🧑🏫 The Future of ChatGPT in Classrooms
Explore how students and educators are utilizing ChatGPT within and outside of the classroom.
Evaluating the credibility and biases of algorithmic outputs requires the very higher-order skills educators aim to cultivate - questioning sources, identifying misinformation, and considering multiple perspectives. Rather than memorizing textbook facts that AI can recite, students have opportunities to develop vital creativity, collaboration, and communication abilities. How might AI influence the development of students' critical thinking abilities in the classroom?
Here is an overview of today’s email:
Recent statistics on the usage of ChatGPT among teenagers in the US
Maximizing the power of ChatGPT as a writing and editing tool
Concerns about AI tool accessibility stemming from disparities in income levels
If you have any resources you would like to share with our community of 4.8k+ educators in our future newsletters, please submit them below!
🚀 Practical AI Usage and Policies
According to a recent survey conducted by the Pew Research Center, roughly one in five US teens ages 13 to 17 say they have used ChatGPT to assist them with their schoolwork. Usage skews higher for upper-grade levels - with around 25% of 11th and 12th graders utilizing ChatGPT, compared to 17% among 9th and 10th graders, and 12% among 7th and 8th graders. While two-thirds of US teens reported they were aware of ChatGPT, awareness levels varied based on race, ethnicity, and household income. Despite its increasing popularity, teens themselves have expressed uncertainty regarding whether ChatGPT should or shouldn’t be employed in specific educational contexts. For further details on this research, check out the source here: https://www.pewresearch.org/short-reads/2023/11/16/about-1-in-5-us-teens-whove-heard-of-chatgpt-have-used-it-for-schoolwork/
Instead of banning ChatGPT, some teachers have decided to embrace it in classrooms. The guide, "An Introduction to Teaching with Text Generation Technologies", by Tim Laquintano, Carly Schnitzler, and Annette Vee from TextGenEd provides a robust set of resources for instructors looking to incorporate AI text generators into their curriculum: https://wac.colostate.edu/repository/collections/textgened/
This resource includes 34 undergraduate-level assignments across a range of areas - “AI literacy, rhetorical and ethical engagements, creative exploration, and professional writing text gen technology”. Alongside these ready-made assignments, the authors have also included practical content to assist educators in evaluating, selecting, and emphasizing the most relevant learning goals for their specific courses and student needs when adopting new AI tools.
However, a major challenge in developing guidelines for AI usage is that the rapid pace of advancement makes it difficult to establish regulations that are simultaneously forward-thinking yet responsive to emerging capabilities. In fact, this past month, OpenAI shared some of their latest advancements of ChatGPT in their opening keynote at their first developer conference. One of its newest features is the ability for users to build their own GPTs without needing a single line of code. Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, recently shared his process of trying out this feature and creating his own GPT to provide customized feedback on student writing. He discussed the limitations and potential for these personalized GPT “agents” in automating tasks for work and school:
The ability to build one's own GPT opens the door to a wave of personalized AI applications. As models like GPT-3 become more accessible, we can see the role of customized chatbots, writing assistants, tutors, and advisors in equipping students with personalized learning resources. While it has become easier now to create your own GPT, it is important to take factors like safety, accuracy, and quality of output into consideration when building a chatbot. For instance, a recent blog from Khan Academy shares some of the considerations they had when developing their AI-powered tutor, Khanmigo: https://blog.khanacademy.org/prompt-engineering-using-ai-for-effective-lesson-planning/
As AI tools continue to advance in their capabilities, a critical priority will be ongoing collaboration and open dialogue between all stakeholders - ranging from policymakers and platform developers to instructors and students.
📝 Latest Research in AI + Education
Stanford Accelerator for Learning
What do AI chatbots really mean for students and cheating? ↗️
The article discusses the impact of AI chatbots, like ChatGPT, on student cheating. Stanford Accelerator for Learning's Victor Lee and Denise Pope argued that concerns about AI facilitating cheating are overblown. Their studies reveal that cheating rates among U.S. high school students have been high for years and have not significantly increased with the advent of AI tools. They highlight that cheating is more a symptom of systemic issues in education, such as overwhelming pressure and lack of engagement, rather than a direct result of technology access. The article also emphasizes the importance of teaching AI literacy and using AI responsibly, rather than focusing solely on the detection and prevention of AI-assisted cheating. They suggest integrating discussions about the ethical use of AI into the classroom to help students understand and critically engage with the technology.
