Teaching and Learning in an Era of Generative Artificial Intelligence
1. Generative Artificial Intelligence
Generative Artificial Intelligence (GAI) refers to a subset of artificial intelligence techniques that involve generating new content, such as images, text, music, or videos, that is similar to existing data it has been trained on. GAI has the potential to open up new avenues for research and innovation in various fields, but it also introduces ethical concerns in how one uses the technology safely, responsibly, and securely. Teaching students to critically evaluate the capabilities, limitations, and biases of GAI may be an essential skill for helping future generations utilize the power of GAI for societal good.
In Summer 2023, the Provost invited a multidisciplinary group of faculty to strategize how Fordham faculty might engage with GAI and related technologies in ways that would enhance and not compromise Fordham’s teaching and research mission. The AI Vision Committee wrote a comprehensive report with a helpful list of recommendations.
The following webpage, inspired by the committee’s report, is intended to help instructors explore the use of various GAI tools and consider the degree to which they want to encourage or discourage their students to use such tools in their courses. The webpage lists some GAI products on the market along with resources for detecting AI-generated content. It provides guidance for how instructors can design course assignments that promote deep learning and academic integrity in an age where GAI tools are readily available to students.
2. Common GAI tools and their capabilities
Below is a partial listing of GAI tools that are commercially available to generate original content in response to user prompts:
- Claude: from Anthropic. Marketed as being a more helpful and honest tool.
- GPT4: most recent version by OpenAI. Marketed as being more inventive and accurate.
- ChatGPT: most common version used by the public.
- ChatSonic: developed by the technology company Writesonic. Integrated with Google Search to create content with real-time data. Generates visuals, voice commands, and more.
- AlphaCode: programming capabilities in Python, C++, and several other languages.
- GitHub Copilot: a code completion Artificial Intelligence tool.
- Bard: a chatbot and content generation tool developed by Google.
- Synthesia: a tool for creating videos. With little to no work, it rapidly generates and broadcasts videos of professional quality.
- DALL-E2: OpenAI’s recent version for image and art generation.
- Copy.ai: creates variants of marketing texts for specific goals and target demographics.
- Murf.ai: an online tool that uses AI to generate high-quality voice-overs for videos, presentations, and text-to-speech needs.
3. Sample Course Activities that Train Students to Become Critical and Proficient Users of GAI-Generated Content
- Conduct in-class discussions analyzing AI-generated writing to understand its strengths and limitations.
- Assign students to revise and edit AI-generated texts to elevate them to their own standards. Students will submit both the original AI draft and their final version.
- Organize in-class presentations comparing and contrasting AI writing with human writing. Prompt students to reflect on elements replicable by ChatGPT and aspects unique to human authors in their work.
- Explore refinement techniques by having students compose variations of the same prompt to fine-tune AI-generated results. Require students to submit the prompts used for GAI and assess their ability to effectively customize and adapt AI-generated content to fit specific contexts or target audiences.
- Scaffold engagement with AI tools by encouraging students to interact with AI, using it for brainstorming or divergent thinking exercises.
- Assess the practicality and usefulness of AI-generated content in real-world scenarios, such as marketing materials or informational texts.
4. Sample Course Assignments/Activities that Deter Student Use of GAI Assistance
- Use oral presentations rather than written assignments to assess students' understanding and communication skills.
- Employ interactive, in-class exercises to promote active learning and real-time application of concepts, fostering a deeper understanding of the subject matter.
- Engage in case studies based on current events (within the past year). Given ChatGPT's reliance on established empirical datasets, this approach effectively prevents students from obtaining recent information from a GAI tool.
5. Sample Syllabus Statements to Clearly Communicate the Instructor’s Expectations around Usage of GAI
Faculty need to clearly communicate the degree to which it is unacceptable for students to use GAI tools in their course. The course syllabus is an ideal place to communicate such expectations. Below are example syllabus statements for courses where the instructor wishes to (a) ban the use of GAI in the course, (b) allow students to engage with GAI tools under certain conditions, or (c) signal how students may responsibly use GAI tools freely in the course.
a) For a "No-AI" approach:
"Generative AI tools are not permitted in this course. Students [or learners] must rely on their own originality, creativity and critical thinking skills to complete all assignments and engage with course material."
b) For a "Limited-AI" approach:
"Limited usage of generative AI tools may be allowed for specific assignments in this course, enabling exploration of ideas, complex data analysis, and creative solution development, when explicitly permitted by the instructor. When using these tools, it is mandatory to clearly indicate the sections of your work that were generated using them for proper attribution and transparency, and indicate the prompts and software versions that were used. It is critical to adhere to ethical standards by refraining from activities like plagiarism or creating misleading content. Additional guidelines or restrictions will be provided for specific assignments."
c) For a "Full-AI" approach:
"This course allows the use of generative AI tools to facilitate exploration of innovative ideas, complex data analysis, and creative solution development. Students must clearly indicate the sections of the work that were generated using generative AI tools for proper attribution and transparency, and indicate the prompts and software versions that were used. It is critical to adhere to ethical standards by refraining from activities like plagiarism or creating misleading content. Additional guidelines or restrictions will be provided for specific assignments."
6. AI content detection tools
AI-detection tools attempt to identify if a submitted text has been written by a GAI engine. Some of the more reliable AI-detection tools are listed below, but none are 100% foolproof. All of these tools can, at times, deliver false positives or false negatives.
- GPTzero: https://gptzero.me
- ZeroGPT: https://www.zerogpt.com/
- Writefull GPT Detector: https://x.writefull.com/
- GPT-2 Output Detector: https://openai-openai-detector--mqlck.hf.space/
- TurnItIn: https://help.turnitin.com/ai-writing-detection.htm
Other approaches to detecting a student’s unauthorized use of GAI tools include:
- Compare the quality and creativity of take-home assignments with in-class work, considering factors such as coherence, style, and relevance.
- Evaluate content accuracy and relevance in addressing specific assignment objectives.
7. Resources and References
a) Forums, Tutorials, and Webinars on GAI
- UMichigan Institute for Data Science (MIDAS) Data and AI in Society Forum: AI in the Classroom
- UTexas: Forum for Artificial Intelligence
- MIT AI Policy Forum
b) Other Resources and References for Faculty and Students
- Fordham’s Academic Integrity Policy
- The Sentient Syllabus Project
- Bias and discrimination in Generative AI:
- Dwivedi et al., 2023. “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642.
- Sun, L., Wei, M., Sun, Y., Suh, Y. J., Shen, L., & Yang, S. (2023). Smiling Women Pitching Down: Auditing Representational and Presentational Gender Biases in Image Generative AI. arXiv preprint arXiv:2305.10566.
- Ferrara, E. (2023). Should chatgpt be biased? Challenges and Risks of Bias in Large Language Models. arXiv preprint arXiv:2304.03738.
- Srinivasan et al. (2021) Biases in Generative Art: A Causal Look from the Lens of Art History. arXiv preprint arXiv:2010.13266
- AI Privacy concerns