Resource Center

Generative AI in Teaching and Learning at the GSD

A computer-generated image spells the words Made by A I with pink mylar balloons

Image generated by Adobe Firefly (beta)

This page provides policies, information, and guidance for courses regarding the use of generative AI in teaching and learning at the GSD. Note that AI technologies are developing rapidly, and this page will be updated periodically.

Introduction

Updated 8/21/2023

Generative AI is artificial intelligence trained on large data sets that can predict letters, words, sounds, images, code, etc., based on the likelihood of those so-called ‘tokens’ occurring together. Note that generative AI is not intelligent and doesn’t think or apply reason as we understand it but produces output that mimics the data it was trained on, flaws included. Generative AI technologies have existed for years and are used in tools such as Microsoft Office, Grammarly, and others, but the release of OpenAI’s ChatGPT in November 2022 and other products since has made this technology publicly accessible in unprecedented ways and has given rise to much debate about its impact on human knowledge work and higher education.

Policies

Policies

What do I have to keep in mind when using generative AI in teaching and learning?

  • Compliance and copyright

    • You are responsible for the accuracy and compliance of your content
    • Depending on the tools and parameters you use, AI-generated content can be inaccurate, misleading, entirely fabricated, or may contain material protected by copyright
    • Do not submit work to which you don’t have rights into a generative AI tool and always be prepared to disclose your usage of any such tool

  • Academic integrity

    • The GSD’s academic integrity policy can be found in the Student Handbook and any reference to unauthorized human or non-human support and aids in producing academic work applies equally to generative AI tools
    • As a rule of thumb, wherever it is inappropriate for you to ask a human contributor to do work for you or where you don’t have explicit permission to share the work of one person with another, it is equally inappropriate for you to prompt an AI tool to do work for you or upload the work of others into an AI tool
    • Instructors determine what constitutes appropriate use of generative AI in their courses just like they have the authority to determine what constitutes appropriate use of established technological aids and reliance on human collaboration
    • Students are expected to be familiar with and abide by the school’s standards for academic integrity and conduct, and consult their instructor if they need clarification
    • It is suggested that instructors be proactive and communicate expectations for academic conduct and the use of generative AI tools for their courses (see “Guidance for GSD Courses” below)

Information

Tools and General Information

What tools are available, and where can I learn more about generative AI?

Generative AI tools available at Harvard

Generative AI tools available at Harvard

  • At this time, the only generative AI tool available to all Harvard users via an enterprise license is Adobe Firefly for image generation, accessible via the Adobe Creative Cloud; the Photoshop desktop app also uses Firefly as a built-in feature
  • HUIT is working on securing an enterprise agreement for ChatGPT and/or similar large language models (LLM) such as Google Bard and may soon be able to provide protected access to other tools such as Google Vertex, Amazon’s Bedrock, and Llama upon request
  • Enterprise licenses provide additional security via HarvardKey and may make it possible to input private or otherwise protected data into generative AI tools in the future; doing so with free, non-enterprise accounts likely constitutes a violation of existing policies on data security
  • If you have questions about the risk of using a specific tool, or are interested in learning whether HUIT may be able to provide a secured environment for experimenting with a specific tool, please contact [email protected]
Harvard Resources on Generative AI

Harvard Resources on Generative AI

Non-Harvard Resources on Generative AI

Non-Harvard resources

GSD Courses

Guidance for GSD Courses

What usage of AI tools is appropriate in GSD courses?

Communicating expectations for the use of AI tools in courses

Communicating expectations for the use of AI tools in courses

  • Instructors are encouraged to communicate clear expectations surrounding the acceptable use of generative AI tools for their course and/or individual assignments on their course syllabus and may require, encourage, or restrict specific uses and tools based on their course pedagogy
    • The required use of AI for a given course or assignment means that students must utilize an AI tool to complete the requirement
    • The encouraged use of AI means that students are not obligated to utilize AI tools but may do so if it fits the goals of their project
    • The restricted use of AI means that there are clear exclusions for certain uses of AI, keeping in mind that a total ban on the use of AI tools by students is widely seen as unrealistic and unenforceable except in controlled environments such as in-person exams
  • Note that any use of generative AI tools requires proper referencing, as is the case for all other resources and aids used in academic work
Common uses for students

Common uses for students

  • In alignment with course policy requiring disclosure and excluding certain scenarios, students might want to use AI tools in coursework for tasks such as
    • Formulating initial ideas and starting points for research and asking high-level non-specialized questions about their goals
    • Proofreading or correcting existing text similar to what is provided by tools such as Grammarly (which relies on AI in its main functionality)
    • Gathering references and resources for research, with great caution towards unreliable and fabricated content (sometimes called “hallucinations”)
    • Summarizing large datasets that are either publicly available or don’t otherwise violate data privacy policies. An example would be extracting verdicts from hundreds or thousands of publicly available legal cases
    • Analyzing existing and non-protected sets of data for correlations or possible patterns
    • Generating images with caution towards possible copyright infringement (note that Adobe Firefly is trained exclusively on licensed or freely available content and thus poses no risk regarding copyright)
Common uses for instructors

Common uses for instructors

  • Instructors are expected to use great caution in employing AI technologies in teaching and are responsible for ensuring accuracy. However, like traditional internet searches, AI-generated content can provide a useful starting point and inspiration for
    • Drafting lesson plans, exercises, or quizzes; note that output will almost always contain problems and requires careful review
    • Summarizing, simplifying, or customizing existing material such as lecture notes; when asked to edit text, AI generally won’t introduce new or misleading concepts, but vigilance is key
    • Note that while it is inappropriate for instructors to use AI for providing feedback on student work, it may be appropriate to ask students to seek AI-generated feedback on their work as part of a carefully framed assignment (e.g., AI tools may be getting better at auditing a design project for compliance with codes or other specifications)
Covering the cost of AI tools

Covering the cost of AI tools

  • Instructors are encouraged to utilize AI tools in their courses that Harvard or the GSD provides on enterprise agreements as they become available. In partnership with HUIT, the GSD is actively exploring how to make desirable tools available at no additional cost to users, and contract negotiations are ongoing
  • Instructors may also require students to purchase individual licenses to a specific generative AI tool not provided by Harvard or the GSD and are asked to include information and the expected cost on their syllabus, treating it the same as any other expected expense associated with courses, such as materials for fabrication
  • Instructors may not use course budgets to pay for or reimburse students for using AI tools in their courses but may use available research funds to pay for AI tools for their own research and experimentation; contact [email protected] to learn whether HUIT can offer support or access to a desired tool for your course needs