Will 2023 be the year of Artificial Intelligence?

*Updated 13/01/2022 with minor additions and formatting changes.

Happy New Year! I’m making my first post in my refreshed online portfolio, with a new design and look for 2023! Thanks again for joining me and for taking the time to read this and other posts I’ve shared!

AI hits the mainstream

As the final weeks of 2022 rolled by, social media was ablaze with the release of ChatGPT, a chatbot which represents one of the most advanced generative AI experiences for everyday interaction.

After the launch of DALL-E 2, a text to image model that generates realistic images and art from a natural language description in April, and the release of text to video models by Meta and Google in October, it seemed like the future was now, but it was ChatGPT, with its freakish capacity to respond to seemingly any question with human-like dialogue that created the realisation that this type of AI had truly come of age.

Since its launch in late November, there has been debate and discussion about the impact of generative AI platforms such as ChatGPT on education. A tool that can (and does) create meaningful essay length text in any genre requested is an obvious threat to many learning and assessment tasks that ask students to do just that – exacerbated by the fact that the text is ‘original’ – and will not be identified by traditional plagiarism checkers. Articles with headlines stating that academics were stunned by ChatGPT’s essay writing skills and claiming the death of the college essay proliferated online, and some education systems made the decision to block the platform or return to paper and pen exams.

Like the calculator, the search engine, and countless other technological developments of the past, claims that students’ learning will be short-circuited by this tool are not surprising. ChatGPT is far from infallible (read here a fascinating and evolving list of ways in which the current iteration of ChatGPT fails) and it can only respond with data it has been ‘trained’ on – which does not include knowledge of events after 2021. Nor can it browse the internet or access information behind paywalls (such as journal databases). When it doesn’t know the answer, sometimes it ‘hallucinates’ – the AI term for ‘makes stuff up’. ChatGPT’s hallucination rate has been cited between 15-21%. It depends on what you’ve asked the chatbot as to whether it admits it doesn’t know. Here are two examples (click the image for full screen):

Samples of responses demonstrating ChatGPT’s capacity for accuracy.

From these you can see that the way the prompt is crafted will determine how the chatbot responds. You can also see that while it is able to generate a correctly formatted reference, the reference in fact is to an article that does not exist! This points to two things; firstly, no matter how convincingly the text is written or presented, it is not necessarily accurate, and MUST be evaluated; and secondly, to use ChatGPT effectively, some skill must be applied to the construction of the prompts – something people are already discussing in depth. Note also in this Reddit thread the interesting points of view regarding whether using ‘please’ and ‘thank you’ in prompts unneccessarily humanises the interaction, and the philosophical implications of this!

What does this all mean for teachers, and more specifically, for Teacher Librarians working in schools?

There are many fantastic educators and teacher librarians already sharing incredible reflections and collections of resources and ideas for how we can respond to this technological development. At the end of this post, I’ve included links to some of the ones I feel are most relevant or useful for teacher librarians – although if I don’t include something you’ve found useful, please share in the comments; there’s a lot being published at the moment!

Having said that, I would like to share my own emerging reflections on the role of the TL and generative AI.

For me, it feels as if generative AI (tools that can generate art, text, video, code, music etc) gives TLs the opportunity to (once again!) point out their expertise to school leaders and teachers, and to highlight all of the ways in which collaboration with the TL can lead to much stronger student outcomes – particularly in light of what these tools can do. Let me explain further.

1. The TL as leader of information and digital literacies
   
   The TL has always been a leader in information literacy and digital literacy – and has often led the charge in understanding new technologies, identifying their implications for learning and teaching and working with teachers to model and scaffold how these technologies could be used either by students in the process of learning, or by teachers to enhance their pedagogy and practice. The conversations around ChatGPT and awareness of how AI is developing is an opportunity for TLs to once again initiate balanced and informed understandings of the real implications for learning and teaching. Yes, the IT staff will also be keen to be involved; but no other role has the pedagogical and curriculum knowledge, combined with a deep understanding of academic integrity, information literacy and fluency and critical evaluation of information that the TL brings to the table.
2. The TL as leader of digital citizenship and media literacy

   TLs are well positioned to design and implement digital citizenship and media literacy programs that include algorithmic and AI literacy. These are not really new additions to what TLs already teach, but the evolution of generative AI certainly has brought them to the forefront. Being aware of the ethical implications of AI technologies – such as algorithmic bias, copyright and intellectual property issues, use of private data, environmental impacts – as well as the increased need for critical evaluation of online information brought about the facilitation of disinformation generation, amplification of bias etc generated by these tools are just part of what the TL can teach students (and teachers) or can consult with teachers to embed within current curriculum.

