Generative AI tools are rapidly evolving. While we will periodically update the information in this guide, please be aware that the content may become outdated quickly.
Last updated: August 16, 2024
Introduction to Generative AI
Relevant Terminology
Artificial intelligence: theory and development of technology that enables computers to perform tasks that normally require human intelligence, like pattern recognition, speech recognition, and problem solving
Machine learning: a subset of artificial intelligence that focuses on the development of algorithms and statistical models that enable computers to learn patterns and make decisions based on data, without explicit programming
Generative artificial intelligence: a subset of artificial intelligence that uses machine learning models to generate new, original content (such as text, images, video, or audio) based on patterns and statistically likely relationships learned from training data.
Large language model: a type of generative artificial intelligence system that can produce natural language text based on a given input.
What is Generative AI?
As a disruptive technology, generative artificial intelligence (GAI) tools present both new opportunities and challenges for teaching and learning. Since the release of OpenAI’s ChatGPT in November 2022, these technologies have continued to evolve, expanding the opportunities for their use.
Generative artificial intelligence is a subset of artificial intelligence that uses machine learning models to generate new, original content (such as text, images, video, or audio) based on patterns and statistically likely relationships learned from training data. This capacity is achieved through advanced algorithms and neural networks that are trained on vast amounts of data to respond to prompts provided by humans. In response to prompts, these tools can provide contextually relevant, coherent output.
Examples of GAI tools include Microsoft Copilot, OpenAI’s ChatGPT, Claude’s Anthropic, Perplexity AI, OpenAI’s Dall-E, Midjourney, Adobe Firefly, Runway, ElevenLabs, Suno, etc. These tools are capable of performing a wide range of tasks, from generating realistic images and videos to composing music and writing essays. Specialized GAI tools can create slide decks, perform literature reviews, or perform other discipline-specific tasks.
With the increasing ubiquity of GAI tools, stemming from their integration into word processers, search engines, grammar assistants, and more, the importance of clearly communicating with students about what a generative AI tool is (and how to know if they are engaging with one), as well as the appropriate use of these tools in their coursework has grown significantly.
Limitations of Generative AI
It is important to note that while GAI tools can deliver quick and often highly accurate output, they are not human. Therefore, they don’t possess knowledge or comprehension of the materials they generate. In particular, GAI tools:
- Are not fully reliable – they hallucinate information and make reasoning errors
- Exhibit a wide range of bias that reflects the human biases found in the training material and/or training process
- Make it difficult to trace the source and provenance of information incorporated into GAI output
In addition, there exists a wide range of concerns about generative AI tools, from ethical concerns about privacy and IP, to the environmental impacts of generative AI.
Additional Resources
Generative AI in a Nutshell: How to Survive and Thrive in the Age of AI (18 min)
By: Henrik Kniberg
Practical AI for Instructors and Students (10 to 12 minutes each)
By: Ethan and Lilach Mollick
Generative AI and Teaching
Here are a few steps you might take as you decide how or whether to incorporate GAI tools or GAI output into your courses:
Learn: Gain a baseline understanding of how these tools function and how they are commonly applied. How might you use these tools, personally or professionally?
Explore: Experiment with GAI tools relevant to your discipline. Practice using GAI to complete an assignment from your course. Identify opportunities for student learning, as well as areas of concern.
Reflect: Might GAI use support or undermine students in achieving any of your course learning outcomes? How important is it for your students to have experience with GAI tools, or understand GAI-related issues?
- Consider your course learning outcomes, as well as the tasks students complete to demonstrate they have achieved those outcomes. Does GAI leave some of them unchanged, render them moot, or allow you to scale up or enhance some?
- Consider the expectations in future courses or workplaces regarding the understanding and knowledge that your course helps students develop, as well for their responsible and ethical use of GAI tools. Are there areas where your course learning objectives or course assignments might evolve, or even lean into the use of GAI tools or output?
Set and communicate your stance: Decide whether and when students could engage with GAI tools in your courses. Craft a course policy that clearly communicates this stance (see Example Syllabus Policies related to Generative AI, below). Talk with your students about this stance early and often. If students will be allowed or encouraged to engage with GAI tools, we encourage you to:
- Share how students should properly document use of GAI tools or cite use of GAI output.
