This is How We AI: A Practical Approach to Teaching with Artificial Intelligence

Synthesis: Hanshaw, Brewer and Hunt (2026), writing as faculty and academic leaders at Los Angeles Pacific University, argue that AI in higher education is not primarily a tool problem but a teaching-practice shift, and that the governing question is no longer whether students use AI but how it is positioned pedagogically. Drawing on empirical studies of AI course assistants run in LAPU courses, the guide offers a role typology, design patterns, readiness checklists, and low-risk entry points while holding that faculty expertise becomes more valuable, not less. Its central claim is that faculty framing of AI shapes student behavior far more than any tool or policy: two instructors can use the same AI in the same course and produce entirely different outcomes. The guide therefore favors process-based assignments and guided transparency over bans, detection, and surveillance, arguing that ignoring AI only leaves its use unstructured, unexamined, and inequitable. It is a practice guide, not peer-reviewed research, and its evidence comes largely from one institution's own implementations.

Key Findings

1. The guide's typology describes the common roles AI plays in student learning — answer machine, editor, practice partner or coach, tutor, and thought partner — and argues that when instructors do not name the intended role, students define it themselves by what is easiest.
2. Quality depends on framing, not the tool: two faculty members can use the same AI in the same course and produce entirely different learning outcomes, because the difference is the instructional frame, not the technology.
3. A suggested grading shift moves weight from product to process — 60% final product, 20% process documentation, and 20% reflection on AI use — so the journey is valued alongside the destination.
4. A redesigned 5-page argumentative essay adds a process narrative of 1-2 paragraphs plus, if AI was used, 2-3 useful prompts; the reflection adds roughly 10-15 minutes to student work and gives instructors the information they need.
5. Students consistently name three gains from AI — clarity, confidence, and momentum — while still turning to faculty for judgment, disciplinary context, and feedback on thinking rather than grammar.
6. The pre-assignment checklist asks whether AI could complete the task wholesale, whether student thinking will be visible, whether the instructor has clarified AI's role, and whether they would be comfortable if students used AI exactly as designed.
7. The most significant gains accrue to students who feel underprepared, non-traditional and asynchronous learners, students navigating language and literacy challenges, and international or geographically isolated learners; over-reliance on AI tutoring risks learned helplessness.

The AI Role Typology

The guide's most portable contribution is a typology of the common roles AI plays in courses. As an answer machine, AI bypasses cognitive engagement entirely: the student asks, receives, and submits. As an editor, AI can support genuine revision, but only if students document what they kept and rejected and explain why; uncritical acceptance makes editing passive. As a practice partner or coach, AI supplies low-stakes, repeated practice with immediate feedback for mock interviews, language conversation, public speaking, and clinical case simulations. As a tutor, AI personalizes explanation and multiple representations, but risks learned helplessness when consulted before independent effort. As a thought partner, it is arguably the most cognitively demanding use, requiring students to drive the dialogue, defend positions, and judge when AI pushback is useful or misleading. Role confusion, the guide warns, produces mismatches between faculty expectations and student behavior, creating conflict, eroded trust, and misuse that is really just confusion about unstated boundaries.

From Content Expert to Learning Architect

The guide reframes AI as changing where instructional effort is invested rather than whether instructors are needed. With AI able to answer factual questions, explain concepts, and provide examples on demand, the instructor's content knowledge is most valuable when applied to course design: sequencing learning experiences, crafting constraints that promote thinking, and engineering the cognitive friction that supports deep learning. This is the learning architect stance, and it extends to competency: instructors decide when AI should challenge, clarify, or extend student thinking. The guide pairs this with Active Learning patterns already familiar in online courses, noting that AI is most effective when embedded into learning activities rather than added on top of them — guided practice, structured reflection, and iterative feedback positioned as part of the learning process itself. Embedded use, it argues, minimizes extraneous cognitive load, normalizes appropriate use, and keeps instructional authority over how learning unfolds.

Modeling, Presence, and Visibility Without Surveillance

Teaching presence, the guide argues, can be extended rather than reduced by AI. Students learn appropriate use not from policy documents but from watching instructors model thoughtful, disciplinary-grounded interactions — sharing prompts, narrating revision decisions, questioning AI output, and showing where they stop. Because detection tools are unreliable and create false accusations, the recommended substitute is design-based visibility: process-based assignments where AI interaction is visible by design, such as drafts with revision notes, decision journals, and annotated AI transcripts. The guide frames this as guided transparency rather than monitoring, and reframes the underlying question from "Did you use AI?" to "How did you use AI, and what did you learn from the process?" Trust, clear community norms, and Feedback that rewards reflective AI use are presented as the real safeguards, with journals, reflective notes, and artifacts doing the work that detection cannot.

Designing Assignments and Practical Guardrails

The guide contrasts AI-resistant "AI-proof" design with AI-intentional design, arguing that restriction-based approaches are exhausting, create adversarial relationships, and often remove the very challenges that develop Critical Thinking. In their place it offers scalable design patterns: AI as drafting partner (learning happens in revision), as Socratic questioner (devil's advocate against the student's thesis), as role-play character (a client, patient, or historical figure), and as feedback mirror (students evaluate which AI suggestions are useful and which miss the point). It then recommends simple redesign moves faculty can make this term: add a reflection layer, require AI interaction logs, shift grading weight from product to process, and build in multiple checkpoints. Ethical Guardrails include treating AI use as a teaching practice with explicit interaction norms, teaching AI literacy and verification of hallucinated claims, offering low- or no-tech alternatives so access differences do not become achievement differences, and reserving values assessment, relational feedback, and final judgment for humans.

What this means for practice

• Instructors. Name AI's role explicitly in every assignment — expected, optional, or prohibited — and explain why, since undefined roles are filled by AI as an answer machine.
• Instructors. Start with one low-risk change this term: add a reflection prompt, require a brief AI interaction log, and reward documented process, then observe and iterate before scaling.
• Faculty developers. Support the mindset shift from prevention to pedagogical design, and provide models of expert AI use rather than relying on policy statements to change practice.
• Instructional designers. Move grading weight from final product toward documented thinking and revision, and build in checkpoints so student reasoning is visible at multiple stages.
• Administrators and institutions. Favor design-based visibility, trust, and clear norms over detection tools and bans, and offer low- or no-tech alternatives so AI access gaps do not widen outcome gaps.

Limitations

• This is a practice guide authored by faculty and academic leaders at one institution (LAPU), not peer-reviewed research; it is explicitly an invitation to practice rather than a study report.
• It synthesizes the authors' own institutional studies of AI course assistants, but the document itself reports no sample sizes, effect estimates, or control-group comparisons.
• Evidence is drawn from a single university's courses and culture, so transferability to other institution types, disciplines, and student populations is untested.
• The guide is written with acknowledged partnership from a commercial AI vendor, whose tools informed the reported practice.
• It offers no cost, workload, or implementation-effort data, and no comparison against prevention-based or non-AI teaching.
• The typology, activation, and design patterns are descriptive frameworks without evidence that a specific pattern outperforms another.

Citation

Hanshaw, G., Brewer, C., & Hunt, S. (2026). This is How We AI: A Practical Approach to Teaching with Artificial Intelligence. EdArXiv preprint.