multilingual settings, it also reduces the incentive to learn new languages and, by extension, their associated cultures.

One of the most compelling roles of AI, according to Sharples, is as a co-designer in creative processes. He gave examples of AI being used in textbook writing, where it can model different perspectives, and as a tool in reducing conspiracy beliefs through persuasive communication at scale. He posited that while AI is a master of mediocrity in writing, it holds significant potential as a writing companion, helping users develop their voice and enhance their work.

However, he cautioned against relying on AI for teaching as it lacks the intrinsic care that human educators provide. Sharples touched on the optimization dilemma in education: should we prioritize efficiency or care? He highlighted the UK Department of Education’s guidelines, which focus on efficiency in tasks like lesson planning. However, he argued that educational products must be grounded in an understanding of teaching and learning, and that the human element of care is essential and irreplaceable.

The keynote also addressed the inherent biases in AI, given its training on vast datasets and fine-tuning based on cultural perspectives, particularly a liberal American context.

Concluding his talk, Sharples shared his personal experiences with creating a GPT model, emphasizing the ease with which AI can be trained. After 30 years of research and one year of writing a book, it took only 30 minutes to train his GPT model TeachSmart. While the AI provided very good answers, it also included false information, such as references to learning styles. He described the process of fine-tuning the GPT by providing information about educational myths and the respective evidence-informed position. Overall, he urged educators to get hands-on with the technology, a sentiment that was shared in the joint keynote by Jo Tondeur and Sarah Howard.

One of the workshops offered by Martin Ebner and  Benedikt Brünner from Graz University focused on AI-generated videos, specifically multilingual video generation. Tools such as pictory.ai, HeyGen, auphonic.com, and elevenlabs.io were highlighted. Despite some technical issues, the workshop showcased the interesting concept of integrating native speakers for quality assurance and creating AI-generated audiovisuals with expert feedback loops.

Personally, it was a joy to present at AACE EdMedia about the book sprint with Thilo Harth from Fachhochschule Münster, and the collaborative writing class project ‘AI in Education’ at Asian University for Women with Sima Ahmadi and Israt Jahan Oeeshi. Liandi van den Berg and Jacobus Marthinus De Villiers from North West University (South Africa) had thoughtful feedback and great collaboration ideas for book sprints. I thoroughly enjoyed discussing AI ideas with Aaron Jones from Marshall University (Cyprus).

A panel discussion organized by Madalina Tincu (Johns Hopkins University) and Verona Leendertz (North West University) brought together participants from Canada, South Africa, Singapore, the US, and Italy to explore the emerging role of AI in instructional design. It was an amazing example of how the best sessions have the least technology. Without a working computer, the panelists ditched their presentation and focused on an open discussion. I was particularly pleased to connect with participants from last year’s workshop ‘AI in Education’ and foster new connections.

Participants shared a wide variety of experiences. Language learning has benefited from AI through text-to-speech technologies that can record speech and provide feedback. Instructional designers reported time-saving practices leveraging AI components of H5P.com for interactive content, as well as using generative tools to create summaries of readings, questions, quizzes, and multimedia content. Critical discussions also focused on testing AI’s ability to pass student exams, and how modifying and changing prompts can enhance learning. The future applications of AI in education remain uncertain, prompting encouragement for experimentation with free AI tools.

Some faculty expressed frustration over AI-generated student content. While some institutions have set hard limits on AI use, detection remains flawed. Traditional assessment methods, such as sit-down exams, were discussed as potentially more effective in mitigating AI-related challenges compared to longer assignments.  The emotional impact of grading AI-generated low-quality work was also acknowledged.

Throughout the conference, attendees and speakers shared many articles and useful resources on AI in Education that I will certainly follow up on.

• Elements of AI is a free online course by the University of Helsinki to foster AI awareness. It teaches the basics of AI and explains – among other things – types of machine learning.
• Oregon State University has developed an adapted version of Bloom’s taxonomy that takes into account generative AI.
• Georgia Tech has developed a virtual teaching assistant named Jill Watson, which leverages AI to support and enhance the educational experience for students.
• Several projects referenced Helen Crompton’s Framework on AI in education.
• The Australian government discusses assessment challenges surrounding AI, specifically the questions:  “Is the degree still worth the same? Is it devaluing our degrees?”. TEQSA (Tertiary Education Quality and Standards Agency) has released a discussion paper on ‘Assessment Reform in the Age of Artificial Intelligence’, addressing the implications and strategies for integrating AI in educational assessment practices.
• Speaking of assessment: MoodleGPT is a browser extension that allows quiz takers in Moodle to automatically fill in multiple choice quiz selection and answer open ended quiz responses with ChatGPT. It is likely that other learning management system quizzes have the same vulnerability – I checked this for UNC’s LMS Canvas and a simple Google search produced CanvasGPT.

The best conversations do not happen in sessions, they happen during breaks and social events. The conference featured many opportunities for informal connections, in particular the Beer Museum with its stunning rooftop views. It was likewise delightful to explore the city’s many excellent restaurants with international colleagues.

I look forward to next year’s EdMedia conference in Barcelona (May 2025). Hope to see you there!