# Suggested storyline for the IIT Madras Academic Council talk

## Recommended title

**What Should a University Preserve When Intelligence Gets Cheap?**

Good alternates:

- **Education in the Age of Autopilot**
- **If a Bot Can Pass the Exam, What Are We Teaching?**
- **From Content and Control to Judgment and Trust**

## One-line thesis

AI is making routine intellectual execution cheap. That does **not** make education less important. It makes the university more responsible for teaching **judgment, verification, taste, and direction** - and for redesigning assessment so AI use is **visible, governed, and educational** rather than hidden and adversarial.

## Core metaphor and framework

### Core metaphor: education in the age of autopilot

The strongest anchor for this audience is **autopilot**.

- When autopilot arrived, aviation did not stop training pilots.
- It changed **what pilots are trained on**: simulation, instrument reading, exception handling, abnormal situations, checklists, and certification.
- AI is doing the same to education.
- If machines can do more of the routine flying, the human role shifts upward toward **judgment, responsibility, and intervention when things get weird**.

This gives the Council a reassuring frame: the goal is **not to surrender standards**, but to **redesign standards for a new operating environment**.

### Decision framework: enforce, level up, switch, accept

Use this as the recurring structure of the talk.

1. **Enforce**\
   What must students still be able to do without AI because it is foundational, safety-critical, or identity-forming?

2. **Level up**\
   What higher-order skill becomes more important because AI can do the lower-order version?\
   Examples: framing, verification, critique, taste, communication, judgment, problem selection.

3. **Switch**\
   What teaching or assessment format must change because AI has broken the old one?\
   Examples: from content delivery to question design, from output grading to evidence grading, from hidden AI use to observable AI use.

4. **Accept**\
   What skills have become commoditized enough that we should stop pretending they are scarce markers of academic value?\
   Examples: syntax recall, routine drafting, basic lookup, first-pass coding.

### End-state skills to emphasize

If you want a memorable triad to keep returning to, use:

- **Trust** - Can the student or institution stand behind the result?
- **Taste** - Can they tell good from plausible?
- **Direction** - Can they define the problem, choose trade-offs, and steer AI well?

## Narrative arc

### 1. Setup: every student now arrives with an invisible advisory cabinet

Open with a vivid scene:

> A student no longer walks into class alone. She arrives with a tutor, coder, editor, critic, translator, researcher, and lab assistant in her pocket.

Then ask:

> If that student sits for our exam, what exactly are we assessing?

This is a better opening than a generic "AI is changing education" line because it makes the issue immediate, human, and impossible to dismiss.

### 2. Complication: the old academic bargain is breaking

State the old bargain clearly:

- We built much of higher education for a world where intelligence was expensive.
- So we taught through repetition, assessed through unaided execution, and treated content delivery as a major faculty function.

Then introduce the break:

- AI can now draft, code, search, summarize, critique, and iterate.
- Traditional exams, assignments, and even vivas are easier to game.
- The middle of the apprenticeship ladder is collapsing: students can outsource execution before they have built judgment.

This is where you bring in the future-of-work insight:

- The scarce thing is no longer raw answer production.
- The scarce things are **trust, taste, attention, context, and direction**.

### 3. Revelation: this is not mainly a cheating problem - it is a redesign problem

This is the turn of the story.

The central claim:

> The question for a university is no longer "Did the student use AI?"\
> It is "What should we preserve, what should we raise, and what should we redesign when AI is available?"

Then land the autopilot metaphor:

- AI-free work still has value, but more like **manual flying** or **mental math**: useful as training, calibration, and confidence-building, not as the entire definition of competence.
- Human faculty matter more, not less - but less as answer-vending machines and more as **standards setters, directors, reviewers, coaches, and trust anchors**.
- When AIs disagree, that is often a **signal for escalation**, not a failure. Human judgment becomes more valuable exactly where the stakes, ambiguity, or accountability are high.

### 4. Evidence: your experiments are prototypes of the new university

Do **not** try to walk through all 50 innovations. Instead, use 4-6 representative experiments as proof points.

