Editorial
Working in education in the last two years has been interesting to say the least, as generative AI’s emergence has caused severe disruption. Inevitably, substantial progress in integrating generative AI into education at all levels is far from rapid. The systems that govern education, in most of the world, are far from contemporary. They tend to lack flexibility and urgency as well as being very risk averse, not to mention the reality that many facets of education are political tools.
Despite the issues with education systems and societal problems, this year we are starting to see a more measured reaction to generative AI than the focus on plagiarism and the death of thinking. The pockets of experimentation have and continue to grow. Blanket bans are being lifted in many places, and policies that outline when, how and where AI can be used in education are being implemented in institutions. It could be even said that the excitement felt by the early adopters of generative in education is being shared more widely. Professional development sessions that provide information and examples of how this technology can be used are being accessed by educators across the world. Many of these will showcase particular tools, highlighting use in enhancing teaching, learning and assessment. Thus, the popularity of certain tools get passed on from early adopters to greater audiences within education.
Freemium Today, Expensive Tomorrow
The freemium business model is a well-established practice in the technology industry. It reduces the initial barriers to entry for new users, allowing them to test a limited version of the product without any financial obligation. This approach can help increase a company's total customer base, as some users may eventually upgrade to the fully paid version after trying out the free offering. Freemium and seemingly free-to-use tools are generally where most educators are being trained or are experimenting. It could easily be argued that this is no surprise.
However, an emerging trend in respect of generative AI software development is for many freemium or free-to-use tools to quickly and without warning, introduce pricing plans. These plans are often relatively expensive or without education discounts. In other words, freemium is exactly what it promised — only free for a limited time, and free to use is a misnomer, as that is also free for a limited time.
Sudden costs or price hikes cause issues in education. Increasingly, there are examples of where teachers and students have invested time, built up practices and materials around particular tools after experimenting/professional development, but these tools are suddenly out of their reach. Teachers are not accustomed to paying for software and paid-for digital tools used in education systems are often paid for by the institutions or at a higher level across states, boards, authorities and the like. Typically, the pricing for education software is often discounted for volume and designed with the nuances of teacher and student accounts, provision of dashboards, issues, such as data privacy, and facilitates easy account management. Many generative AI tools do not provide for these factors.
The Impact on Primary and High Schools
Arguably, one of the most difficult areas of education to get traction with generative AI is in the primary sector. Most generative AI tools have a lower age limit of 13 in line with social media. Tools that can be used with students under 13 years old are few and far between. Practices are being developed with tools that limit under 13s having accounts where teachers are demonstrating the software. The teachers are using and displaying output to students. Young learners are offering ideas for input, such as prompts or working with output from the teacher's prompts in tools, such as image generators. There are a few tools that permit under 13s access to generative AI, but some of these are developed on the freemium or free for a limited-time model. Hence, when price hikes occur, the negative impacts can seem greater.
Such issues are not confined to the primary school sector, though. They also impact high schools, further and higher education. While it is understandable that generative AI developers want to recoup costs, and the costs can be high when utilizing LLMs, the rate at which the change from free to use or freemium to high cost happens is unlike other areas of software development. It is not how the market operated during the web 2.0 era or since. Freemium tended to last longer or limited-time offers were clearly indicated. Likewise, information was often provided that price changes were going to happen, and provision for education was often a part of the process. Generative AI development seems to be different and perhaps fueled by the culture of rapid and iterative development.
Issues in Scaling AI in the Classroom
Without education pricing, in other words discounted pricing for volume, most generative AI tools are limited in their appeal to educators. It can become very expensive to have whole classes or courses using or reliant on these tools. Having features that provide for managing students accounts are also barriers to education use at scale.
Already, issues in working with popular chatbots, such as ChatGPT, Gemini and Claude, exist, where free versions work on models that perform at a lower standard or only allow limited use on free versions. Paid-for versions where hallucination is less, and thereby, accuracy and reliance on output are greater, features, such as fine-tuning, custom chatbot design and substantial file uploads, mean that educators are not able to truly work with the capabilities of these tools without incurring costs. Questions have to be asked as to whether such factors are hindering developments and adoption of the technology in education.
There are also other signs that cost issues may be about to become exacerbated for the most popular and commonly used generative AI tools. Some reports are suggesting that major providers, such as ChatGPT, may be significantly increasing their subscription costs or widening levels of performance with the high-performing models given luxury pricing. Another significant cost comes in areas such as protection of data. An example is Microsoft Azure, which offers institutions enterprise accounts where data is not shared with LLMs.These are attractive propositions for those in education, given the sensitivity of student data but such provision is expensive, and there are few signs that this will change.
Conclusion: A Call for Sustainable AI in Education
This article outlines how the unpredictable pricing models create instability in education systems. Having such instability leads to trust issues among educators, authentic issues with logistics in scaling use to learners and to larger groups of teachers. The continuation of such issues can only hamper developments, experimentation and the long-term integration of generative AI in education.
Perhaps there needs to be a greater understanding that if education use of generative is seen as both a major tech concern both fiscally (education use of chatbots is a substantial market) and from a humanitarian perspective, there needs to be more thought put into catering for the user needs.
Learn how you can join our contributor community.