- Determine if all students will have fair and equal access to GAI tools.
- Determine what support or education students might need to determine the accuracy and validity of GAI output.
Additional Resources
The AI Pedagogy Project (metaLAB (at) Harvard)
Designing AI-Resilient Learning Experiences (MIT Sloan School of Management)
ChatGPT Assignments to Use in Your Classroom Today
By: Kevin Yee, Kirby Whittington, Erin Doggette, and Laurie Uttich
Generative AI Product Tracker for Higher Ed (Ithaka S+R)
Academic Honesty and Generative AI
At UGA, the default rule for student use of AI on their coursework is that it is not permitted unless it is explicitly authorized by the course instructor.
In each individual course, the instructor of record can determine what constitutes acceptable use of GAI by students. Therefore, it is important to proactively engage students in an open discussion about your expectations regarding GAI. For example, is the use of GAI off-limits, sometimes OK, or always encouraged?
Example Syllabus Policies related to Generative AI
The following examples are general, course-level syllabus policy examples representing varying stances on the use of GAI in a given course (highly permissive, moderately permissive, or maximally restrictive). In addition to these examples, you can explore a growing repository of classroom policies related to generative AI available here.
Example 1: Highly permissive
In this course, thoughtful and creative engagement with AI is encouraged. Therefore, you will not be penalized for using Generative AI tools (GAI) in this course.
However, remember that AI tools are best used to augment, not replace, our skills. It is important to be aware of the limitations of GAI tools. For example, GAI may exhibit bias and inaccuracies. You are responsible for assessing the validity and applicability of any GAI output incorporated into your work.
If you provide minimum effort prompts, you will receive low quality results. Practice careful prompt engineering.
Be cautious with data privacy. Refrain from entering information of a highly personal or confidential nature into a GAI tool.
To gain a deeper understanding of GAI tools and enhance your interaction with GAI output, you can explore the following resources:
Note also that any work not created by yourself should include a proper source citation. For Generative AI, that may include prompts, specific AI used, and date. For full details on how to properly cite AI-generated work, please see the APA Style article, How to Cite ChatGPT.
Keep in mind that this AI-permissive policy applies only to this course. Using AI tools in courses without permission is a violation of UGA’s Academic Honesty policies and is considered unauthorized assistance. Be sure to consult the specific guidelines for each course.
Example 2: Moderately permissive / mixed
The use of Generative AI (GAI) tools is not generally permitted in this course, but will be permitted for select assignments. [Alternative opening statement that is more permissive: You are welcome to explore the use of generative artificial intelligence (GAI) tools for your work, but use of GAI tools should be limited to providing support as you develop your thinking and knowledge base for an assignment.] Specific assignments will include details about whether use of GAI tools is allowable or not, and why. If you are uncertain about using a particular tool to support your work, please consult with me before using it.
Please note that you may not represent output generated by a GAI tool as your own work. Any such use of GAI output must be appropriately cited or disclosed, including quotation marks and in-line citations for direct quotes. Including anything you did not write in your assignment without proper citation will be treated as an academic misconduct case. Suspected unauthorized assistance, plagiarism, or other violations of UGA’s “A Culture of Honesty,” will be reported to the Office of Academic Honesty. For full details on how to properly cite AI-generated work, please see the APA Style article, How to Cite ChatGPT.
If you are unsure where the line is between collaborating with GAI and copying from GAI, I recommend that you do not have your assignment and the GAI tool open on your device at the same time. Instead, take notes in your own words while you interact with the GAI tool, then use your notes to remind you of what you’ve learned and to inform your work. Never copy output from GAI tools into your assignment. Instead, use your interaction with the tool as a learning experience, then close the interaction down, open your assignment, and let your assignment reflect your improved understanding. (Sidenote: This advice extends to AI assistants that are directly integrated into a composition environment or grammar modulation tool.)
