AI AI in Teaching and Learning

AI and learning: what AI does in a task

What role AI plays in a task, what recent research says about offloading, and the AI skills students need. Faculty set the learning goals.

Key takeaways
  • Roles describe what AI does in a task; they are examples, not risk ratings.
  • The same tool can support or bypass learning, depending on how it is used.
  • Faculty set learning goals; take design questions to your teaching center.

Scope

Faculty determine the learning goals and assessment approach. This guide helps clarify the role, limits and expectations for AI within those decisions. It does not prescribe teaching methods.

Six roles AI can play

Often supports Instructional resource · Cognitive assistant
Depends on task Collaborator · Automation tool
Bypasses learning Substitute for effort
Is the learning Object of study

Permission levels say how much AI is allowed. Roles describe what the AI is doing. The six roles below are overlapping examples, not fixed categories or risk ratings: an explanation can still bypass the thinking a task is meant to build. For any use, ask three questions:

  1. Function: what is AI doing? Explaining, generating, critiquing, automating or supporting exploration.
  2. Learning relationship: does that use support, bypass or constitute the intended learning?
  3. Conditions: what verification, privacy, access and disclosure requirements apply?
Often supports

Instructional resource

AI explains, summarizes or gives examples, much like a textbook or tutor.

Asking AI to explain a statistics concept in plain words.

Often supports

Cognitive assistant

AI supports the student’s own thinking with practice, prompts or questions.

Generating practice questions before an exam.

Depends on task

Collaborator

AI and student work back and forth; the student keeps judgment and authorship.

Getting feedback on a draft argument, then revising it.

Depends on task

Automation tool

AI handles routine work that is not the learning goal.

Formatting references in a lab report about experimental design.

Bypasses learning

Substitute for effort

AI does the thinking or skill the assignment is meant to build.

Having AI solve the problem set the course is teaching.

Is the learning

Object of study

AI itself is examined, tested and critiqued.

Comparing AI answers with expert sources and documenting errors.

Cognitive offloading: support or substitute?

Letting AI explain a hard concept is different from letting AI do the reasoning a student is meant to learn. Recent experiments find the same tool can help or hurt, depending on how students use it.

AI supports the learning

  • Explains a concept the student is stuck on
  • Quizzes the student or gives practice questions
  • Critiques a draft the student wrote
  • Handles routine work that is not the learning goal

AI substitutes for the learning

  • Writes the text the student is meant to compose
  • Solves the problems being practiced
  • Does the analysis or reasoning being assessed
  • Makes judgments the student must learn to make
Working paper

In a randomized experiment, AI raised test scores. Gains lasted for students who used AI as a tutor and faded for students who had AI write for them.

Contractor & Reyes (2026) ↗
Peer-reviewed

Students who offloaded more ideas and reasoning to AI while writing showed lower higher-order thinking later, without AI. Self-regulation softened this.

Chen (2026) ↗
Peer-reviewed review

Evidence shows both gains and learning penalties. The difference lies in how AI is set up and used, so offloading should be managed task by task.

Nakamura (2026) ↗

From AI literacy to AI fluency

Three stages of student capability. Recent survey data show AI use has become routine, so literacy needs to go beyond a one-time orientation (From Novelty to Normalisation, 2026).

AI literacy

Understand what AI is, what it does well and poorly, and its risks, including made-up facts and bias.

AI competency

Apply AI appropriately to defined tasks, and check and cite the results.

AI fluency

Integrate AI into complex work while keeping independent judgment and accountability.

Eight AI competencies students should build

Treat AI literacy as a learning outcome, not only an integrity rule. Each competency can be taught and assessed in a discipline.

  1. 1 Understand how AI worksExplains in plain terms how generative AI produces output.
  2. 2 Spot limits and errorsIdentifies fabrications, gaps and outdated information.
  3. 3 Evaluate AI outputChecks claims against reliable sources and discipline standards.
  4. 4 Protect informationKeeps personal, confidential and others’ data out of AI tools.
  5. 5 Recognize bias and ethicsNames bias, ownership and fairness concerns in a use.
  6. 6 Apply AI in the disciplineChooses when AI fits the field’s methods, and when it does not.
  7. 7 Explain and defend useDiscloses use and justifies choices to an instructor or peer.
  8. 8 Exercise independent judgmentMakes the final decisions and takes responsibility for them.

Where to get specialist help

Questions about learning design, assessment validity and disciplinary teaching belong with people who hold that expertise.

Teaching and learning center

Course and assessment design, learning goals and teaching methods.

Instructional designers

Building activities and assessments, including AI-integrated ones.

Department or program

Discipline norms and shared expectations across sections.

Disability or accessibility services

Accommodations, including assistive technology that uses AI.

Privacy, IT and academic integrity offices

Approved tools, data rules and misconduct procedures.