Spector, Carrie. “What Do AI Chatbots Really Mean for Students and Cheating? .” Stanford Accelerator for Learning, 1 Nov. 2023, acceleratelearning.stanford.edu/story/what-do-ai-chatbots-really-mean-for-students-and-cheating/. International Journal of Educational Technology in Higher Education
This paper examines the use of generative artificial intelligence (GenAI), like ChatGPT, in academic writing processes from the perspectives of university students and educators. A survey involving 68 educators and 158 students assessed their views on GenAI's role in different writing tasks such as brainstorming, outlining, writing, revising, feedback, and evaluation. The findings indicate a general agreement between students and teachers on the appropriate use of GenAI, with both groups finding it more acceptable for early-stage tasks like brainstorming and outlining than for complete writing. However, minor disagreements were noted, particularly regarding GenAI's use without disclosure. The study highlights the lack of preparedness at classroom and institutional levels for GenAI integration, emphasizing the need for explicit guidelines and teacher training on ethical GenAI use. It suggests that both educators and students see potential utility in GenAI but express concerns over its impact, calling for transparent and ethical use of these tools in education. The study acknowledges limitations like its non-random sample and the evolving nature of technology and perceptions, suggesting future research to explore these dynamics further.
Barrett, Alex, and Austin Pack. "Not Quite Eye to A.I.: Student and Teacher Perspectives on the Use of Generative Artificial Intelligence in the Writing Process." International Journal of Educational Technology in Higher Education, vol. 20, no. 1, 2023, pp. 1-24, https://doi.org/10.1186/s41239-023-00427-0. Accessed 26 Nov. 2023.📣 Student Voices and Use Cases
We interviewed students to learn more about their experience using AI in schools. This week, we had a chance to speak with Ami Patel, a Master’s student at the University of Toronto studying Psychology and Education. In the following, we present select highlights from these conversations, which have been slightly edited for enhanced clarity and precision:
Q: In your opinion, what are the key benefits of integrating AI into educational settings?
From a student perspective, integrating AI makes things a lot easier. While a lot of educators fear that students won’t do their own work or write their own essays, I don’t think that AI is quite advanced enough to do their work for them. Currently, I use ChatGPT to help me come up with a first draft or an outline but it doesn’t create the entire content of the essay for me. I also use it as a creativity tool, for instance, to generate the title of a presentation. Additionally, I think that AI can help us save time as students. I used to spend hours writing flashcards, but AI tools save me time by quickly generating questions and flashcards from slides and content to help with studying.
Q: Can you share an example of how AI has positively impacted your learning experience?
This term, I am taking a course called “Child Development and Personal History” where we have to design and run a survey-based research study from start to finish in 2 months, which is really intense. From the beginning, we were encouraged to use ChatGPT and other AI tools to brainstorm research questions and identify research gaps. We were able to use ChatGPT to help us formulate questions and design survey items. I was able to formulate hundreds of survey items that I wouldn’t be able to do given our limited human capabilities. While I didn’t end up using all of the generated questions, the process of creating so many items provided me with insights for coming up with new questions. I was also able to use ChatGPT to help me understand different ways I can analyze my final data. For instance, it would suggest to look at how gender plays a role in the results and relationships. At the end of the survey, we had to create our final presentation, and I also had a chance to utilize AI to ideate creative presentation titles.
Q: Have you noticed any challenges or concerns with AI implementation in education? If so, what are they?
I think a major one is the problem of accessibility. ChatGPT’s free version is outdated, with information only up to 2021. The paid version, on the other hand, provides more updated information and helpful tools. As a result, students who are more privileged and are able to afford these paid AI tools have a higher leg up in terms of learning resources.
I also know that some schools offer free access to Grammarly to all their students, but for schools that don’t, the students who do have personal access to these tools will have a higher leg up compared to those who cannot afford these tools in terms of the quality of their written work. At the end of the day, essays get graded on clarity and grammar, so access to these tools can enhance one’s writing and result in better grades. I believe that it is important to consider how we can create an equitable playing field in education regardless of financial status.
