Methods · History, human story & context · Short guide

Bibliography construction

Build a reading list and defend the sequence: so reading order is a designed path, not a random PDF pile.

When to use this: Before you start studying
Learning impact: Clear help, modest gain
Learning evidence: Thin or mixed evidence
Where you are with the topic: Meeting the idea
How often to use it: Build and keep as a habit

When to use it

You have a pile of PDFs and no honest order to read them.
You need a shortlist you can defend, not a fifty-item dump.
A supervisor asked how your readings connect, and you cannot say.

What to ask AI

Type something like: "Question: [ ]. Level: [ ]. Propose a sequenced bibliography with a role tag per item. Mark unverified. I will check existence and cut to [N]. Then grill me on why item 3 precedes item 4."

Never accept citations without checking they exist.

Keep the hard part on your side: you defend the sequence after you cut the list.

How it helps

A reading list is a path, not a pile.

Role tags (survey, landmark, critique, update) force each item to earn its place.

Defending why item three precedes item four is where the thinking happens.

How to run it

State the question and your level honestly.
Ask for a short sequenced list with a role per item.
Mark every unverified citation.
Check existence yourself and cut hard.
Defend consecutive pairs: why this order, not the reverse.

Example exchange

You

I need a reading path for how attention mechanisms changed NLP after 2017, advanced undergrad level. Can you propose a sequenced bibliography with a role for each paper, and mark anything you have not verified? I will cut it to six.

A workable path is survey, transformer landmark, early critique, extension, domain application, then a recent review. Two of those titles I would treat as unverified until you check them. Paste your cut list of six once you have checked DOIs.

You

Here are my six after checking DOIs.

Why does the critique sit before the domain application? Answer in one sentence before we continue.

Copyable prompt

Question: [ ]. Level: [ ]. Propose a sequenced bibliography with a role tag per
item. Mark unverified. I will check existence and cut to [N]. Then grill me on
why item 3 precedes item 4.

The Tell

Here is how you know this method has flipped: you have fifty titles and three opened.

A list the AI wrote for you, that you never sequenced or checked, is not a bibliography, it is a prop.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated weak-to-moderate. Information literacy and curated sequencing are sensible; named experimental evidence for particular reading orders is thin.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

Fabricated citations are a known failure mode. Delivery is only as trustworthy as your existence checks.