Responses

University Responses

In response to the release of ChatGPT in November 2022 ( Citation: OpenAI, 2022 OpenAI, O.(2022, 12/30). Retrieved from https://openai.com/blog/chatgpt ) , several universities including Stevens ( Citation: Lipizzi & Morgan, 2023 Lipizzi, C. & Morgan, G.(2023, 1/19). Retrieved from https://www.stevens.edu/news/is-chatgpt-going-to-eat-or-save-the-world ) , Harvard ( Citation: Mucharraz y Cano, Venuti & al., 2023 Mucharraz y Cano, Y., Venuti, F. & Herreta Martinez, R.(2023, 2/1). Retrieved from https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text-generators-should-academia-adapt-or-resist ) and Stanford ( Citation: Lee, Lemons & al., 2022 Lee, V., Lemons, C., Brown, B., Lising Antonio, A. & Levine, S.(2022, 12/20). Retrieved from https://ed.stanford.edu/news/stanford-faculty-weigh-new-ai-chatbot-s-shake-learning-and-teaching ) responded with articles explaining their view on the new technology. The professors who responded have mixed feelings regarding generative AI because it allows students to complete homework assignments. However, some faculty members believe that the benefits of GPT outweigh the downsides. The following is a list of ideas common among the university responses.

Disclaimer

These responses were given within three months of the release of ChatGPT. This means that, at the time of responding to ChatGPT, the universities have little to no classroom experience.

College Admissions

The college admissions offices will now have to sift through essays written by both students and AI. The question is: Will ChatGPT change who gets accepted into competitive universities? If so, is there a group of people that this new technology will disproportionately impact?

According to Stanford ( Citation: Lee, Lemons & al., 2022 Lee, V., Lemons, C., Brown, B., Lising Antonio, A. & Levine, S.(2022, 12/20). Retrieved from https://ed.stanford.edu/news/stanford-faculty-weigh-new-ai-chatbot-s-shake-learning-and-teaching ) , the accessibility of this technology will lead wealthier people to find new ways of standing out from the crowd. Anthony Lising Antonio, associate professor of education, predicts that a “for-profit cottage industry” will develop to help people design competitive applications that stand out from the AI-generated mass.

Example (AI GPT generated essay) ↕

As my worn-out sneakers pounded the sidewalk, the rhythmic thud of my heartbeat filled my ears. Running was the metronome of my life, steady, unyielding, and more than anything, constant. The sun dipped beneath the horizon, and a sliver of moon illuminated the sky. The question, however, lingered in my mind as I ran: Can I break free from the confines of my small town and create a meaningful impact on the world?

My life began in a rural town so remote that it could have been plucked straight out of a Steinbeck novel. Growing up, I was immersed in a cultural tapestry of cornfields, pick-up trucks, and Friday night football games. While my roots are deep, I found myself yearning for more - a life beyond the finite boundaries of my hometown. My curiosity spurred me to explore, ultimately leading to an unquenchable thirst for knowledge and a deep-seated desire to make a difference.

One fateful day, I stumbled upon an old, dusty book buried beneath a pile of outdated encyclopedias at our local library. It was a collection of essays by Ralph Waldo Emerson. As I read, I became entranced by his philosophies of self-reliance and the pursuit of truth. Emerson’s words resonated within me, igniting a spark that ultimately ignited my passion for literature and a desire to explore the world through the eyes of great thinkers.

This newfound passion led me to establish a monthly book club at our local community center. I was determined to bring people together to discuss literature, exchange ideas, and inspire one another to dream big. Our motley crew of literary aficionados consisted of farmers, factory workers, and high school students alike. The first few meetings were awkward, but as the months progressed, I watched with pride as our group coalesced into a community of engaged, thoughtful citizens.

The power of literature, I discovered, transcends the borders of geography and age. It creates a common ground upon which we can challenge our beliefs, explore the depths of the human experience, and dare to envision a brighter future. It was in those moments, sharing ideas and laughter in our book club meetings, that I realized the true potential of education: to transform lives and unite people in the pursuit of knowledge.

