You Already Think Computationally: Legal Reasoning Meets the Machine
You have done computational thinking since your first moot — this module shows you how to wield AI with it, without becoming the lawyer in Mata.
The hook
“A New York lawyer told a judge that ChatGPT "could not possibly be fabricating cases." It had invented six. The $5,000 sanction is the cheap version of this mistake; in India in 2026 the Supreme Court warned that leaning on fake AI-generated judgments would be misconduct.”
What you'll be able to do
- Establish the core mapping that the four pillars of computational thinking are skills you already practice as legal reasoning (decomposition=IRAC, pattern recognition=analogy from precedent, abstraction=ratio decidendi, algorithm design=applying a legal test).
- Set the course promise and the stakes: using AI well is now a core lawyering skill, and using it badly is a sanctionable, career-ending risk.
- Introduce the Ladder of Misinference (statement to fact to data to evidence to proof) as the course-wide answer-decoder for every AI output.
- Recognize how priming and anchoring move both human estimates and LLM outputs, so you never smuggle the answer you want into a prompt.
In short
Module 1 lays the intellectual spine of the course: legal reasoning IS computational thinking. It opens with the cautionary cases that define the stakes (Mata v. Avianca; Gummadi Usha Rani; the Bombay HC cost order), maps the four CT pillars onto skills students already use, and installs the Ladder of Misinference as the tool for decoding any AI answer. The flagship Mirror Move demonstrates anchoring on the room and then on a chatbot, teaching the first prompt-craft discipline.
The AI bridge
The entire course is framed as one promise: how not to be the lawyer in Mata. Once students see that the four CT pillars are skills they already wield as legal reasoning, prompting and interrogating AI stops feeling foreign — it becomes an extension of issue-spotting, analogy, abstraction, and applying a test. The Ladder of Misinference becomes the reusable decoder they will apply to every AI answer for the rest of the course.
In this module
- 01
The cold-open cautionary cases set the stakes: Mata v. Avianca (S.D.N.Y. 2023) where ChatGPT fabricated six cases and a lawyer trusted that it could not be inventing them; Gummadi Usha Rani (SC of India, a pending 2026 SLP) where the Court observed that a decision built on fake AI-generated judgments would be misconduct; and the Bombay HC Rs 50,000 cost order for phantom precedents. This is why decoding an AI answer is a professional duty, not a nicety.
- 02
The four-pillar mapping (Section 4): decomposition is issue-spotting/IRAC; pattern recognition is reasoning by analogy from precedent; abstraction is extracting the ratio decidendi; algorithm design is applying a legal test step-by-step. The payoff: you already think computationally, so you can already prompt AI well once you see the bridge.
- 03
Two further resonances: stare decisis behaves like caching/memoization (don't re-derive settled law), and the canons of statutory interpretation are heuristics a system applies — reinforcing that legal method and computational method are the same muscle, now pointed at AI.
- 04
Why 'use AI well' is now a core lawyering skill: AI is already in Indian courts and firms, so the question is no longer whether to use it but how to use it effectively and responsibly — prompting it, interrogating its output, verifying it.
- 05
The Ladder of Misinference (Edmans, May Contain Lies) is installed as the course spine: statement to fact to data to evidence to proof. A hallucinated citation is a statement costumed as binding proof; every later module climbs this ladder on AI output.
- 06
Edmans's warning that smart people are better at biased search lands hard for top law students: intelligence and verbal fluency do not protect you from a confident wrong answer — they help you rationalize it. The remedy is disciplined verification, not cleverness.
The interactive demos
Every idea is a Mirror Move
Run it on the room, show it inside the machine, prove it live on a real AI, then name the skill.
Anchoring / Priming on a Legal Estimate
On us
Split the room (Group A/B style) and prime half with a low anchor and half with a high anchor before a legal estimation question (e.g., what percentage of bail applications in a given category succeed, or what damages would you expect), then plot the two group averages to reveal the gap the anchor created.
In the machine
LLMs are just as suggestible to a number or a framing planted in the prompt — the model drifts toward whatever value or stance you seed, exactly as the two primed halves of the room did.
Live AI
Give a chatbot a high anchor and then, fresh, a low anchor before the same estimation question, and watch the answer move with the anchor.
The skill
Never smuggle the answer you want into your prompt — and beware leading framing in your own legal analysis. A neutral prompt is the first verification discipline.
The lab
Spot the Rung
Students take three confident statements — one made by a human, one drawn from a case headnote, and one AI-generated — and climb the Ladder of Misinference on each, classifying where each claim actually sits (statement, fact, data, evidence, proof) rather than where it presents itself.
Deliverable
A short worked classification placing each of the three statements on the correct rung of the Ladder, with a one-line justification per statement explaining why its apparent rung differs from its actual rung.
Key sources & cases
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)
Judge Castel; $5,000 sanction; six fabricated cases (Varghese, Martinez, Shaboon, Petersen, Miller, Estate of Durden); the lawyer's fatal assumption that ChatGPT could not be fabricating cases. The flagship cautionary tale and the course's framing device.
Gummadi Usha Rani v. Sure Mallikarjuna Rao (SC of India, SLP (C) No. 7575/2026, Narasimha & Aradhe JJ.)
A pending Special Leave Petition in which the Supreme Court, on a trial court order built on fake AI-generated judgments, observed that such a decision "would be a misconduct and legal consequence shall follow" and issued notice. Not a final holding; the India-first anchor for why verification is a duty.
Bombay HC Rs 50,000 cost order — Deepak v. Heart & Soul Entertainment Ltd. (7 Jan 2026)
Cost order for relying on phantom precedents — a concrete Indian consequence for unverified AI citations.
Alex Edmans, May Contain Lies (2024)
Source of the Ladder of Misinference (statement to fact to data to evidence to proof) and the observation that smart people are better at biased search — pitch-perfect for sharp law students.
Readings
- Alex Edmans, May Contain Lies (2024) — the Ladder of Misinference and biased search
- Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)
- Gummadi Usha Rani v. Sure Mallikarjuna Rao (SC of India, 2026)
- Bombay HC Rs 50,000 phantom-precedent cost order (2026)
Next module
Module 02 / 08
How the Machine Learns (and Why It Invents Cases)
Demystify the machineBring a credit-bearing AI course to your students
A 16-hour course that treats using AI well as a professional duty — one a council can approve, and a graduate can defend in court.