Faculty judgment beyond the textbook
Modified
Input
AI faculty must do more than transmit correct theory or recount professional experience; they must connect the two without corrupting either. The distinctive faculty contribution is the selection, sequencing, and criticism of models under realistic conditions. “Practical” teaching is rigorous when experience is converted into inspectable cases, assumptions, and evidence.
The Faculty Problem
An AI program can recruit distinguished researchers, experienced professionals, or skilled teachers. None of those labels independently guarantees strong education.
A researcher may know the frontier but struggle to decompose it for learners. A practitioner may possess valuable experience but generalize from one institution. A polished lecturer may explain supplied material without being able to redesign it for a new problem.
The school needs faculty who can translate among:
That translation is faculty judgment.
A Complementarity Model of Teaching
Let faculty educational contribution be
where:
- $K$ is command of foundational knowledge;
- $R$ is active research and model criticism;
- $P$ is relevant professional or empirical experience;
- $T$ is pedagogical capability; and
- $A$ is the ability to integrate them.
The multiplicative form captures bottlenecks. Experience without theory can become anecdote. Theory without empirical contact can conceal measurement and operational constraints. Research without pedagogy can overwhelm students. Pedagogy without sufficient subject command can make a weak model sound clear.
The institution should therefore evaluate the whole product, not one credential.
What “Beyond the Textbook” Means
Teaching beyond the textbook does not mean dismissing textbooks.
A good textbook supplies definitions, derivations, canonical examples, and a stable sequence. It preserves disciplinary knowledge and protects a course from becoming one instructor's improvisation.
Faculty judgment adds four things:
- Selection: Which parts matter for the program's outcomes?
- Connection: How does one formal result relate to other disciplines and decisions?
- Boundary: Under which conditions does the textbook solution fail?
- Reconstruction: How should the model change in an unfamiliar setting?
Table 1. Forms of faculty judgment in AI education
| Textbook contribution | Faculty contribution |
|---|---|
| Canonical model | Reason the model belongs to this problem |
| Standard assumptions | Evidence for and against those assumptions |
| Worked example | Perturbed case and alternative interpretation |
| Correct procedure | Decision about whether the procedure is appropriate |
| Known result | Connection to current empirical or professional conditions |
| Exercises | Feedback that diagnoses the student's reasoning |
The textbook is a foundation. The faculty member makes it a living intellectual system.
From Experience to a Teachable Case
Professional experience is not automatically educational material.
Suppose an instructor encountered repeated unauthorized login attempts. The raw incident contains confidential details, operational noise, and hindsight. To become a case, it must be reconstructed:
- define the observational unit;
- distinguish normal traffic from an attack hypothesis;
- identify base rates and alternative explanations;
- specify which information was available at each time;
- construct features without leakage;
- define costs of escalation and non-escalation;
- remove identifying information; and
- create evidence that supports more than one defensible initial answer.
The decision may be written
The instructor's experience supplies realism. Formalization turns it into education.
The Selection of Examples Is an Intellectual Act
Examples determine what students believe a method is for.
If logistic regression appears only in clean classification datasets, students learn that it is a software option. If it is used to examine a binary outcome, interpret odds, analyze selection, calibrate risk, and construct a decision threshold, students learn a family of reasoning.
A faculty member chooses an example to expose a specific relation:
An authentic case can be educationally weak if operational detail obscures the concept. A stylized case can be powerful if the simplification is declared and later relaxed.
Explaining Why, Not Only How
Consider standardization:
A procedural explanation shows the calculation. Faculty judgment expands the question:
- Why does scale matter for this objective?
- From which sample are $\widehat\mu_j$ and $\widehat\sigma_j$ estimated?
- What happens under a changing distribution?
- Is a standard deviation meaningful for a heavy-tailed feature?
- Does standardization change interpretation?
- Which models are scale-invariant?
The “why” is not an inspirational addition. It determines whether the operation belongs in the pipeline.
Faculty as Model Critics
Students often encounter polished results before they develop a habit of criticism. Faculty should make criticism observable.
An instructor can think aloud through a model:
- define the target;
- inspect the DGP;
- identify the strongest assumption;
- construct an alternative explanation;
- propose a falsification or stress test;
- compare the result with the decision; and
- narrow the claim.
The instructor should also expose uncertainty. Pretending that every case was obvious in advance teaches hindsight, not judgment.
A useful phrase is:
Given this information, the model is defensible for this claim; it is not yet defensible for the stronger claim.
Research Faculty and Practitioner Faculty
A specialized school benefits from both research and practice, but roles should not be caricatured.
Read as a diagnostic rather than a second scorecard, the framework becomes clear: Methodological researcher — Potential contribution: Formal depth and frontier criticism; Required safeguard: Connect method to empirical conditions. Domain researcher — Potential contribution: Measurement and substantive structure; Required safeguard: Make assumptions and computation explicit. Experienced practitioner — Potential contribution: Constraints, failures, implementation; Required safeguard: Distinguish evidence from local convention. Systems specialist — Potential contribution: Reproducibility and deployment; Required safeguard: Connect engineering choices to the claim. Case instructor — Potential contribution: Decision and communication; Required safeguard: Preserve technical depth.
