Constructive Alignment in the Age of Generative AI: How to Realign Learning Outcomes, Activities, and Assessment Evidence

When Australian educational psychologist John Biggs formulated constructive alignment in the 1990s, he established a standard for university curriculum design: learning happens when students perform cognitive activities that build conceptual understanding. Three components must interlock: Intended Learning Outcomes defining what students should know, Teaching and Learning Activities providing practice, and Assessment Tasks evaluating whether students achieved the target standard.

In a constructively aligned course, the verb in the outcome governs the curriculum. If an outcome asks students to analyze an epidemiological dataset, classroom activities must practice epidemiological analysis, and the assessment task must require students to execute that analysis. When the task demands the exact cognitive behavior named in the outcome, grading yields genuine evidence of learning.

Generative artificial intelligence breaks the operational link between the outcome verb and the assessment artifact. Large language models synthesize essays, evaluate code, and draft business plans from brief prompts. When a student submits an assignment generated by synthetic software, the document appears aligned with an outcome requiring critical evaluation. The student bypassed the cognitive act, and the assessment evaluated the artifact rather than student thinking.

To restore constructive alignment, universities must realign assessment tasks through direct demonstration, scenario decision-making, and source-grounded spoken defense.

The Broken Triad: Why Generative AI Decouples the Verb from the Task

Traditional constructive alignment assumed that generating a complex deliverable required human cognitive struggle. Writing long served as an indirect proxy for higher-order thinking. Instructors assigned research papers under the premise that articulate prose reflected internalized understanding.

Generative tools eliminate that struggle. Today, a student can paste course objectives into a prompt box, upload readings, and receive a structured essay in seconds.

This creates an operational breakdown. On a syllabus audit, the course appears aligned:

1. Intended Learning Outcome: "Critique competing theories of social stratification."
2. Teaching and Learning Activity: Reading assignments and seminar discussions.
3. Assessment Task: A 2,500-word critical essay.

In practice, the task measures prompt engineering and text curation rather than theoretical critique. The student never wrestled with conflicting sociological paradigms. The course exhibits unaligned assessment: the grade reflects an activity decoupled from the intended learning verb.

In our analysis of the collapse of the artifact economy in higher education, we demonstrated why evaluating static submissions fails when software separates text production from cognitive effort. Awarding credentials based on unverified text undermines institutional credibility.

The SOLO Taxonomy in the AI Era: Distinguishing Surface Structure from Relational Depth

To diagnose how generative models disrupt learning outcomes, educators must examine the companion framework Biggs developed with Kevin Collis: the Structure of Observed Learning Outcomes, known as the SOLO taxonomy.

The SOLO taxonomy categorizes student understanding across five progressive levels:

1. Prestructural: The student misses the core concept.
2. Unistructural: The student understands one relevant aspect.
3. Multistructural: The student identifies several independent facts.
4. Relational: The student integrates components into a coherent whole, explaining relationships and applying principles.
5. Extended Abstract: The student generalizes integrated concepts to unfamiliar domains, generating hypotheses and new perspectives.

Generative AI mimics multistructural and nominal relational prose with ease. Language models identify statistical associations across training data, producing essays that list historical factors, cite precedents, and use sophisticated transitions.

The deliverable creates an illusion of high-level SOLO attainment. An essay appears relational because it links legal precedents smoothly. When an instructor questions the student directly, however, the facade vanishes. The student cannot explain why one precedent outweighs another under altered facts, revealing that understanding remains unistructural.

As explored in our examination of Bloom's Taxonomy in the age of generative AI, machines generate the surface language of evaluation without semantic comprehension. Evaluating true relational and extended abstract learning requires verifying whether students can adapt principles spontaneously under questioning.

Why Text Detectors and Stricter Policies Cannot Restore Alignment

When generative AI entered higher education, many institutions attempted to defend traditional assessment tasks through surveillance and regulation. Colleges purchased commercial AI detection software to flag synthetic text, while academic committees updated student honor codes.

