To Use or Not to Use

Calvin Cecil is a junior double-majoring in Math and English. For many, concerns about water usage, governance, job losses, and a million other things are reason enough not to use AI. However, in Calvin's view the best way to make an impact on a technology is to understand it. For him, that meant a summer in Maine with the Hastings Initiative.

---

I have spent the summer exploring AI. Through projects with professors, collaborations with peers, and personal investigations, I've been amazed by how far this tool can be pushed given the right instructions and context. In the absence of classes, I've been able to commit myself to better understanding such a controversial and mysterious technology. And from this I feel I have a newfound respect for both the potential and risks of the technology.

As this summer ends, I have begun thinking about how I will apply what I've learned about AI to the academic year. My AI usage largely falls in three categories: academics, career preparation, and extracurricular projects. Although each deserves their own scrutiny, I generally accept AI for both career preparation (editing resume, practicing for interviews, etc.) and extracurricular projects. Academics, however, remain the area I believe deserves a second look.

As could be expected, there is little consensus in how AI is approached in education. Some professors encourage using AI tools to help brainstorm a paper; others prevent it. Many don't discuss it. Some professors allow for AI in creating study guides or to answer questions. Using AI to help assist with homework is almost always frowned upon, but using AI to copyedit is almost always allowed. So where is the line?

To (attempt) to answer this question, I'll lead with a few lessons I've taken away from these last eight weeks and beyond.

1. AI usage can be a slippery slope. An LLM's urge to be helpful can push it past the role originally assigned. This summer, I've found that an LLM often ends up drafting an entire solution when I just ask it to help me brainstorm a project. At other times, falling into this trap can be my own doing. I've seen this when using AI to help me understand a confusing concept. This starts genuine and innocent enough, yet due to the ambiguity of what the "concept" is or the appeal of a further shortcut, the LLM often ends walking me though the homework problem where the "concept" originated from. And many times, due to the hours of back and forth and my attempts to really work through the problem, it becomes difficult for me to distinguish between what I came up with and what the AI told me. In the end, even with the best of intentions and desire to be as academically rigorous as possible, AI can subtly pull one into overreliance.
2. AI can become a "crutch" that disincentivizes human help. Simply knowing that I have an LLM as an option makes me less likely to attend office hours or study with peers. Establishing a rule that prevents me from utilizing this tool is crucial in forcing myself to be proactive in seeking this human help. For me, this can mean completely eliminating AI in certain classes (such as English) or projects (such as this article).
3. It takes time to learn how to utilize AI, but far more to become an effective driver of it. Over these last eight weeks, I have begun to develop a fairly comprehensive grasp of AI and its use cases across a variety of contexts. While it has taken some time, there are multitude of resources online to help (including the LLMs). The real challenge is having the creativity, intuition, and criticalness to best drive the technology; skills I believe are best developed in the classroom, conversations with peers, in a job, etc.
4. There is no one size fits all solution. I have found any one sentence answer to the question of AI usage (e.g. where AI enhances critical thinking without replacing it) to be overly reductive and vague to the point of unhelpfulness. A more accurate answer considers more context and priorities. It necessitates forethought and means examining each use case and its potential effects. It means imposing strict rules so the "crutch" effect isn't in the back of my mind. And most of all, it requires having a clear and specific picture of what my goals are. For me, these rules and intuition have come mostly through conversations with peers, mentors, and professors; as well as individual reflections on AI's role in my academics. While this role will look different for everyone, having an honest conversation with oneself about what an ideal level of AI involvement will look like is a critical beginning step.

Ultimately, AI use in education is undeniably an open question.  And while I’ve found it to be counterproductive to many of my goals, I do think it is important for people to engage critically with the tool and honestly examine their usage. AI is an incredibly contentious and dangerous technology, but the more level-headed people with a desire to understand the technology in the conversation, the better.