At Harvard, I see an opportunity to delve deeper into this transformative power of learning. I hope to collaborate with my peers, engage with brilliant professors, and work together to solve the complex challenges facing our world. The hallowed halls of Harvard hold the promise of intellectual growth and the chance to develop the tools necessary to make a lasting impact on the world.

As I stride forward, my sneakers still beating a familiar rhythm against the pavement, I am filled with a sense of purpose. I am ready to embrace the challenges and opportunities that lie ahead. I am prepared to step beyond the confines of my small town and embark on the journey of a lifetime.

So here I stand, at the precipice of a new beginning, with the words of Emerson echoing in my mind: “Do not go where the path may lead, go instead where there is no path and leave a trail.” At Harvard, I am determined to blaze a trail that leads not only to personal growth but also to a brighter, more connected world for all.

Helps Teachers

ChatGPT can make teachers more effective. In February 2023, Harvard described the positive and negative impact ChatGPT would have on the classroom. Regarding the positives, they focused on how teachers can use it to create review sessions, automatically grade homework, and emulate conversation for language learning ( Citation: Mucharraz y Cano, Venuti & al., 2023 Mucharraz y Cano, Y., Venuti, F. & Herreta Martinez, R.(2023, 2/1). Retrieved from https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text-generators-should-academia-adapt-or-resist ) .

Overreliance

If a student graduates from high school or university using ChatGPT to solve their homework, the question becomes: Did the student graduate, or did ChatGPT graduate? The Harvard Business Publishing states that students can overuse generative AI and “miss out on important learning opportunities” ( Citation: Mucharraz y Cano, Venuti & al., 2023 Mucharraz y Cano, Y., Venuti, F. & Herreta Martinez, R.(2023, 2/1). Retrieved from https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text-generators-should-academia-adapt-or-resist ) .

In general, when people depend on unreliable technology, it can lead to problems. For example, I always use Google Maps when I drive, and while the technology is more reliable than my memory, my phone battery is unreliable. When my phone does, I am lost and must stop to charge or drive out of the way until I recognize a road. This example demonstrates that becoming reliant on a technology that is only sometimes available can be problematic. ChatGPT is far from reliable as it requires an internet connection, paid subscription (to have access when servers are overloaded), and often outputs fabricated facts. If tech companies can fix these problems and make language models reliable, the case against overreliance is diminished.

Be Cautious

ChatGPT is a tool, and the output of this tool can mislead students. I asked Professor Schaumann, a computer science professor at Stevens, how ChatGPT will impact education. In contrast with the articles released early this year, Professor Schaumann has had an entire semester of classroom experience with these tools. As a result, his response differed from the standard response of either trying to promote or detect cheating using ChatGPT. Instead, he focused on the danger ChatGPT has to students.

If students use ChatGPT to learn the information they do not know, how will they verify that the output of ChatGPT is correct? The problem is that ChatGPT can generate text that looks reasonable, but the facts it points out are often false. The same problem occurs when students trust the first article they read after Google searching a question. Therefore, professor Schaumann recommends that, when using Google or ChatGPT to answer questions, students ask themselves, “(a) what you’re asking, (b) if what you’re asking is actually what you need to know, (c) if what you found does actually answer your question, and (d) what assumptions or shortcuts the answer is making.”

Academic Responses

GPT For Good?

Although there are many ways that ChatGPT can be abused to get out of doing work in an educational context, if used correctly, ChatGPT leads to a better educational system ( Citation: Kasneci, Sessler & al., 2023 Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., Stadler, M., Weller, J., Kuhn, J., Kasneci, G., Chiriatti, F., et al., D., et al., T. & et al., V. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. ) . In an article submitted in February titled “ChatGPT for good? On opportunities and challenges of large language models for education”, Kasneci et al. elaborate on the potential benefits of using AI in education. Here are several domains of education that generative AI can assist.

