The Challenge of AI in Education

Without a doubt, students are using generative AI and turning it in as their own work. Not only is this a clear violation of the honor code, but students end up shortchanging themselves by not learning the skills and content we have diligently crafted in our courses. The knee-jerk reaction would be to ban the tool outright; no use of AI in my class would mean students would do the assignment as planned and have demonstrated in the past. But let me tell you, this is not a solution. Not only will students continue to use AI regardless of our restrictions, but we will shortchange the student by not teaching them appropriate use of these tools. As businesses and organizations quickly adapt AI tools in a myriad of ways, we need to prepare students for that reality by engaging with these tools in a productive and “transformative” way.

Transactional vs. Transformative AI Use

Most of our students’ use of AI has been “transactional” rather than “transformative.” They take the question or prompt you have provided, drop it into a large language model (LLM), like ChatGPT, and then copy and paste the output as their submission to you. For all of us, this is easy to spot:  strangely placed punctuation, vocabulary well out of their realm of knowledge, complex thoughts and arguments beyond their current level. 

“Transformative” use of generative AI is different. It focuses on keeping the human in the loop through this process by crafting a good prompt on the front end and modifying the output of the AI on the back end. Rather than blindly adopting whatever the AI produces, we should be using our expertise and knowledge of the context surrounding the prompt to modify the output to match our tone, voice, and intent. This paring of human and AI working together to produce excellent work is the desired outcome for my classroom and office.“Transformative” use of generative AI is different. It focuses on keeping the human in the loop through this process by crafting a good prompt on the front end and modifying the output of the AI on the back end. Rather than blindly adopting whatever the AI produces, we should be using our expertise and knowledge of the context surrounding the prompt to modify the output to match our tone, voice, and intent. This paring of human and AI working together to produce excellent work is the desired outcome for my classroom and office.
 

Practical Application: Writing a Letter of Recommendation

For faculty, writing a letter of recommendation is a great place to start leveraging these tools in a transformative way to save ourselves time and energy. I have written many of these letters over the course of my career. Their formal tone and rigid structure provide an opportunity to use generative AI to help save me to time and produce and quality letter. I invite you try this out on your own.

Step 1: Creating the Prompt

To create a new letter, I first start with creating a rough outline including the character qualities and situations I wanted to feature about this student as well as the broad details of the intended employer. I went to ChatGPT and entered in this information in sentence form with same framing context:

“Please write a letter of recommendation for a current college student majoring in Computer Information Systems who is applying for a job as a product manager for a large tech company. I am in no rush, kindly take your time and consider what makes an excellent letter of recommendation to a company of this type. He excelled at the skill of public speaking by engaging with the audience excellently, through strong delivery, use of humor, and adapting to the context of the room. He took leadership with his final group project with vision and direction and even developed video content far beyond the expectations of the project. His ability to speak to any audience with authenticity and clarity will take him far in his career.”

Step 2: AI-Generated Response

ChatGPT’s response:

"[Your Name]
[Your Position/Title]
[Your Institution/Organization]
[Your Address]
[City, State, Zip Code]
[Your Email Address]
[Your Phone Number]
[Date]
[Hiring Manager's Name]
[Company's Name]
[Company's Address]
[City, State, Zip Code]

Dear [Hiring Manager's Name],

I am writing to offer my highest recommendation for [Student's Name], who is applying for the Product Manager position at [Company's Name]. As a professor in the Computer Information Systems program at [Your Institution], I have had the pleasure of teaching [Student's Name] in several courses, including Intro to Public Speaking. Over this time, I have witnessed [his/her] remarkable growth, leadership, and unwavering dedication to excellence.

[Student's Name] possesses a unique blend of technical knowledge and outstanding communication skills, a combination that is essential for a successful Product Manager. In my public speaking class, [he/she] consistently demonstrated an exceptional ability to engage with the audience. [His/Her] presentations were marked by strong delivery, effective use of humor, and an impressive adaptability to the context of the room. [He/She] has a rare talent for making complex concepts accessible and engaging, which is critical in bridging the gap between technical teams and end-users.

One of [Student's Name]'s most impressive achievements was [his/her] leadership in the final group project. [He/She] took charge with a clear vision and provided direction that inspired the entire team. Under [his/her] guidance, the group not only met but exceeded the project requirements. [Student's Name] developed video content that was innovative and far beyond what was expected, showcasing both [his/her] technical prowess and creative ingenuity.

What truly sets [Student's Name] apart is [his/her] ability to communicate with authenticity and clarity to any audience. This skill will serve [him/her] well as a Product Manager, where clear and effective communication is key to understanding customer needs, conveying technical details to non-technical stakeholders, and driving product success.

In addition to [his/her] communication and leadership abilities, [Student's Name] has demonstrated a strong work ethic, intellectual curiosity, and a commitment to continuous improvement. [He/She] is always eager to take on new challenges and approaches every task with a positive attitude and a determination to excel.

I am confident that [Student's Name] will be an invaluable asset to your team at [Company's Name]. [His/Her] combination of technical expertise, leadership skills, and exceptional communication abilities make [him/her] the ideal candidate for the Product Manager position. I wholeheartedly endorse [his/her] application and am excited to see the positive impact [he/she] will undoubtedly have at your company.

Thank you for considering [Student's Name] for this opportunity. If you have any further questions or require additional information, please do not hesitate to contact me.

Sincerely,

[Your Name]
[Your Position/Title]
[Your Institution/Organization]"

Step 3: Human Refinement

This output gave a great head start on crafting this letter. From here, I was able to add the appropriate student and business information and modify the letter in my own tone and language. This saved me time and enabled me to respond to the student’s request in a timelier manner.

Key Elements of Effective Prompts

There are a few elements to point out in my initial prompt.

1. Specify the deliverable – A letter of recommendation is a defined document out in the known world, which an LLM more than likely can identify and create. Consider specifying how long or what format of text you are wanting the LLM to create.

2. Add context – By identifying who the document is for and what I am trying to do, the LLM will craft something closer to my desired outcome. In this case, a letter of recommendation already has formal business connotations, and the output reflects this context with its tone and format.

3. Supply specific content – Without specifics, the LLM would attempt to write a letter anyways and could potentially make up details about this student to achieve the desired outcome. Adding in specific content to include or an example of what you are looking for will increase the chances the quality of the LLM’s output.

4. Be nice – Adding pleasantries to your prompt has shown to improve an LLM’s outcome in many demonstrations contrary to what you might think. With LLMs being designed by humans, it is not too far fetched to consider that adding a human touch to your interaction is appreciated similar to any human in our daily lives.

These prompt elements can be used to specify many different types of documents, from lesson plans to meeting agendas. Consider utilizing an LLM to double check your work or same you time on tasks which are your least favorite. By gaining familiarity with these tools, we will be better equipped to turn around and teach our students about ethical and transformative uses of generative AI.

 

References:

• Conrad, H. (2024, May 14). Exploring artificial intelligence in the classroom. University of Texas at Austin News. https://news.utexas.edu/2024/05/14/exploring-artificial-intelligence-in-the-classroom/ 
• OpenAI. (2024). ChatGPT (Mar 18 version) [Large language model]. https://chat.openai.com
• Schell, Julie. (2024, June 26). Transformational GenAI. [Session 1: Insights into GenAI in Practice]. Adapting to Innovation: Facilitated Learning Network, Texas Higher Education Coordinating Board, Austin, TX, United States.