Generative AI Guidelines

“Generative AI (GenAI) is beginning to disrupt all aspects of society, from teaching and learning to art and design to engineering innovation. While GenAI holds immense potential to augment productivity and minimize redundant tasks, its use also carries significant risks. U-M has the intellectual depth, resources, and international and national connections and networks to be the leader in the development and appropriate use of GenAI.”

—Generative Artificial Intelligence Committee Report (PDF)

Introduction

These guidelines establish principles for the responsible use of generative artificial intelligence (GenAI) in communications at the University of Michigan’s Office of the Vice President for Communications. In alignment with our commitment to academic excellence, integrity and innovation, these guidelines seek to harness AI’s potential while upholding our core values.

GenAI tools are impacting communications on our campus and beyond. This set of guidance endeavors to place some guardrails as well as point to areas of opportunity for its use in our professional practice. GenAI can be a tool to enhance our work. As an institution that creates and curates knowledge and innovation, we believe that humans should be the primary authors of such work. GenAI is helpful to automating repetitive work or facilitating faster work. (ie: reformatting text or AI noise reduction in images)

Our guiding principle is that we need humans to remain in the loop. We should not cede important decision making to an algorithm. This is not to say that we cannot employ AI tools as a component of a decision-making process. People are responsible for the final output of all products.

Purpose

  • To guide the responsible use of AI in enhancing communication and outreach efforts.
  • To safeguard the integrity, originality and accuracy of our communications.
  • To ensure compliance with relevant university policies related to privacy, copyright and data security.

Principles and Guidelines

Ethical Adoption and Use

  • AI tools should be used to complement human decision-making and skill, not replace it.
  • Approval from relevant supervisors in consultation with the ITS Privacy team should be obtained for purchasing and adopting new AI tools, ensuring alignment with institutional privacy and data storage standards.
  • AI should enhance, not replace human creativity and decision-making.
  • The role of AI in content creation should be clearly defined, from idea generation to content development.

Originality and Expertise

  • People at U-M remain the original source of content in areas where we hold subject matter expertise.
  • AI-generated content should not replace expert opinions and original scholarship from our community.
  • Use generative AI as a tool to enhance creativity, not to deceive or mislead.
  • Fully synthetic GenAI imagery should be a rare exception. And in order to avoid confusion between photographs and photorealistic constructions, these GenAi illustrations should be made to look like drawings or illustrations, and be labeled clearly. Consider traditional stock images when original images cannot be captured.
  • In sensitive communications, priority must be given to human responses. AI should not replace the personal voice necessary in times of crisis or hardship.
  • Respect copyright and licensing terms when training or prompting generative models.

Accuracy, Accountability and Transparency

  • Human editors are responsible for verifying the accuracy of AI-generated content and analysis, ensuring it aligns with institutional messaging strategies.
  • AI-generated imagery or analysis that cannot be fact-checked should be avoided.
  • Proper attribution should be provided for AI-generated content, with citations used where necessary.
  • The use of AI in producing content should be disclosed if such transparency would be required for human-generated content.

Confidentiality and Privacy

  • Personal and proprietary data must be handled with care, never inputting sensitive information into AI tools without appropriate consent and safeguards.
  • Outputs from AI should reflect our commitment to inclusivity, avoiding any form of bias or prejudicial content.
  • Never use AI to generate synthetic representations of real individuals without their explicit permission.

By adhering to these guidelines, we can ensure that our use of AI in communications enhances our institutional reputation for excellence and innovation while safeguarding ethical standards and values. Regular reviews and updates to these guidelines will also be undertaken as AI technologies and applications develop.