How can Artificial Intelligence – AI – help researchers and students in their work? How do faculty members understand students’ use and misuse of AI in their papers and presentations? What are the ethical considerations of this new tool?
Ruopeng An, associate professor at Washington University in St. Louis, outlined some answers to those questions in a May workshop for the College of Community Engagement and Professions. His presentation focused on the “Use of Chat GPT to Enhance Research, Teaching and Learning.”
The presentation covered the basics of ChatGPT, particularly GPT 4, preparing courses and teaching using AI, and the ethics involved in using these tools.
Designing effective prompts to gather helpful information
An opened the workshop by introducing participants to the concepts behind AI. The basic definition of ChatGPT is that it is a large language model designed by Open AI that uses artificial intelligence to generate human-like responses to natural language inquiries.
AI has become a tool to guide self-driving cars, facilitate speech recognition, generate written text for a chatbot, and assist researchers and students, among its many uses.
ChatGPT has become widespread and generated an incredible amount of media attention. As of January 2023, the application had more than 100 million users, making it the fastest-growing consumer application of all time, according to An’s presentation.
However, there is a large gap between those who can use AI effectively and those who can’t and don’t.
According to An’s presentation, the key to using ChatGPT and other generative AI tools is designing effective prompts to gather helpful information. The process involves choosing and crafting specific words, phrases, or code snippets to achieve the user’s goal.
Ten principles for creating effective prompts:
- Be specific and transparent in your prompt
- Break down complex questions into simpler parts
- Experiment with different prompt formulations
- Set context and provide instructions
- Ask for step-by-step explanations or pros/cons
- Request for sources and citations
- Ask for alternative viewpoints (arguments and counterarguments)
- Use constraints to control response length or format
- Provide examples (few-shot learning) to guide the model
- Encourage critical thinking and exploration
An offered numerous examples for each of these principles. Here’s an example he gave of different prompt formulations:
“If you aren’t satisfied with the response to “How do genetic factors influence obesity?” try asking, “What are the specific genes and pathways involved in the development of obesity?”
Or, instead of asking, “What are the risk factors for Alzheimer’s disease?” provide context: “Describe the modifiable and non-modifiable risk factors for the development of Alzheimer’s disease.”
Chat GPT can be helpful for faculty in different stages of research, from identifying research topics and generating questions to data collection and management and preparing the final presentations.
Implications for teaching
The workshop also offered faculty insights into the impact of ChatGPT on teaching, including how to assess student work and provide guidance to students on the ethical use of ChatGPT and other AI programs.
As the use of AI becomes more prevalent, instructors must work closely with students to ensure academic integrity. Faculty can sometimes review work by looking at sentence length and structure, for example. According to An’s presentation, AI writing tends to have sentences of short, uniform size, monotonous text, predictable language, and hedging phrases like “some might say,” in a similar way. AI-generated artwork/photos often include some watermarks that computer programs can detect.
As the use of AI becomes more prevalent, instructors must work closely with students to ensure academic integrity.
An outline of several AI detection programs that use algorithms to perform similar functions was provided. These programs can help assess whether the student wrote the content based on writing patterns, syntax, and linguistic factors. An’s presentation offered links to many AI detection tools. The accuracy of these tools varies from 64% to 85%, so he said it’s essential not to use them as the sole indicator of artificial content generation.
However, An concluded, as AI programs become more sophisticated and students and others use these tools in collaboration with their own writing and design work, decisions about academic approaches become more complex.
The workshop also offered sessions on facilitating course preparation and teaching using ChatGPT and customizing ChatGPT for personal tutoring and self-learning.
Ethics and AI
An also discussed the issue of ethics in the use of AI.
The presentation offered examples of how AI could be misused in creating deceptive photos and videos to spread misinformation, showing, for example, an AI tool used to create a fake video of Barack Obama making phony speeches.
Another example from a 2019 New York Times story focused on a couple who applied for Apple cards. Even though the wife’s credit score was better and the couple shared finances, assets, and tax returns, Apple gave the husband a credit limit 20X higher than hers. The husband, the entrepreneur, wrote in “Fast Company,” magazine: “Apple has handed the customer experience and their reputation as an inclusive organization over to a biased, sexist algorithm it does not understand, cannot reason with, and is unable to control.”
According to the presentation, the use of facial recognition technologies is another ethical challenge.
In a 2019 study of facial recognition programs, researchers found that Blacks, Asians, and Native Americans were much more likely to be misidentified by these programs than white people.
An offers an Advanced AI certificate program for those interested in learning more. Application for the program is open till August 21, 2024.