The strongest courses may involve co-design rather than a sequence of unrelated guest appearances. Faculty need a shared vocabulary and one assessment standard.
Current Knowledge Without Fashion Chasing
AI faculty must update courses, but recency is not itself relevance.
Let a new method $m$ provide educational value
where $\Delta C$ is additional conceptual capability, $\Delta T$ transfer value, and $B$ the burden of displacing foundations or introducing unstable detail.
A new architecture belongs in a course when it reveals a durable idea, solves an important limitation, or materially changes a decision. It need not enter merely because it dominates current discussion.
Faculty judgment protects the curriculum from both stagnation and trend dependence.
The Work of Feedback
Faculty value appears most clearly after the lecture.
Two students can produce the same wrong answer for different reasons. One made an algebraic mistake; another defined the wrong target. Automated feedback may identify the first. The second requires a person—or a carefully designed system—to reconstruct the student's reasoning.
High-quality feedback identifies:
This work is time-intensive. Programs that evaluate faculty only by content delivery understate the resource needed for serious education.
Faculty Judgment Must Be Auditable
Expert judgment should not become unreviewable authority.
An institution can require:
- written learning outcomes;
- shared assessment rubrics;
- peer review of cases and examinations;
- documented reasons for major curriculum changes;
- multiple readers for dissertations;
- student evidence rather than popularity alone; and
- periodic comparison with external disciplinary standards.
The point is not to eliminate judgment. It is to make judgment inspectable and correctable.
Hiring for Educational Evidence
Faculty hiring should include evidence of teaching judgment, not only a list of publications or employers.
A candidate can be asked to prepare a short lesson from a supplied technical result and then respond to three perturbations:
- explain the result to a student missing one prerequisite;
- connect it to a realistic data-generating process; and
- revise the lesson after an assumption fails.
The evaluation can score
where $K$ is subject command, $P$ problem formulation, $T$ teaching structure, and $C$ response to criticism.
A polished sample lecture may demonstrate presentation and still conceal weak diagnosis. The perturbations reveal whether the candidate can teach outside a prepared script.
For practitioner faculty, the school should also ask what evidence would contradict the lesson learned from experience. For research faculty, it should ask how the formal result changes a professional decision. These questions test whether the candidate can cross the textbook boundary in both directions.
Faculty Development
Faculty also need education.
A researcher entering the classroom may need case design and feedback methods. A practitioner may need formal review and research ethics. A technical instructor may need training in assessment validity. A strong teacher may need protected time to update empirical knowledge.
Faculty development can use the same correction loop expected of students:
Teaching quality is not a fixed personality trait.
The SIAI GSB Faculty Principle
SIAI GSB's institutional standard should be that faculty live neither only inside nor only outside the textbook.
They should command formal knowledge, understand where it encounters data and institutions, and be able to turn that encounter into a disciplined learning sequence. Their cases should be real enough to expose constraints and formal enough to support criticism.
The purpose is not to make faculty the center of the school. It is to make their judgment part of a reproducible educational system.
Faculty judgment also includes the discipline to leave material out. A field moving as quickly as AI produces more papers, tools, and demonstrations than any coherent course can absorb. Selection must follow the learning objective: a new method belongs in the course when it clarifies an enduring problem, changes a relevant decision, or exposes a limitation in the existing framework. Novelty by itself is not a curriculum argument.
The need for faculty judgment is visible in recent evidence. Teacher AI Literacy Is the Real Test of AI in Education argues that institutions cannot develop student capability when instructors lack the confidence to evaluate AI-supported work. SIAI’s research on cognitive outsourcing places verification at the center of teaching design, while its review of agentic machine science shows why the boundary between checkable output and contextual judgment matters. Faculty value increasingly lies in selecting that boundary and making it teachable.
Conclusion
Faculty judgment beyond the textbook is the capacity to connect theory, evidence, practice, and learning.
It selects the right example, reveals the hidden assumption, converts experience into a case, and gives feedback at the layer where reasoning failed. It also knows when a fashionable method deserves curricular attention and when it does not.
A specialized AI school cannot be built from content alone. It needs faculty capable of explaining not only what a method is, but why it belongs, where it fails, and how a student can reconstruct it.
That is the difference between delivering information and educating independent practitioners.
References
Lee S. Shulman, “Signature Pedagogies in the Professions”, Daedalus 134, no. 3 (2005): 52-59.
Donald A. Schön, The Reflective Practitioner, Basic Books, 1983.
Allan Collins, John Seely Brown, and Susan E. Newman, “Cognitive Apprenticeship: Teaching the Craft of Reading, Writing, and Mathematics”, 1987.
The Economy Editorial Board (2026) ‘Teacher AI Literacy Is the Real Test of AI in Education’, The Economy Review, 22 June.
Swiss Institute of Artificial Intelligence (2026) ‘Cognitive Outsourcing in Education: Why AI’s Real Classroom Crisis Is Verification, Not Cheating’, SIAI Working Papers, 24 July.
Swiss Institute of Artificial Intelligence (2026) ‘What Agentic AI Can Prove and What It Still Fabricates: The Real Frontier of Machine Science’, SIAI Science Review, 15 August.
Similar Post
