Relay Library for Faculty and Staff

What is Generative AI?

Generative AI is technology that can create new content such as text, images, or videos. It functions as an advanced tool for content creation across multiple media formats.

Large Language Models (LLMs) are one type of generative AI that specializes in creating text. These systems learn by analyzing millions of examples of human writing, then apply that knowledge to generate new text.

LLMs work similar to an advanced autocomplete feature. They predict what word should come next based on patterns learned during training. This allows them to produce confident-sounding responses, but they do not actually understand topics the way humans do—they excel at predicting what sounds appropriate based on their training data.

Sample Applications of Generative AI:

• Content Creation: Writes articles, generates code, creates music or images
• Accessibility: Makes content more accessible by simplifying complex language, generating alternative text, or presenting material in alternative formats (e.g., creating a podcast from an article).
• Personalization: Tailors content and experiences to individual contexts, preferences, and needs
• Writing Enhancement: Assists in editing and improving existing text for clarity and readability.
• Information Summarization: Processes large amounts of information quickly and extracts key points.
• Concept Explanation: Breaks down complex topics and supports exploration of new ideas.
• Idea Generation: Produces ideas on specified topics, presents different perspectives, and helps organize thoughts into structured formats.
• Scenario Simulation: Can adopt different roles or personas for training, practice, or educational purposes.
• Research Support: Provides topic overviews for research purposes, suggests relevant search terms, may suggest relevant sources.

Limitations and Concerns of Generative AI

• Limited Knowledge Base: Training data does not include the most recent information or specialized content that is not publicly available online.
• Accuracy Concerns: AI systems do not truly understand topics—they recognize patterns. This can result in incorrect information, fabricated sources, or oversimplified explanations of complex subjects.
• Inherent Bias: AI systems reflect the biases present in their training data.
• Copyright and Privacy Issues: AI development involves using extensive content without explicit permission from original creators. Further, users should exercise caution when sharing personal or confidential information with AI.
• Environmental and Ethical Impact: AI systems require significant energy resources and may involve problematic labor practices in their development and operation.
• Potential Learning Impact: Excessive reliance on AI for writing or summarization tasks may reduce opportunities for skill development and critical thinking that come from performing these tasks independently.

Sources:

Yale Poorvu Center

Claude AI was used to edit this content. The prompts used were:

• "Revise the following content to make sure it is concise, accurate, consistent, and professional."
• "Revise the text to make it easier to understand"
• "Make the tone more professional and less conversational, but keep the simpler language and clear structure"

The final output was lightly edited to ensure that the original ideas behind the content were maintained.

Additional Resources:

Link to AI tools and resources for education, links to disclosing it effectively.

Uses of AI for Teachers

• Lesson planning
• Creating "tutor bots"
• Ask program to write a response/exemplar to an assignment from your class (consider having students critique this!)
• Using AI to make documents accessible
• Generate case studies - these are fictional anyway, so the fact that AI makes things up is fine! 
• Writing CFU questions and feedback - make sure you check accuracy, consider providing it with specific information (e.g., through PlayLab)
• Design a rubric (make sure to review)
• Generate discussion prompts - Can generate human-like opinions and perspectives that may spark discussion, be up front about the source, and ask students about biases that may be present. Could help explore diverse perspectives
• Customized instruction and feedback

Uses of AI for students

• Accessibility
• Have students use AI as a tutor - or have students act as a tutor to AI!
• Be sure to remind students about AI literacy, critically review its accuracy
• Overcome writer's block - draft intro lines, topic sentences, etc
• Improve writing - often, bad writers are penalized, even if they know the actual skills required for the assignment

Source

How AI can be used meaningfully by teachers and students

Instructors are understandably concerned about the impact of generative artificial intelligence (GenAI) on academic integrity. However, relying on AI detection software is unreliable, and designing "AI-proof" assignments is not a long-term solution given AI's rapid advancements.  Further, teaching students to ethically and effectively use AI is crucial for preparing students to live in a world where the technology is becoming unavoidable.    

Fostering a strong class culture that emphasizes intrinsic motivation, honesty, and transparency is more effective at encouraging students to use AI in ethical and appropriate ways.   