3. The TL as leader in curriculum and assessment planning
   
   TLs can help educators understand the academic integrity implications of these tools, and can work extensively with teachers to design learning that focuses on critical thought and creativity, as well as more innovative assessments that reduce opportunities for cheating. Students cheat for many reasons (we are all humans), but with their extensive knowledge of inquiry learning, TLs can work with teachers to focus on of the research/inquiry learning PROCESS. Generative AI tools may be fantastic at creating a finished product, and to some extent they can produce a description of the steps to reach that output – but is this final product truly a demonstration of learning? When tasks are designed with a focus on the final product, they open themselves up to issues with academic integrity – if you want an essay on any topic, ChatGPT can give it to you – but it can’t provide  critical reflections on learning with examples drawn from personal experience; nor can it create drafts of work with annotations that explain how it could be improved, how thinking has evolved over time or how the student felt at different stages of the process. These are true demonstrations of learning, and when assessment is built into the inquiry process, rather than added at the end, learning is more meaningful and so is assessment, as it helps students understand how they are developing and what they need to do next. The TL is poised to support teachers to collaboratively develop units of work that do just that.

Generative AI platforms have many implications for teaching and learning, and this will be a moving feast – it is very early days in the development of these tools. The whole reason so many generative AI platforms are currently freely available is so that they can continue to learn – the more humans engage with them, the better they will become. Such is the nature of machine learning. What I believe is the most important thing for TLs is to be open to how these platforms operate, to look for ways they can educate with and about these tools and to resist knee-jerk reactions to block or ban. While the focus at the moment is heavily directed toward ChatGPT, it is necessary to remember that this is just one platform – generative AI is going to continue to develop and already underpins many different tools, some of which we use without a second thought.

As expected, tools have now emerged that attempt to identify AI created text – ChatGPTZero is one of the most well known currently, but others will follow. In fact, ChatGPT creators OpenAI claim to have a protype themselves which will watermark AI generated content. While I agree completely that it is unethical to generate text through AI and claim it as your own thinking, and believe strongly in the need for academic integrity, I have always felt that rather than try to ‘catch’ those who replicate text, it is better to design learning opportunities where the ‘answers’ cannot simply be reproduced from another source. I understand the situation is not black and white and is far from simple; however to design more effective learning tasks and assessments that personalise the response to the student or that capture the learning as it develops through the entire process seems far better than simply placing our hopes in tools to compete against technology, in a race they are unlikely to win.

Educators cannot swim against the tide by pretending these capabilities do not exist. They do, and they will only get better. TLs can be (and in many cases, already are) at the forefront of establishing new ways of working that acknowledge this reality. It really is an exciting (and challenging) time to be involved with learning and teaching!

The librarian urge to curate is strong, and along with the work shared below, I’d like to share my Flipboard collection of articles on ChatGPT and AI in Education. This resource will continue to grow, and has a focus on articles about the implications of generative AI in general and ChatGPT specifically.

Click the image to access the Flipboard curated collection.

Before I close, I’d like to acknowledge the amazing work being generously shared by fellow TLs and educators in this new area. I’ve been honestly overwhelmed by how much creativity, innovation and sharing that has been happening. Here are just some of the best collections of resources – and if you know of others, please share in the comments!

5 reasons why School Librarians are Essential for Teaching Research Skills – a ChatGPT experiment by School Library Consultant Elizabeth Hutchinson

AI, chatbots, ChatGPT and School Libraries – a terrific post that not only includes a super list of highly relevant resources but also shares some more thoughts about TLs and AI by Madison Dearnaley

(If You) USEME-AI – a draft model for adapting to AI in Schools – an incredible early post that presents a structure for discussing AI in school education, and many interesting resources by Stephen Taylor

ChatGPT, chatbots and AI in Education – a massive post for educators, with sections on implications for the classroom, ways to use AI tools, should we block/ban? and resources by the Matt Miller of Ditch that Textbook

ChatGPT and Education – a comprehensive slide deck, generously CC licenced from Torrey Trust

*Disclaimer – this post was written entirely by myself in the ‘old fashioned’ way :). I do believe that one change this technology brings is the need to acknowledge when and how it is used in the construction of texts, and to establish referencing solutions to meet this need :). See these Syllabus Resources (designed for higher education) for an idea of what I’m referring to.