6.1 Teaching About (the Risks of) AI

This section explores ideas and activities for building student AI literacy: teaching about AI, including risks and concerns about AI use in their learning, work, and lives.

Teaching About AI

Begin by reflecting on what aspects of AI matter most in your course.

Ask Students About their AI Knowledge and Experiences

Invite students to share their prior knowledge of using assistive or generative AI. This information can give you a sense of what students know (and do not know) about AI.

• Why might you want to use AI as a student?
• When has AI helped you?
• When has AI not helped you?
• When might it be a good idea to use AI? When is it a bad idea?

You may also wish to invite students to ask questions they have about AI. Using an anonymous question collection tool, which reduces the risk students feel you are trying to catch them using AI, may lead to more submissions.

Use an Explainer Video

Use a simple video to start a class discussion about how AI works and its capabilities and risks. See the “Watching” section of the Hub’s AI Literacy for Educators post for videos you can share with your students.

Explain the Benefits and Limitations of Using AI

Consider sharing with students some of the benefits and limitations of using AI for learning in your course. You may even ask students what they believe about how generative AI can help or harm their learning!

Teach About the “AI Sandwich”

The article “Use the ‘AI Sandwich’ for a Human Approach to New Technology” emphasizes integrating AI with human insight and decision-making to value humanizing work with AI (Russell Reynolds Associates, 2023). Emphasizing the importance of humans driving AI use helps to avoid what Perkins, Roe, and Furze (2024) describe as “the illusion of finality” (p. 12), the tendency to accept AI outputs without any critique of their completeness or authority.

The AI sandwich suggests starting with human insights, such as establishing the criteria for a quality output. AI is then used for efficient creation or analysis. Finally, human judgment is used to finalize decisions: to accept, revise, or reject the AI output. This approach ensures that AI enhances rather than replaces human creativity and critical thinking.

Share “Do’s and Don’ts” Guidelines

Here are some guidelines for students if they are permitted to use AI in class for learning tasks and/or assignments. You are encouraged to review and adapt each statement to be relevant to your course. Pair these written guidelines with an in-class active learning activity to reinforce their meaning and relevance to your students’ learning. You may even wish to devise your own “do’s and don’ts” guidelines together as a class!

Use the accordion to explore sample statements for students to do, don’t do, and try doing. You may wish to select or adjust what you provide to students based on your course’s permitted level of use.

Teaching About the Risks of AI Use

AI use has many pitfalls, or hidden dangers, that students should know about. The risks of AI use are broad (economic, environmental, social) and narrow (academic integrity violations, lack of skills development, surface learning).

How do we help students avoid AI risks and pitfalls? See 8 ideas and associated activities below.

1. Provide clear and transparent AI Use Guidelines

Give a verbal and written set of clear AI use guidelines when the course starts and then again when introducing specific assignments. This strategy assists students who need clarity about appropriate and inappropriate use, including what constitutes an academic integrity violation. It can also help adult learners understand the “why” of the permitted or restricted level of AI use.

• • See the The Optional Use of Generative Artificial Intelligence (GenAI) in Assessments

2. Engage students in active learning to recognize AI’s limitations, myths and risks.

The Downsides of AI: IMPACT RISK

The IMPACT RISK acronym was developed by John Ippolito (2024). It is a mnemonic designed to help remember the potential downsides of generative AI. This acronym helps to remember the broad economic, environmental, and social risks associated with AI. You may wish to focus on ideas that relate specifically to your course or industry.

See the IMPACT/RISK website for an explainer video and infographic (CC 1.0 license).

The Downsides of AI in Learning and Research: LEARN AI

The LEARN acronym was developed by Fatima Zohra (2024). It focuses on elements of risk and potential harm with respect to the use or misuse of AI in the context of higher education.

Use the accordion below to learn more about the potential problems of using AI for learning and for research, as well as some suggested solutions to address the downsides.

Explore the “Misinformation, Security and More” Section of the GenAI Toolkit

AI makes mistakes, errors, biased outputs, stereotypes, and omissions. AI also risks information and data privacy. The GenAI Toolkit for Students contains a section with information, examples, and activities relating to the risks of GenAI.

Consider screen sharing in class any of the following information and activities to discuss with students class:

• H5P Some Considerations of the Harms of LLMs
• Examples of AI hallucinations
• Examples of bias and discrimination
• “AI Can Make Mistakes Too” Video
• The Most Likely Machine algorithm activity

Exploring Bias and Stereotypes in AI Text Outputs

Many myths and misunderstandings exist about how AI works with respect to saving time, creativity, and learning (Salvaggio, 2024). By analyzing AI outputs and comparing AI-generated and human-written content, students can begin to recognize differences in AI outputs with respect to quality, reliability, accuracy, and diversity (Kentucky Chamber, 2024).