Strategies for Assessments in an Age of AI:

• Foster AI Literacy: Encourage appropriate use of AI 
  • Open Conversations About AI: Initiate open dialogues with students about AI's role in education, including potential intellectual trade-offs from overuse.
  • Discuss GenAI's Capabilities and Limitations: Explore GenAI's strengths and weaknesses, and model ethical AI use and disclosure yourself. 
  • Have Students Practice Critical Use of AI: Assign tasks where students use generative AI in ways that they might use it in their professional lives, then require them to critique and refine the AI-generated content based on established criteria. Ask students to share their AI prompts and outputs, along with reflections on their choices and the AI's effectiveness.
• Emphasize Learner Agency through Universal Design for Learning: Design assessments that emphasize intrinsic motivation, so students want to complete them without over-relying on AI. Reduce barriers so that students feel competent enough to complete assignments independently. The following UDL guidelines are especially relevant: 
  • Welcoming Interests and Identities: Ensure that assignments are relevant to student goals and interests to boost motivation for completion
  • Sustaining Effort and Persistence: Clearly articulate to students why the course content matters and how the assignments will help them learn that content. Foster collaboration and offer action-oriented feedback to encourage effort and persistence.
  • Expression and Communication: Help students ethically and effectively use GenAI as a tool to support communication while preserving their critical thinking and their own unique voice and perspectives
  • Strategy Development: Guide students in planning for challenges, monitoring progress, and organizing information and resources so that they can manage the demands of their course without resorting to unethical AI use.
• Focus on Strong Assignment Design: Build rigorous assessments that focus on process and assess content rather than writing.
  • Assess Content, Not Writing: Design rubrics that evaluate knowledge and application of content rather than solely writing mechanics.  While generative AI can write well, it can't engage in critical thinking about content. Assessing thinking over writing is also more equitable for students who grasp and can apply content but may struggle with Standard Academic English. 
  • Require Higher-Order Thinking: Design assignments that demand original thought and higher-order thinking. AI often struggles with nuanced or highly specialized content; prompting AI for meaningful content requires topic expertise and skill. Further, because AI refers to previously created content to generate responses, posing questions that require nuanced analysis of topics still under debate not only increases rigor but also limit's AI's ability to generate relevant content.
  • Assess Process Over Product: Shift the focus to evaluating the learning process rather than just the final product. Consuder implementing multi-stage assignments where students receive feedback throughout the process (from peers, instructors, or even AI), allowing them to refine their work and understanding.  Consider requiring students to show different drafts and reflect on their revisions and how they incorporated feedback.
• Develop AI Resistant Assignments: While designing "AI proof" assignments is not a long-term solution, the following tips may make it more challenging for students to rely solely on AI: 
  • Test Your Own Assignments: Proactively run your own assignments through generative AI tools. If AI can competently answer the question with limited prompting, consider revising the assignment.
  • Require Inaccessible Resources: Require students to use materials and information not readily available to most generative AI tools, such as library databases, the integration of specific course content/concepts, or context-specific information (e.g., knowledge of particular students or school contexts). While students may feed this information to AI, it requires more critical engagement.
  • Alternative Formats and Tools: Encourage students to demonstrate understanding through non-textual formats like video recordings or voice posts, or ask them to use specific templates, software, or annotation tools (e.g., Perusall or Hypothes.is). While generative AI can still assist, these format require more user intervention.
  • Increase In-Class and Group Work: Increase the number of in-class activities, group projects, discussions, and presentations.

Sources:

https://educational-innovation.sydney.edu.au/teaching@sydney/how-ai-can-be-used-meaningfully-by-teachers-and-students-in-2023/
https://melbourne-cshe.unimelb.edu.au/ai-aai/home/ai-assessment/designing-assessment-tasks-that-are-less-vulnerable-to-ai

https://educational-innovation.sydney.edu.au/teaching@sydney/what-to-do-about-assessments-if-we-cant-out-design-or-out-run-ai/

https://educational-innovation.sydney.edu.au/teaching@sydney/chatgpt-is-old-news-how-do-we-assess-in-the-age-of-ai-writing-co-pilots/

https://www.oneusefulthing.org/p/all-my-classes-suddenly-became-ai

https://udlguidelines.cast.org/

Additional Resources: