Learning with AI

October 2024

Artificial Intelligence and Teaching in Higher Education

A series of timely workshops, discussions, and course design sessions focused on AI.

Approved UMaine AI Guidelines

Generative AI Guidelines for Teaching and Learning

What is generative AI?

generate statistical probabilities of images or narrative writing in response to simple prompts, such as: “describe a Maine coastline from the perspective of an economist.” Their output can seem competent, especially as a description, utilize good grammar and syntax, and could be hard to differentiate from student work. These tools are simple to use, fast, and, currently, some are free.

On the other hand, they generate statistically probable results. Not real or actual facts. This means quotes, citation, and other data may be assembled words and/or numbers, but they were generated by the AI not drawn from an actual source.

We may benefit by differentiating between five kinds or points on the spectrum of generative AI which we use today (more are coming):

Embedded AI or the kind that is infiltrating our email clients, productivity software, graphics software, search engines, operating systems, phone apps, etc.
General purpose, stand-alone genAI, such as ChatGPT, Claude, Dall-E
AI agents that complete tasks such as information gathering and presentation
Discipline-specific genAI such as iNaturalist, Merlin, OverflowAI
GenAI used as engines for processing large volumes of data

Of course, there is blurring between these and an individual might use ChatGPT to check spelling and grammar as well as to process their own large data sets.

Why are we talking about it?

Students can use generative AI to complete assignments.

Some may use their instructor’s assignment as a prompt with an AI service and simply hand in the AI generated results. Others may start with their own thoughts, use AI, incorporate some material and jettison others, rework on their own, and finish with a grammar check (a very similar approach would also work for programming assignments, especially at the introductory level).

Currently, companies, such as Turnitin, with whom UMaine has a contract for assessing plagiarism of student submissions, claims to have the technology to detect AI-generated text. Assessments of Turnitin’s results don’t support this claim. We are not aware of any vendor who has succeeded.

It is highly probable that today’s students, upon completing their education, will go into a field in which their employers and/or customers/clients will expect a high level of proficiency in the use of AI.

What should you do about this in the classes you teach?

Faculty members could approach this in a range of ways.

In each case, though, instructors should explain to students their rationale for including/excluding specific types of generative AI in their course.

Course outcomes include the effective use of AI in the context of the course. Descriptive citations are required with each assignment submission.

Faculty will need to design their course to meet these specific learning outcomes. Faculty will also need to arrange for approved use, payment, security, and accessibility for the tool(s).

Tip: This approach may be most effective once students have gained a significant degree of mastery in the discipline and are ready to apply their skills at analyzing or developing sophisticated hypotheses with new content. Ask them to submit their topic to one or more generative AI tools and critique the results.

The instructor expects students to use generative AI services to develop content, programs, or other materials which will contribute to final assignment submissions. Descriptive citations are required with each assignment description.

If faculty require students use specific tools or services, the faculty will need to arrange for approved use, payment, security, and accessibility for the tool(s).

Tip: Talk to students about how to use and not use AI. Discuss ways AI can help as well as hinder their creative and analytical work. Share your expectation for their use of it. Students should have enough knowledge and skills within the discipline to gauge the utility (or lack of utility) of AI.

Faculty may present students with AI generated content and/or assign students to use generative AI services themselves in order to provide instruction on critically assessing generative AI services. Instruction may include helping students to identify bias, the impact of the dataset on the results, inaccuracies, and/or to critically evaluate the technology that supports the service. Faculty may work with students to assess methods of crafting prompts and the results produced by the AI services.

Tip: Discuss with students the direct relationship between the datasets as well as the approach to developing the technology and the results they generate. The goal of the lesson will be for the students to become informed evaluators and users of generative AI services.

Students may use generative AI services to help develop and explore responses to assignments. This would include using such services to uncover potential gaps or missing components in the analysis, presentation, code, calculation, and/or graphic. All content and materials submitted must be authored or developed solely by the student or the student members of a group project. No citation for this use is required.

Tip: Continue to build students proficiency in the discipline as well as appreciation of the limits of AI. Introduce AI as a detailed search engine that can help them find a range of approaches to assignments which may support or refute the argument they are considering.

After completing an assignment, but prior to submitting it, students may use generative AI services to check grammar, code, the clarity of their presentation, and style. No citation for this use is required.

Prior to submitting assignments, ask students to run materials through a generative AI tool and share the results and the students’ response/critique of the results with the instructor. Design and develop assignments which will not work well with generative AI tools.

Tips: Introduce this use of AI as analogous to an ever-present and tireless tutor-editor helping with clarity, completeness, and presentation. Consider constructing assignments not easily or proficiently answered by AI, such as:

Ask students to write about a specific and current example of a phenomena or event.
Ask students to write about/compare two current journal articles that they access via Fogler subscriptions.
Ask students to write about a local topic or setting, which could involve interviewing a faculty or postdoc researcher, a staff member from a regulatory agency, dam worker, etc.
Assign students a specific and current data set to interpret and or develop a strategy to accommodate.
Give students assignments which require college-level critical thinking, analysis, and creativity rather than any type of summary or basic highlighting of themes or ideas. Ask them to speculate why things happened rather than narrate what happened.
If possible, ask that they write on a topic which would allow them to reflect on their personal experience or goals within the field.

Students are not to use any generative AI service to complete any assignments in this course.

At some level, this will need to be on the honor system. There are grammar and citation tools included with Google Docs, Microsoft 365, and Brightspace. There is no real way of knowing how or if students have used most generative AI tools.

Tips: This may be a suitable approach when students are new to a discipline and are still developing basic skills. In addition to the tips under Minimal Use, require students to prepare assignments in Google Drive and share the file with the instructor to track the version history. If it is feasible, have them come to office hours to talk about their research for their assignments.

Citing or acknowledging the use of AI services.

You may wish for students to cite their use of AI tools in foot or endnotes, or via blanket acknowledgment.

Citation in foot/endnotes: “Service used, version, date. Text of prompt. Link to service.”
Blanket acknowledgement: “Portions of the text and ideas in this document were created or edited using Service.”

If you would like help crafting assignments and assessments for your courses, use this form to Request a consultation with an Instructional Designer.

Other Resources

Jon Ippolito from New Media created an online Learn with AI Toolkit. Here you will find a sortable collection of AI related Resources, Strategies for Using AI, and an FAQ.

For those interested in a deeper dive into AI, the University of Maine also hosts an Artificial Intelligence Initiative (UMaine AI). The goal of this effort is to bring together university, industry, government, and community collaborators from Maine and beyond to advance the field of artificial intelligence, and, through development of innovative technologies and applications, find transformative solutions to enhance human life and societal well-being.

In September of 2023 the Margaret Chase Smith Policy Center at UMaine posted a Maine Policy Matters pod cast on The Impact of AI on Research and Higher Education, featuring Ali Abedi, Salimeh Sekeh, and Peter Schilling.