Assessing AI: A Report Card for Artificial Intelligence in Australian Education
Written by Grok Academy
Permission granted. Just as teachers were breaking for the holidays in December 2023, the Australian Framework for Generative Artificial Intelligence (AI) in Schools [1] was released, lifting the blanket ban of ChatGPT in public schools to give way to an enabling policy that empowers educators to bring AI into their classrooms.
Globally, TeachAI was a leader in this space of policy, producing the AI Guidance for Schools toolkit [2], with our very own Grok Academy serving on the advisory committee. However, on Australian soil appreciation should be attributed to the work of South Australia’s Department of Education, whose bold move in 2023 was not to ban AI but to instead trial EdChat, a generative AI chatbot. The chatbot addresses the many safety concerns of other AI, playing a pivotal role in shaping the Australian Framework. It provides the experience of a real chatbot: students can input prompts to access information on topics, breakdown tasks, and receive creative suggestions, all within a secure online environment. Max Leong, Team Lead (APAC) of Insight Enterprises who led the development of the chatbot explains, “EdChat harnesses the power of AI to personalise learning, creating innovative pathways for students and teachers to thrive in the digital age.”
In February 2024 NSW followed suit with the NSWEduChat, initially piloted in 16 public schools. Internal benchmarks indicate that the app outperforms the two most popular free generative AI tools. However, these same benchmarks also highlight that subjects like Mathematics Extension 1 & 2, Chemistry, and Physics pose challenges for the app. [3]
Notably, many private and Catholic schools across Australia have already been integrating genAI (short for generative artificial intelligence) into their classrooms for over a year, benefiting from their use as the ban did not extend to these sectors. However, Anna Howarth, the director of strategy at the National Catholic Education Commission (NCEC), points out that concerns persist, which extend beyond privacy, security, safety, and academic integrity, to include “the existential risks to human relationships, human connection, and human creativity.” [4]
The Digital Technologies subject has adapted seamlessly to the emergence of genAI, underpinned as it is by a set of enduring concepts, which afford opportunities to respond to changing technological and society contexts. Paula Christophersen, a co-writer of the Australian Curriculum: Digital Technologies and Grok Academy's Curriculum Chapter Lead, explains, “we took an agnostic approach to technologies when writing the Australian Curriculum: Digital Technologies, which has allowed for the complementary inclusion of new AI elaborations.’ For instance, within Digital Technologies for Years 7 and 8 alone, there are now references to AI in eight content descriptions.
The early trials of genAI in education in Australia and abroad have (disclaimer: for now) dispelled many misconceptions, while validating others. Let’s look at a few.
1 – Students will no longer have to think hard.
According to Martin Westwell, Chief Executive of the SA Department for Education [5], the results have been quite to the contrary. Teachers using SA’s EdChat have observed students engaging in deeper thinking and developing critical thinking skills. These teachers reported that learners began to interrogate the AI outputs by cross referencing this information thus showing increased meta-cognitive awareness.
DEBUNKED
2 - Students will no longer need teachers.
In an AI-integrated classroom, teachers become orchestrators of, and motivators for learning [6]. According to Emeritus Professor of Artificial Intelligence Benedict du Boulay, the essence of human-to-human interaction, particularly in the complex area of motivation, is irreplaceable. Mutual respect is a key ingredient, and learners cannot genuinely feel respect toward a piece of technology. Whilst genAI is an excellent tool, student success is still dependent on the relationship between teacher and student.
In SA, the EdChat tool freed teachers to interact more meaningfully and purposefully with learners. These teachers were able to provide more 1:1 support and engage in deep discussions with students about their learning [7]. In short, genAI has not replaced the teacher.
DEBUNKED
3 - There will be a lot more cheating.
Although it is early days, the data collected by Stanford Graduate School of Education through student surveys suggests a slight decrease in the frequency of cheating [8], yet anecdotally it appears on the rise. What might explain these results? Aside from methodological concerns [9], when it comes to genAI in take-home assessments students might genuinely be oblivious to their transgressions. Indeed, the question of what constitutes plagiarism is currently a grey area that is not limited to the field of education.
Therefore, we begin to question. Are we witnessing a shift in the definition of academic integrity, rather than unreported results?
VERDICT PENDING
4 - AI will increase the digital divide.
Sadly, we can see evidence of this already in Khanamigo, Khan Academy’s AI tutor. Although this software is already showing a positive impact on student learning, according to founder, Sal Khan [10], the non-profit’s AI tutor is not free, potentially limiting access for students from low socio-economic backgrounds.
Closer to home, despite our best efforts with initiatives such as SA’s and NSW’s purpose-built educational AI’s, the bubble of equal access bursts the moment students return home. Some will have access to the free generative AI tools readily available on the web; however, others may be able to enjoy the benefits of the significantly superior tools of paid subscriptions. Students without a device or an internet connection are further disadvantaged.
What has been confirmed is that the introduction of AI tools compounds the existing inequality, deepening disparities in educational opportunities.
VALIDATED
5 - AI will eliminate the need to learn to code.