#### A. Make AI use visible instead of hidden

- `Ask AI` embedded in assessments
- model choice exposed to students
- telemetry on AI use and outcomes

Message: **do not drive AI underground; instrument it**

#### B. Redesign integrity instead of relying only on surveillance

- personalized or dynamic questions
- server-side validation
- repo-grounded viva
- execution logs, terminal recordings, verification notes

Message: **integrity can be engineered into the task, not only policed around it**

#### C. Move assessment upward from output to judgment

- test cases as tutors
- directional feedback
- mental-model diagnosis from code traces
- explanations, defenses, and recovery from mistakes

Message: **the gold standard becomes framed, verified, defended work**

#### D. Build governance into the system

- traffic-light / green-amber-red human-in-the-loop models
- policy documents becoming executable rules
- retrieval-first institutional memory
- shadow-mode and canary rollout for academic AI workflows

Message: **scale requires governance, not just enthusiasm**

### 5. Implication: the Council's role is to redesign the academic operating system

This should feel like the natural conclusion of the story.

The Council is not being asked:

- "Should IITM like AI?"
- "Should we ban AI completely?"
- "Can we preserve the exact same exam forever?"

The Council **is** being asked:

1. Where must IITM create deliberate **AI-free zones**?
2. Where must IITM shift standards upward to **judgment, verification, and defense**?
3. How should IITM make AI use **auditable and contestable**?
4. What new evidence should count as academic work: **specs, logs, prompts, citations, code history, oral defense, peer review**?
5. How do we train faculty and TAs to manage AI workflows with rigor?

### 6. Ending: position IITM as a leader, not a late adapter

End with ambition, not anxiety.

Suggested closing move:

- IITM already has the ingredients to lead: scale, engineering culture, online delivery experience, and respect for standards.
- So the opportunity is not merely to "adopt AI tools".
- It is to become one of the first universities to build a **trustworthy academic operating system for the AI era**.

Possible closing line:

> In the age of AI, the university's job is not to prove that students worked alone.\
> It is to ensure that they can frame problems well, use tools wisely, verify results rigorously, and stand behind what they submit.

## Key messages to repeat

These are the lines worth repeating in different forms throughout the talk:

1. **AI is not only a cheating problem. It is a curriculum, assessment, and governance redesign problem.**
2. **When intelligence gets cheap, judgment gets expensive.**
3. **Do not hide AI use. Instrument it.**
4. **The new academic gold standard is not unaided output. It is framed, verified, defended work.**
5. **Faculty become less like broadcasters and more like directors, critics, and trust anchors.**
6. **AI-free work still matters - but as deliberate exercise, calibration, or safety check, not as the default for everything.**
7. **Institutions that govern AI well will have an advantage over institutions that merely prohibit it loudly.**

## How to make the talk engaging and memorable

### Use one memorable question throughout

Keep coming back to:

> **What should a university preserve when intelligence gets cheap?**

That question is philosophical enough for a Council, but concrete enough to drive decisions.

### Use strong contrasts

These contrasts will help the talk stick:

- **hidden AI use** vs **visible AI use**
- **grading outputs** vs **grading evidence**
- **content delivery** vs **problem framing**
- **policing** vs **designing for integrity**
- **lecturer as broadcaster** vs **faculty as director**

### Use only a few experiments, but make them vivid

Do not present the innovation inventory as a catalogue. Use a handful of examples that dramatize the shift:

- "I put `Ask AI` inside the exam."
- "I used telemetry to see which models improved grades."
- "I used repo-grounded viva and execution evidence."
- "I used tests not just to grade but to teach."
- "I treated policy and workflow as machine-readable rules."

Those are memorable because they sound counterintuitive, concrete, and evidence-backed.

### Use the autopilot metaphor visually

A single recurring slide can do a lot of work:

- left side: **old university** - content, recall, hidden AI, output grading
- right side: **AI-era university** - questions, verification, visible AI, evidence grading

And one simple 4-box graphic for:

- **Enforce**
- **Level up**
- **Switch**
- **Accept**

### Give the audience relief, not just alarm

The talk should not sound like "everything is broken".

It should sound like:

- the disruption is real
- some old assumptions are breaking
- but there is a credible design response
- and you have already been prototyping pieces of that response

That combination of danger plus design will make the talk persuasive.

## Suggested section-by-section flow

1. **Hook:** every student now has an advisory cabinet in their pocket.
2. **Tension:** if a bot can pass the exam, what exactly are we teaching?
3. **Metaphor:** AI is autopilot for education.
4. **Framework:** enforce / level up / switch / accept.
5. **Proof:** a few sharp experiments from your own teaching and assessment practice.
6. **Governance:** what the Council must now decide.
7. **Conclusion:** IITM can build a trustworthy academic operating system for the AI era.

## Bottom line

The talk should feel less like a tour of AI tools and more like a **governance argument**:

**When execution becomes cheap, the university's value shifts upward - from delivering content and policing behavior to cultivating judgment, building trust, and certifying responsible capability.**