Neither approach solves the breakdown of constructive alignment. Text detectors evaluate probabilistic linguistic markers rather than student learning. As documented in our guide on alternatives to AI detection for college assessments, commercial detectors generate unacceptable false positives, particularly among multilingual students and neurodivergent writers. Accusing students based on unreliable probability scores creates institutional liability and damages trust between faculty and learners.

Surveillance cannot fix a curriculum design defect. If an assessment task no longer measures the intended cognitive verb, policing the tool does not restore validity to the grade. Even if an instructor could identify AI writing with absolute accuracy, the take-home essay would remain an indirect proxy vulnerable to automated generation.

Constructive alignment is an instructional design challenge. When an assessment task stops eliciting the target cognitive verb, the professional response requires redesigning the task so that evidence of learning is direct, authentic, and verifiable.

Curricular Walkthrough: Three Realigned Course Blueprints

To see how constructive alignment operates under realigned assessment tasks, consider three academic disciplines where faculty replaced vulnerable written assignments with interactive demonstration and spoken defense.

Scenario 1: Health Policy and Management

• Intended Learning Outcome: Evaluate how Medicare adjustments impact hospital margins, defending trade-offs to a clinical board.
• Legacy Assessment Task: A 15-page policy memo analyzing bundled payment rules.
• AI Vulnerability: Language models generate polished memos without student calculations.
• Realigned Assessment Task: Students submit financial models paired with oral defense: "In table three, you cut nurse staffing by eight percent to preserve margins under bundled rates. Defend that allocation against clinical safety guidelines."
• Alignment Outcome: Students demonstrate relational understanding by explaining trade-offs verbally without notes.

Scenario 2: Software Engineering and Distributed Systems

• Intended Learning Outcome: Design fault-tolerant distributed architectures, justifying consensus protocol selections under network partitioning.
• Legacy Assessment Task: An architecture specification document and postmortem.
• AI Vulnerability: Generative models draft architectural descriptions from basic prompts.
• Realigned Assessment Task: Students submit architecture diagrams paired with an adaptive defense: "Your database encounters network isolation across region two. Defend choosing eventual consistency over strict serializability during this partition."
• Alignment Outcome: The oral defense verifies individual systems design and trade-off comprehension.

Scenario 3: Environmental Economics

• Intended Learning Outcome: Critique efficiency and distributional equity in cap-and-trade systems versus carbon taxation, applying economic theory to regional energy markets.
• Legacy Assessment Task: A comparative term paper analyzing state climate policies.
• AI Vulnerability: Chatbots assemble literature reviews with standard citations requiring minimal synthesis.
• Realigned Assessment Task: Students submit brief position papers followed by adaptive oral defense: "Suppose industrial lobbies secure free permit allocations for top facilities in your model. Explain why this exemption alters abatement incentives, and analyze the resulting incidence on household consumers."
• Alignment Outcome: Students demonstrate extended abstract thinking by generalizing economic principles to altered market conditions.

Realigning Teaching and Learning Activities: Practicing Spoken Defense in the Classroom

Constructive alignment requires that teaching and learning activities mirror the demands of the final assessment task.

If an instructor assesses students through oral defense and scenario handling, classroom activities must provide regular practice in verbal justification. Expecting students to defend technical decisions on microphone without prior rehearsal creates speech anxiety and measures public speaking confidence rather than conceptual mastery.

Faculty can structure teaching activities to support spoken demonstration throughout the term:

1. Low-Stakes Verbal Check-Ins: Instructors ask students to record sixty-second spoken reflections summarizing the core tension in weekly readings rather than posting to discussion boards.
2. Peer Questioning Drills: In classroom sessions, students work in pairs to cross-examine project assumptions, using instructor-provided stems to probe conceptual edge cases.
3. Formative Rehearsals: Before submitting capstone projects, students complete an ungraded practice defense to experience adaptive questioning and identify conceptual gaps.
4. Rubric Transparency: Faculty share the evaluation rubric early, clarifying that grading rewards analytical depth and evidentiary support rather than vocal polish or absence of hesitation.