No, at least not yet. Based on the usage data of GitHub Copilot, a popular AI-coding tool, more than 50% of the code generated by users is created by AI [11]. Empirical data from the company even suggests that developers using this tool can complete tasks 55% faster [12]. However, the effectiveness of AI-coding tools like Copilot largely depends on your existing programming knowledge. According to Thomas Dohmke, CEO of GitHub, the takeaway from the developer community is that it should be used as a pair-programmer copilot, as it cannot operate without human supervision [13]
Data scientist and former computer science lecturer Bryn Jeffries warns that novices are drawn to discipline-specific AI tools like GitHub Copilot because they represent a shortcut. As a result, they may devote a lot of effort in surface learning rather than taking a deep learning approach to coding. Bryn states “failing to invest sufficient time in mastering the skills these tools aim to accelerate could result in a future community capable of using them but lacking the expertise to do so effectively, therefore not realising the full potential of these AIs.” Indeed, according to one study, there is already evidence that greater dependence on AI is leading to poorer code quality [14].
The time may come where tools like GitHub and other discipline-specific AIs reach incredible accuracy, but regardless, won’t humanity always need their super experts?
As educators, it remains our responsibility to kindle curiosity and light that fuse for the future super experts within our classrooms who will carry the torch for the various disciplines in our society.
DEBUNKED
Reflecting on Grok’s own learning
Just like everyone else, at Grok Academy we have been placed back into the learner’s seat as we engage with this fast-evolving technology. Whilst most past technologies have been highly predictable, Sanjin Dedic, a member of Grok Academy’s AI working team reflects that “genAI is like the ocean, full of waves and winds. As you engage with genAI, you learn to sail on top of the waves and through the wind. You get to your destination most of the time, but it has these unpredictable elements, just like nature.”
Sanjin articulates that engaging in co-design alongside genAI is an enlightening journey, wherein the breadth and depth of knowledge gained surpasses anticipation. The learning process defies predictability, offering a dynamic and transformative experience that continually surprises and enriches participants. Learning with AI is inherently different from the way we traditionally design learning tasks where outcomes are often predetermined. Although Sanjin acknowledged the unpredictability in all learning tasks, acknowledging some students go on learning tangents even with traditional methods, he highlights that with a technology like this, the tangents and unpredictability of what will get learned along the way are much more prevalent. Sanjin promises “it’s going to be an enlightening learning experience.”
What might be the implications for the future of AI in Education?
As educators, are we ready to relinquish our traditional stronghold on learning outcomes, liberating student agency to foster transformative learning experiences? What essential skills, mindsets, or dispositions must we cultivate in students to unlock their full potential in the realm of AI-assisted learning? Could the adaptable minds of new generations evolve to grasp learning with AI in incomprehensible ways, giving rise to innovative educational paradigms?
In our quest to understand how humans naturally interact with AI and how AI reciprocates, Bryn Jeffries aptly draws a comparison to Google's personalised online experience. Will AI evolve to tailor itself to individuals, adeptly discerning unique learning styles and preferences? Might AI adapt to humans, negating the necessity for us to acquire new learning skills? And if so, what role remains for human educators?
Professionals in every field are likely pondering a similar question, and to add salt to the wound of our existential crisis is the news that big bucks are now being poured into the pursuit of artificial general intelligence – machines that could potentially surpass human intelligence.
To address this collective dilemma, we all need to keep looping back to one central question: What are the unique strengths and capacities that define us as humans?
As educators entrusted with the education of this collective, the time has come to ponder: What might an education that prioritises the very essence of our humanity look like?
[1] Australian Framework for Generative Artificial Intelligence (AI) in Schools (2023)
[2] AI Guidance for Schools toolkit (2023)
[3] NSWEduChat -Performance and Benchmarking (2024)
[5] & [7] Microsoft News Center (2023)
[6] du Boulay, Benedict. "The Many Roles of Human Teachers within AI in Education." Keynote Address, AIED 2023, University of Sussex, UK, July 2023.
[8] What do AI chatbots really mean for students and cheating? (2023)
[9] Nguyen, H.M., Goto, D. Unmasking academic cheating behavior in the artificial intelligence era: Evidence from Vietnamese undergraduates. Educ Inf Technol (2024). https://doi.org/10.1007/s10639-024-12495-4
[10] Meet ‘Khanmigo’, Khan Academy’s AI assistant (2023)
[11] & [13] AI Is Rewiring Coders’ Brains. Yours May Be Next (2024)
[12] Measuring GitHub Copilot’s Impact on Productivity (2024)
[14] Coding on Copilot: 2023 Data Suggests Downward Pressure on Code Quality (2024)
* Generative AI (ChatGPT3.5, Copilot, and Gemini) was used in the writing of this article to refine draft sentences written by the author (to achieve higher clarity of expression, detect errors in grammar and punctuation, or reduce length). The output was then readjusted by the author to align with intentions, surrounding context and personal style.
For more information and to register for for the Grok Stream: Building Digitally Capable Learners
Melbourne: https://www.nationaleducationsummit.com.au/melbourne/grok-stream
Brisbane: https://www.nationaleducationsummit.com.au/brisbane/grok-stream