Another issue I’ve seen is the use of AI screening tools, especially for admissions. One of my classmates mentioned the recent use of AI screening tools in medical school admissions, and this was alarming because of the possible biases that may be embedded within these systems. AI screening tools raise concerns about its inherent biases.
Q: How do you envision the role of teachers changing with the increasing use of AI in classrooms?
Here in my province in Canada, we’ve been experiencing an issue with increasing class sizes among primary and secondary schools. This can lead to classrooms having too many students for one teacher to handle. Students may not get enough time and attention, causing some to fall behind. With AI in classrooms, this may allow teachers to spend more time focusing on teaching and less so answering little individual questions. AI can also help reduce the load on teachers and enable them to focus on more important things. Ultimately, I don’t think that AI can completely replace teachers. Technology can’t really change the social aspect of the teacher, being there for students and encouraging them. Interpersonal skills are especially important for students’ development, and teachers play an important role in this.
📰 In the News
The New York Times
Chatbots May ‘Hallucinate’ More Often Than Many Realize↗️
Key takeaways:
Chatbot Hallucinations: Research by Vectara, a new start-up, shows that chatbots like ChatGPT and Google's systems "hallucinate" or invent information, with rates ranging from at least 3% for OpenAI to 27% for Google.
Testing Methodology: Vectara tested chatbot accuracy by asking them to summarize news articles, a task where they often introduced errors, indicating a fundamental problem in AI summarization capabilities.
Industry-Wide Issue: The study reveals varying hallucination rates among major AI companies: OpenAI at 3%, Meta at around 5%, Anthropic's Claude 2 system at 8%, and Google's Palm chat at 27%.
Consequences and Responses: These inaccuracies pose risks for uses involving sensitive information like court documents or medical data. Despite efforts by companies to minimize these issues, the complete elimination of such errors remains uncertain.
Research Implications: Vectara's research, conducted by former Google employees, aims to raise awareness and encourage industry-wide efforts to reduce hallucinations in chatbot technology, though the challenge persists due to the probabilistic nature of these AI systems.
Nature Journal
ChatGPT has entered the classroom: how LLMs could transform education ↗️
Key takeaways:
Educational Integration of ChatGPT: The article discusses how large language models (LLMs) like ChatGPT are being explored in education. Educational psychologist Ronald Beghetto created creativity-focused chatbots at Arizona State University (ASU) that encourage students and lecturers to think creatively, showing positive feedback in enhancing creativity and generating more possibilities.
Potential and Challenges: ChatGPT's potential as a personalized, conversational educational tool is highlighted. It could serve as a cost-effective, always available 'thought partner', augmenting traditional tutoring methods. However, there are concerns about its reliability, with risks of misinformation and dependency on quick answers without understanding.
Concerns About Misuse and Privacy: The article raises concerns about students using ChatGPT to cheat on assignments and the privacy implications of student data being stored by OpenAI. It emphasizes the need for understanding ChatGPT's strengths and risks.
Innovations and Experiments in Education: Various initiatives are mentioned, like Khanmigo by Khan Academy, which uses GPT-4 to assist students, and ASU's experiments with RAG (retrieval-augmented generation) to minimize errors and hallucinations. There's an emphasis on ensuring the AI doesn't insult or belittle students and maintains a supportive tone.
Future Prospects and Ethical Considerations: The article concludes with the uncertain future of AI in education. It highlights the need to address the digital divide, ensure unbiased information, and reconsider teaching and assessment methods in light of AI advancements. The importance of understanding AI's limitations while embracing its potential is stressed.
“Chatgpt.” ChatGPT, OpenAI (GPT-4), openai.com/chatgpt. Accessed 28 Nov. 2023. Based on the results from our previous newsletter poll, the two AI usage tools that educators are most interested in learning about include educational chatbots and curriculum development. We hope that this newsletter provided a deep dive into chatbots and we look forward to continuing to deliver relevant content catered to you!
And that’s a wrap! If you enjoyed our newsletter and found it helpful, please consider sharing this free resource with your colleagues, educators, administrators, and more.
I liked this piece from Ben Thompson (Stratechery) suggesting that our role as humans, and students as a derivative, in the era of generative AI is to become editors instead of memorization machines, which is to a large extent the premise of the old/current education system stratechery.com/2022/ai-homework