Elementary School. ChatGPT can be used to analyze student responses to correct grammar and provide feedback on critical thinking skills.
Middle & Highschool. Used to generate learning material such as quizzes. Provides example solutions that students can follow.
College Students. Summarize text to increase learning rate. It helps students retrieve information.
Disabilities. The flexibility of AI leads to more flexible learning strategies. It can help rework topics in ways that students can better understand.

Kasneci et al. also describes how researchers used GPT3 to generate student quizzes. I thought this would be a fun experiment to recreate, so I created a GitHub repository that does that. You can view it here.

Bullshit Spewer

One of the more critical reviews of ChatGPT titled “ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?” guides how to adapt to this new technology ( Citation: Rudolph, Tan & al., 2023 Rudolph, J., Tan, S. & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?. ) . To set the stage, Jurgen et al. say that the problem arises from technology progressing faster than the educational system. In a slightly joking manner, they state that it will take…

“10 years for academia to face this new reality: two years for the students to figure out the tech, three more years for the professors to recognize that students are using the tech, and then five years for university administrators to decide what, if anything, to do about it.”

One workaround is to asses students on assignments that AI models are deficient in. This is a short-term solution as the models continue to improve, but hopefully, it will give the school enough buffer to adapt to the new technologies. The limitations of the model are the following:

Image understanding. ChatGPT cannot understand images, so assessing students based on video or images is one way to degrade the ability of students to cheat using AI tools.
Recent Events. AI models often have a data cutoff in which a model is unaware of any events that occur after that date. Assigning relating to current events is harder for students to solve with AI.
Niche Topics. While the breadth of knowledge in a GPT model is impressive, it is not limitless. The more Niche a topic is, the less likely an AI model is to generate a correct response.
Personal Experience. By assigning homework specific to a student, teachers can increase the difficulty of cheating.
Cross Referencing. ChatGPT needs help creating valid references. By assigning homework that requires students to reference different research papers, they will need more time to leverage AI.

Even if an educator uses the tips provided here, students can still find ways to bend the rules. In addition, developments in GPT4 are already showing abilities to understand images and increase contextual awareness. For example, if a student is asked to write an essay about a current event, they can paste a recent news article into GPT4, and it will understand the current event.

Bibliography

Rudolph, Tan & Tan (2023): Rudolph, J., Tan, S. & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?.
Kasneci, Sessler, Küchemann, Bannert, Dementieva, Fischer, Gasser, Groh, Günnemann, Hüllermeier, Krusche, Kutyniok, Michaeli, Nerdel, Pfeffer, Poquet, Sailer, Schmidt, Seidel, Stadler, Weller, Kuhn, Kasneci, Chiriatti, et al., et al. & et al. (2023): Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., Stadler, M., Weller, J., Kuhn, J., Kasneci, G., Chiriatti, F., et al., D., et al., T. & et al., V. (2023). ChatGPT for good? On opportunities and challenges of large language models for education.
Mucharraz y Cano, Venuti & Herreta Martinez (2023): Mucharraz y Cano, Y., Venuti, F. & Herreta Martinez, R.(2023, 2/1). Retrieved from https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text-generators-should-academia-adapt-or-resist
OpenAI (2022): OpenAI, O.(2022, 12/30). Retrieved from https://openai.com/blog/chatgpt
Lee, Lemons, Brown, Lising Antonio & Levine (2022): Lee, V., Lemons, C., Brown, B., Lising Antonio, A. & Levine, S.(2022, 12/20). Retrieved from https://ed.stanford.edu/news/stanford-faculty-weigh-new-ai-chatbot-s-shake-learning-and-teaching
Lipizzi & Morgan (2023): Lipizzi, C. & Morgan, G.(2023, 1/19). Retrieved from https://www.stevens.edu/news/is-chatgpt-going-to-eat-or-save-the-world