Note that it is essential to guide and debrief these activities so that students are aware of the harm of stereotypes and bias, including psychological distress, anxiety, depression, and internalized bias, where people adopt the negative messages they hear about their group.

3. Explore Case Studies as Cautionary Tales

Use real-world industry or workplace examples of AI failures to illustrate potential harms. For instance, discuss cases where AI systems have exhibited bias or spread misinformation in your industry or field. Or, discuss the impact of AI errors, mistakes, and fabrications not caught by employees and the ethical, legal, economic, or reputational repercussions. These discussions can help students understand the real-world implications of AI risks.

See the website and sample repository, AI Gone Wrong: An Updated List of AI Errors, Mistakes, and Failures (Drapkin, 2024). See also the AI Incidents Database. More examples may be available in your particular field or industry.

4. Provide Decision-Making Practice Time Using Mini-Scenarios

Create or customize mini-scenarios related to course assignments and permitted use levels, then discuss them with your students. These can anticipate likely situations in which AI could be used, misused, or overused and the repercussions.

• A student uses an AI tool to generate an entire essay and submits it as their work without proper citation.
• For a group project, one member relies heavily on AI to complete their portion of the work, while others contribute manually. The AI-generated content is noticeably different in style and quality.
• A student uses an AI tool to solve complex math problems but doesn’t understand the underlying concepts. They submit the correct answers but fail to explain their process during an oral exam.
• A student uses AI to translate their homework instead of attempting to write it themselves in English.
• A student uses AI to generate data for a research project without verifying the accuracy or ethical implications of the data.

For each scenario, consider asking students to reflect on and discuss the following questions:

1. What is the problem in this situation? Who is being negatively impacted or harmed?
2. Can AI be used without breaking academic rules?
3. How might AI be used without doing all the work?
4. How can work be checked to ensure AI-generated content is accurate?
5. How can you show you understand the material, not just the answers?

5. Reflect on how AI May Negatively Affect Others

AI use may have negative effects, including on the perceptions of others. Ask students to consider the broader consequences of AI use on their reputation and relationships with employers, clients, colleagues, and other stakeholders.

Create a Reflection Discussion or Debate

Do people still expect “human” responses? What happens when people receive information, communications, advice, or research from a “robot” or AI? Is AI doing people more harm than good?

You may wish to facilitate discussions or reflections on the social and ethical implications of AI, exploring scenarios where AI decisions may conflict with human values. Students can debate the ethical considerations of using AI in various contexts, such as hiring or law enforcement. See examples of statements that might be debated:

• “AI does its research in a bubble.”
• “AI erodes human creativity.”
• “AI undermines trust and human relationships.”
• “AI algorithms do more damage than good for society.”

Explore Risks of Using AI in Work and Life

Students can use AI tools available to learn about the risks of AI:

• Google’s Teachable Machine: Train a simple image recognition model using your own images and discuss potential applications and ethical considerations.
• AI Fairness 360 (AIF360): Experiment with fairness metrics and bias mitigation algorithms to see how biases can be identified and addressed in AI systems.
• Fairlearn: Explore how different fairness metrics can impact model performance and outcomes.
• What-If Tool: Investigate model performance on different subsets of data to understand how biases manifest in AI systems.
• FAT Forensics: Evaluate AI systems’ fairness, accountability, and transparency.
• Themis-ml: Understand how different fairness interventions can be applied to machine learning models.
• FairTest: Test models for bias and discrimination.
• TensorFlow Fairness Indicators: Evaluate and improve model performance for fairness criteria.

6. Help Students Manage Stress, Time, and Workload

Helping students avoid AI’s pitfalls also means giving them resources and support for independent learning. Provide in-class time, completion templates, and recommendations to get help from the college so they are less tempted to use AI.

7. Demo Being an “Active Operator” of AI In Class

Teach about the risks of AI by doing active demos in class and walking students through the work of accessing, prompting, and saving AI outputs. If you invite students to use AI, demo how to critique, describe, document its use.

Remind students that human oversight of AI is essential an that AI outputs should never ben used without close critical review and verification.

8. Foster a Supportive Classroom Environment

Help students know you value their authentic work by showing care and compassion, encouraging their curiosity, and giving them choices in how they can follow their interests. Invite students to share thoughts and questions about AI. This builds trust and encourages open communication, making students more likely to seek guidance and use AI responsibly.

Next page >>