Supporting Appropriate Student Use of AI

Help Students Understand the Line Between AI Use That Enhances Their Learning and Use That Impedes Their Learning

There has been much attention given to the observation that, since AI's can generate plausible (if not correct) responses to many typical assignments (e.g. lab reports, problem sets), they make cheating dramatically easier. At the same time, AI's capabilities could make it an important learning aid if used correctly. But what does 'used correctly' look like, and how does a student tell whether their particular use is helping them learn or just serving as a substitute for learning? How can we structure our courses and assignments to help students understand where this line is and to dissuade them from AI use that strays in the wrong direction?

• Be clear about the learning goals for each assignment. If a student knows what knowledge and skills they are supposed to be exercising, they can ask themselves the critical question: if I use an AI tool in this way, at this moment, am I going to be learning the way the assignment intended me to, or is the AI tool going to be doing work that deprives me of that challenge.
• Address this issue explicitly with your students, at least in your syllabus, perhaps on the first day of class, or ideally at the level of the individual assignment. Be clear about when and what type of AI use is allowed, and explain why these limits make sense given the learning goals you have for the course or specific assignment.
  
  In this climate data analysis assignment, you may not use AI to interpret the temperature and precipitation trends and draw conclusions about regional climate patterns for you. While AI could process the datasets and generate explanations, the learning goal is for you to develop your own skills in reading meteorological data, recognizing patterns, and understanding the physical processes that drive climate variability. You may, however, use AI to help you understand statistical methods or clarify atmospheric science concepts that come up in your analysis.
• Promote the metacognitive growth that the critical question above exemplifies. Encourage and support your students to develop the skill of reflecting on and regulating their own learning -- including their decisions about how and when to use AI. You can learn more about metacognition with the metacognition resources found in Teach the Earth.

How have you helped your students navigate AI ?

 

Use Teaching Strategies and Assignments that Discourage AI Shortcuts

Your teaching approach can make taking AI shortcuts less appealing. Students who feel personally invested in their learning, rather than seeing a course as just a hoop to jump through to obtain a grade, will be less inclined to take shortcuts.  Involve students in goal-setting and decision-making about how and what they are learning. Design learning experiences that are relevant to them as people. Personalize assignments so they connect to issues and ideas students care about. Use approaches like service learning that make their work about more than just a grade. You can learn more about this affective side of geoscience teaching in this On the Cutting Edge site about the Affective Domain

The temptation to use AI shortcuts is perhaps most tempting on graded assignments. There are a variety of strategies that have been suggested to make taking an AI 'shortcut' less appealing.

• In-class proctored assessments. This includes the classic exam or quiz, but can also include things like in-class presentations and one-on-one interview-style tests.
• Multi-stage assignments where students turn in and get feedback on drafts of work in progress.
• Assignments that draw from the student's personal experiences.
  
  Analyze the geological hazards (earthquakes, landslides, flooding, etc.) that pose risks to your hometown or a place where you've lived. Interview a family member or local resident about any natural disasters they remember, research the local geological setting, and create a hazard assessment plan for your community.
• Hand-written work (copying answers out of the AI is certainly possible, but more work).
• Group work that includes peer assessment.
• More frequent smaller assessments. Big, high-stakes assessments, especially at the end of a term, are more likely to catch students in a time crunch, where using an AI shortcut seems like the only good option.
• Just raise the bar. Use assessments that require students to do work that AI's simply can't on their own. The educators that suggest this approach design assignments that expect AI use: use an AI tool to help you do work that is more sophisticated than you could without it. This approach requires careful calibration of the difficulty of the assignment, current AI capabilities, as well as consideration of whether the assignment disadvantages students with weaker AI skills and especially students who don't want to use AI (see our AI Ethics section).

This article in Vox touches on recent data around student AI use and cheating in K-12 education.

 

What are your experiences with how AI is impacting how we assess student learning?

 

Use AI to Enhance Student Practice and Self-Assessment

While AI can make accurate summative assessments more challenging, it has the potential to help provide more numerous formative assessment opportunities. A student's ability to frequently and accurately assess how their learning is progressing is foundational to their ability to self-regulate and guide their own learning efficiently. Exams and quizzes serve as good benchmarks, but are time-consuming to produce. So, faculty often save the 'good questions' for their exams. As we have described in the educator strategies page, AI can be useful for brainstorming new questions. This can be coupled with the approach of asking AI to make multiple variants of an existing question to produce collections of strong assessment questions more efficiently. This can open the door to giving students access to large numbers of 'good' questions they can use on their own to probe their understanding.

Based on these lecture notes about igneous rock classification [student pastes in their notes about igneous textures, mineral compositions, and formation environments], create 15 practice questions to help me test my understanding. Include a mix of question types: some that ask me to identify rock types from descriptions of texture and mineral content, some that test my understanding of formation processes, and a few that require me to explain the relationship between cooling rate and crystal size. Make sure to include both felsic and mafic examples. Provide the answers separately so I can check my work after attempting the questions.

This can be taken a step further. You might encourage your students to ask an AI to quiz them in the same way a tutor or study partner would. This can be effective if they also provide the AI with materials from your course (perhaps their own notes) so that the AI can ask relevant questions. There are a variety of AI prompts available that ask the AI to take on this tutor role that you can adapt. There are also commercial AI tutor tools that wrap this into a product. Just make sure your students are aware both of the privacy/copyright issues involved in sharing the relevant source materials, and the issue that the AI may end up producing bad or even misleading guidance. It's a strong strategy (and even a useful learning experience) to start with the critical stance, "is the example problem the AI just generated for me a relevant and appropriate problem for what I'm trying to learn?"

Your Syllabus and Framing AI's Place in Your Course

With AI just a click away for most of our highly connected students, and the news full of stories about the impact it may have on the job market they will be soon facing, AI is on your students' minds and in play in your classroom. There are a few key issues you'll want to be sure you've grappled with, regardless of the course you're teaching and your stance on AI.

• How might AI change the landscape your course is preparing students to navigate? What skills and knowledge will they need, and how might that differ from what you've engaged past students with?  Consider the topics in our AI in the Geoscience Discipline section.
• What are the ethical concerns raised by AI ?  What perspectives on these might your students hold, and how might that impact how you teach?
• What institutional resources and policies will impact your students' AI-use experience?  Are there equity issues with who has access to AI tools, and what AI expertise students arrive with?

Your thinking on these topics can be reflected in your syllabus, perhaps in an AI-specific section. We encourage you to explore this collection of syllabus AI statements contributed by educators across a range of disciplines to inspire your thinking about what is right for your courses:

• Google Sheet of Syllabus AI Statements from AI + Education = Simplified. 

 

 Have an AI section in your syllabus?    

Consider sharing it in the Google Sheet above, or Share it with some commentary through our form »