In the article “Imagine how great universities could be without all those human Teachers” discuss how technology could be used in the future of education. They explore the possibilities of virtual professors and TAs. They specifically use the example of “Jill”, an AI teaching assistant at Georgia Tech who is considered “the best damn teaching assistant you could ever want” (Schrager and Wang, 2017). The professor who created her, Ashok Goel, was feeling incredibly overwhelmed by the sheer amount of inquiries and work he would get from students. The class had 300 students and already had 8 TAs with Jill becoming the ninth. Originally, students did not know they were talking to AI.
The company Pearson also has an AI tutor that they are working on. It is designed to quiz students, can pick up on when students are talking about concepts incorrectly, and will possibly teach in the future. They are also working on software that can grade written answers instead of just multiple choices. This is one example of technology that companies are working on, but the boundaries still need to be explored.
The authors also discuss the cost disease. The cost of universities keeps increasing while it is a field that cannot be automated through technology, thus has a lot of expenses through wages. The authors suggest that through AI taking on the more “tedious” parts of teaching, that universities can hire less professors and TAs, saving them money. But they point out that is where the problem starts. When does there become too little personal interaction that there is no inspiration?
A 2025 journal did a survey of nine research papers to find the effects of AI in schools (Klimova & Pikhart, 2025). They found that a common pro is being that they could tailor individual lessons to students, which can create less stress and better mental health for students. A common con, however, is as the authors of the other article stated that there can become an over-reliance on AI that can lead to negative impacts on social and interpersonal skills. Many also have anxiety over not understanding or knowing how to use AI as well as security and data privacy.
This also brings in anxiety about job displacement, especially in education. Another 2025 journal looked at generative AI (GenAI) and its impact on jobs in education (Joshi, 2025). There were few findings in regard to job displacement. First, low-skill jobs are most likely to be displaced while high-skill jobs are most likely to benefit from GenAI. Second, the statistical predictions indicate significant changes in the workforce, with many being displaced, but with many new job opportunities being created.
However, at the root of this is it worth the cost? The article addresses the cost disease. If AI is added to the education field, those who are still educators have to learn to use it and work with it. How much will it cost to implement and train them? If new job opportunities are created because of AI displacing certain jobs, what will the training requirements be? Will they be required to get another certification or degree? How much will that cost? It turns into a cycle of putting money into the next new technology without addressing if it is worth it or if it is actually effective. We are still unsure if AI will actually help education or if it is just another tool that will go out of fashion or cause harm.
A Stanford Scale Initiative report discusses the research on AI and the impact (Stanford Scale Initiative, 2026). It had three key findings for students. First, some AI tools increased performance, but once access was taken, results were mixed. Second, AI tools can alleviate mental burden and help with a positive learning environment, but may not help with deeper thinking skills. Third, the design of AI chatbots matters as tutor ones are much more helpful than general use ones.
For educators, four key findings were found. First, AI can reduce repetitive tasks without evidence of quality loss. Second, AI can give automated feedback to improve instruction and outcomes. Third, AI can give real-time feedback, particularly in message based settings. Fourth, AI pedagogical support seems to be most beneficial to less experienced and lower rated educators.
The caveat of these findings is that rigorous evidence and research is still thin. Out of over 800 papers that the team reviewed, only around 20 were high-quality causal studies that rigorously studied the impact. Many also only focus on short term effects rather than long term. We do not know what using AI in education would look like in 20 years and so far, even in the short term, results are incredibly mixed. So, is it worth the cost?
As the authors of the article stated, some tools like the ones that Pearson is developing could be beneficial. However, as stated in the Stanford report, not all tools are created equally. Would this also create a resource and access issue, where certain schools could only afford the cheaper, ineffective, poorly designed AI tools? When used as Jill is to answer common questions and perhaps help with grading, it can be beneficial to both teachers and students. However, once you get into long term effects and the cost, how worth it is AI in education?
References
Joshi, S. (2025). Introduction to Generative AI: Its Impact on Jobs, Education, Work and Policy Making. https://doi.org/10.2139/ssrn.5206473
Klimova, B., & Pikhart, M. (2025). Exploring the effects of artificial intelligence on student and academic well-being in higher education: a mini-review. Frontiers in psychology, 16, 1498132. https://doi.org/10.3389/fpsyg.2025.1498132
Standford Scale Initiative. (2026, March 11). Understanding the evidence base on AI in K-12 education [Review of Understanding the evidence base on AI in K-12 education]. Scaleinitiative. https://scale.stanford.edu/research-in-action/understanding-evidence-base-ai-k12-education
Leave a Reply