Considerations and Risks when using Generative AI

Considerations:

• Without a Privacy Impact Assessment (PIA), VCC cannot assess the impact of data collected, for what purposes the data will be used nor who will have access to the data.
• GenAI models use material taken from the internet, generally without the permission of the creators, potentially infringing on copyright and intellectual property rights.
• It may be unlawful to enter material into a GenAI tool that will save the data for training if the material is protected by copyright, intellectual property laws, or confidentiality.

Guidelines:

• Avoid entering personal or high-risk information. 
• Students can choose not to use GenAI tools that require account creation for instructor-assigned activities and an alternative must be provided.
• Students, staff and instructors should be aware of the user agreement of the tool and if input data becomes part of the tools training set, and what copyright exist for materials input. For example, do not upload a journal article accessed through VCC library subscription. You may not have permission to upload a copyright PowerPoint from a textbook.
• Do not submit student work for grading or to AI detector tools.

Considerations: 

• The training data for GenAI models has inherent biases so generated data may reproduce biases and can perpetuate sexism, ableism, racism, heterosexism, etc.  

Guidelines: 

• Check the output before sharing; assess for accuracy and bias (KPU, 2023). 
• Encourage students to use the tools with critical reflection and analysis (KPU, 2023). 

Chapman University provides an introduction to the places where bias can show up in AI as well as identify the types of bias that we can watch out for and apply critical analysis to whenever we are engaging with content  generated by AI. 

Considerations: 

• GenAI tools can pose risks to Indigenous data sovereignty (see the First Nations Principles of OCAP), Indigenous cultural and intellectual property rights, and may not respect Indigenous cultural protocols for information sharing and governance.
• There are also risks of cultural appropriation and perpetuation of stereotypes. 

Guidelines: 

• Use GenAI tools in ways that respect and support Indigenous data sovereignty and protocols.   
• Evaluate the output of the tools for risks of cultural appropriation and perpetuation of bias/stereotypes about Indigenous peoples, histories and cultures. 

Considerations:

• While GenAI can personalize learning, it cannot replace the value of human interaction in education.
• Over-reliance on AI could lead to a lack of meaningful interactions between students and instructors, and between students. This can impact social and emotional learning.

Guidelines:

• Design learning activities with ways that students interact with each other meaningfully.
• Include small group work and discussions.
• Provide avenues for check ins and interactions between yourself and students.

Considerations:

• Do students have equal access to GenAI tools (both due to digital skills and financial access)?
• Do students have access to technology with a reliable internet connection?

Guidelines:

• Design your learning activities so that students are not disadvantaged if they are not using the paid versions of the tools.
• Offer options and alternatives that will allow students to achieve learning outcomes without the use of the tools.

Considerations:

• Chosen tools ought to be navigable to all students and be aligned with assistive technology for those who need it.

Guidelines:

• If you are teaching a course to students using assistive technology, check that the tool is accessible.
• Offer options so that students are not required to use the tools.
• See the CTLR pages on UDL and Accessibility for more

Considerations:

• GenAI tools will generate believable inaccurate information to satisfy the prompt. They will attempt to predict the “best” though not necessarily most accurate, answer (AI “hallucination”).
• GenAI also tends to frame its responses to prompts as absolutes; it does not have the discernment built-in to determine whether the generated content is based on fact or not.
• GenAI may give references that are unrelated or nonexistent.

Guidelines:

• Recognize that AI generated content is a predicted output, not a fact.
• Check that the generated content is reliable and accurate before sharing (unless inaccuracy is part of the learning activity for students to do critical analysis).
• Include AI literacy training into your courses. You can adapt this lesson plan to facilitate discussions with your students.