By integrating verbal explanation into routine learning activities, instructors develop evaluative judgment in higher education. Students learn to monitor their own understanding, evaluate competing evidence, and articulate reasoning before encountering summative examinations.

How Rocketproof Operationalizes Governed Alignment Under Faculty Control

While faculty recognize the pedagogical validity of oral defense, scheduling individual interviews has historically created an unsustainable grading burden. Administering thirty-minute oral exams for one hundred students requires fifty faculty hours, making spoken assessment impractical for large cohorts.

Rocketproof provides infrastructure to scale authentic assessment under faculty control:

• Faculty-Defined Rubrics: Instructors define learning outcomes and rubrics, ensuring questioning reflects course standards.
• Source-Grounded Questioning: Rocketproof analyzes uploaded papers or models and generates defense questions anchored directly to student work.
• Asynchronous Spoken Defense: Students complete oral defenses asynchronously in their browser, responding to structured questions and follow-up probes.
• Structured Evidence Cards: The platform synthesizes student performance into an instructor dashboard with timestamped quotes and rubric alignments.
• Faculty Grading Authority: Instructors retain full grading authority, reviewing evidence and finalizing scores in two to three minutes.

Rocketproof connects with campus infrastructure through LMS integration supporting Canvas, Blackboard, Brightspace, and Moodle via LTI 1.3 Advantage.

The platform supports accessibility accommodations aligned with Universal Design for Learning, including extended timers and typed text alternatives. Standardized rubric criteria evaluate subject matter understanding rather than vocal polish, supporting fairness as outlined in our guide on grading bias in oral assessment.

Explore these workflows on our authentic assessment platform, review our oral defense platform, test interactive AI simulations, or schedule a platform demo.

Frequently Asked Questions

How does generative AI specifically invalidate John Biggs's constructive alignment?

Constructive alignment requires that completing an assessment task demands the cognitive process named in the learning outcome. When generative tools produce essays and code automatically, students submit deliverables without engaging in the intended cognitive struggle, breaking construct validity.

Does realigning a course require abandoning written deliverables entirely?

No. Writing remains valuable for structuring complex arguments. Realigned courses preserve written projects while adding an oral defense. The written paper serves as the artifact under examination, while the spoken defense verifies that the student understands the underlying analysis.

How can large-enrollment courses implement spoken assessment without overwhelming instructor grading time?

Asynchronous oral defense platforms eliminate live scheduling bottlenecks by allowing students to record responses independently. Rocketproof transcribes defenses and drafts preliminary rubric alignments. Faculty review the synthesized evidence card in two to three minutes per student, retaining complete grading authority.

How does constructively aligned oral defense support regional and specialized programmatic accreditation?

Regional accreditors such as SACSCOC, HLC, MSCHE, and WSCUC require direct evidence of student learning mapped to published outcomes. Spoken defenses provide verifiable records that students achieved competencies independently, satisfying standards detailed in our analysis of accreditation and generative AI.

How do instructional designers guide faculty who are hesitant to redesign established assignments?

Instructional designers can recommend backward-compatible adjustments that preserve established course materials. Rather than rewriting an entire syllabus, faculty add a brief, source-grounded spoken defense to existing assignments, restoring alignment between outcomes and evidence without increasing preparation burdens.

Restoring Pedagogical Integrity to Higher Education Curriculum Design

John Biggs gave higher education a clear standard: teaching, learning, and assessment must operate as an integrated system focused on what students do.

For three decades, universities treated the submission of written artifacts as synonymous with student learning. Generative artificial intelligence dissolved that assumption permanently. When computers write articulate papers, continuing to grade static text turns curriculum design into an exercise in self-deception.

Realignment requires courage and clarity. By restructuring assessment tasks around direct demonstration, scenario decision-making, and source-grounded spoken defense, higher education institutions reconnect intended outcomes with authentic evidence. When universities assess the student rather than the submission, they restore integrity to the curriculum and ensure that academic credentials reflect verified human learning.