Finally, GAI is highly vulnerable to inaccuracy and bias. You should assume GAI output is wrong unless you either know the answer or can verify it with another source. It is your responsibility to assess the validity and applicability of any GAI output used.
Example 3: Maximally restrictive
To ensure you develop and master the foundational knowledge and skills in this course, the use of generative AI (GAI) tools is strictly prohibited. This includes all stages of your work process, even the preliminary ones. This prohibition extends to AI writing tools like Grammarly and Wordtune, as well as GAI tools like ChatGPT, Copilot, Writesonic, Rytr, and Rtutor [this list might be amended to include discipline or task-specific tools that apply to your course]. If you are uncertain about using a particular tool to support your work, please consult with me before using it.
Turnitin AI Writing Detector
Turnitin’s AI Writing Detector analyzes student writing, flagging segments of text it believes were either copied directly from generative AI output, or created using an AI-paraphrasing or word modulator / spinner tool. Turnitin’s AI Writing Detector is the only AI Writing Detector approved for use at UGA. UGA instructors should not use AI detectors that are not supported by UGA, as these other tools have not been vetted by UGA’s information security team, for FERPA compliance, or for protection of student’s intellectual property.
Turnitin’s AI Writing Detector can be found within the Similarity Report of a Turnitin assignment on eLC. For the Turnitin AI Writing Detector to function, work submitted to the Turnitin assignment must meet the following requirements (as of July 2024):
- File size must be less than 100 MB
- File must have at least 300 words of prose text in a long-form writing format
- File must not exceed 30,000 words
- File must be written in English
- Accepted file types: .docx, .pdf, .txt, .rtf
How should I interpret Turnitin AI Writing Detector scores?
The AI Writing Report Score indicates the amount of qualifying text within a submission that the writing detection model believes is likely generated by an AI tool. Turnitin’s AI Writing Detector will produce scores of 0% or between 20% and 100%. Turnitin testing has determined that there is a higher incidence of false positives when the score is between 1-20%. As of July 8, 2024, scores between 1% and 19% are replaced by an asterisk to indicate that the score is less reliable.
It is important that Turnitin scores not be used as a definitive measure of misconduct. Instructors should review sentence-level data from Turnitin and determine the appropriate next steps. If you suspect a piece of work was completed using generative AI, carefully analyze the work and compare it to past student writing. If you believe a violation of A Culture of Honesty may have occurred, instructors have a responsibility to report to the Office of Academic Honesty. A facilitated discussion will be scheduled for you to talk with the student about their process for completing the assignment and whether a violation occurred.
A note of caution about Turnitin’s AI Writing Detector
Although Turnitin’s AI Writing Detector is the sole, preferred AI writing detector for use at UGA, instructors should use Turnitin’s AI Writing Detector with caution.
False positives: Turnitin reports a sentence-level false positive rate of 4%, but studies show that this false positive rate is higher for students who speak English as a second language, as well as for text that was modified by a grammar modulator. Presumably, Turnitin’s AI Writing Detector would also flag students who inadvertently choose words in an order that Turnitin’s probability models say they should not.
Lack of source text: Turnitin’s Similarity detector provides contextual information for any flagged content. This allows instructors to determine if flagged content truly matches what Turnitin has identified as similar, by accessing and reviewing the suspected match. However, this is not the case for Turnitin’s Writing Detector – it is not possible for any AI detection tool to offer such context or suspected source.
AI advances outpace detectors: Turnitin’s AI Writing Detector was trained on output from GPT-3.5. Those with the resources to access more sophisticated tools than GPT-3.5 are better able to avoid detection.
Looking for more information on the AI Writing Detector? You can visit this Turnitin Guide.
Questions about Generative AI & Teaching?
Contact our teaching & learning experts for a one-on-one consultation today!
- Teaching Resources
- Active Learning
- Digital Learning Tools
- Establishing your Syllabus
- Evidence-Based Teaching Strategies
- Reflection, Feedback & Evaluation of Teaching
- TA Resources
- Teaching Amid Disruption
- Scholarship of Teaching & Learning (SoTL)
- Teaching Policies & Procedures