conversation-topics

An open dataset of 910 conversation topics, icebreakers, debate motions, ESL speaking prompts, deep questions and writing prompts, in one JSON file, plus a tiny dependency-free picker that never repeats a topic within a session.

Every entry is hand-written and tagged by category (what you use it for), audience (who it suits) and depth (light, medium, deep). Debate and speech topics come with talking points; ESL prompts come with follow-up questions.

• Data: data/topics.json, one topic per line. Use it from any language.
• Code: src/index.js, ESM with TypeScript types. Node 20.10+ or a modern browser.

What's inside

category	general	students	kids	esl	workplace	total
icebreaker	112	67	78	0	44	234
conversation	369	263	78	60	4	511
deep-question	49	0	0	0	0	49
debate	308	308	8	18	4	338
speech	296	296	0	0	0	296
writing-prompt	70	53	6	0	0	76
journal	95	95	0	0	0	95

A topic can have several categories and audiences, so columns don't add up to 910.

• Audiences: general (adults, mixed groups), students (high school and college: debate, speech, essays), kids (roughly ages 6 to 12, classroom and family), esl (English learners; depth doubles as level: light = beginner, medium = intermediate, deep = advanced), workplace (team meetings and stand-ups).
• Tags: a subject for the debate/speech/conversation topics (science, technology, philosophy, psychology, history, art-culture, food-travel, relationships, education, politics, entertainment, sports, business, nature, health, weird-fun) or a theme for the rest (classroom, team, reflection, fiction, gratitude, self-discovery, beginner, ...).

Schema

{"id":"ct-0001","text":"The ethics of gene editing in human embryos","categories":["conversation","debate"],"audiences":["general","students"],"depth":"deep","tags":["science"],"talkingPoints":["CRISPR technology capabilities","Designer babies controversy","Medical vs enhancement uses"]}
{"id":"ct-0728","text":"What do you usually eat for breakfast?","categories":["conversation"],"audiences":["esl"],"depth":"light","tags":["beginner"],"followUps":["Do you eat at home or on the way?","What did you eat for breakfast as a child?","What is a typical breakfast in your country?"]}

field	type	notes
id	string	ct-0001 ... stable; never reused
text	string	the topic or question
categories	string[]	icebreaker, conversation, deep-question, debate, speech, writing-prompt, journal
audiences	string[]	general, students, kids, esl, workplace
depth	string	light, medium, deep
tags	string[]	subject or theme, lowercase
talkingPoints	string[]?	angles to cover (debate, speech, discussion)
followUps	string[]?	follow-up questions (ESL prompts)

Picker

git clone https://github.com/jackships/conversation-topics.git
cd conversation-topics
npm test
node examples/cli.js --category icebreaker --audience workplace -n 3
node examples/cli.js --audience esl --depth medium --seed lesson-4

import { random, filter, createSession } from "./src/index.js";

random({ category: "debate", audience: "students" });
// { id: "ct-0233", text: "Standardized testing: measuring knowledge or test-taking ability", ... }

random({ category: "deep-question", seed: "2026-10-07" }); // same seed, same topic

filter({ audience: "kids", category: ["icebreaker", "writing-prompt"] }).length; // 84

// No repeats within a session, even across different filters
const session = createSession({ seed: "class-3b" });
session.random({ audience: "esl", depth: "light" });
session.take(5, { category: "icebreaker", audience: "workplace" });
session.remaining({ audience: "esl" }); // how many are left unused
session.reset();

function	does
random(filters & { seed? })	one matching topic, or null. Stateless.
filter(filters)	every matching topic.
createSession({ seed?, recycle? })	a picker with random(filters), take(n, filters), remaining(filters), reset(). It returns null once the matching pool is used up, or starts over if recycle: true.
rng(seed)	the seeded PRNG used internally (FNV-1a + mulberry32).

Filters: category, audience, depth, tag (each one value or an array, matched as "any of") and search (case-insensitive substring). Unknown category, audience or depth values throw a RangeError, so typos don't silently return nothing.

Where the topics come from

All entries were written by hand. About 730 of them are the curated lists behind randomtopics.app; the ESL, kids, workplace, deep-question and story prompts were written for this release. Nothing is scraped or machine-generated in bulk. See CONTRIBUTING.md to add or fix topics.

License

• Code (src/, examples/, test/): MIT © 2026 Jack Ships
• Data (data/topics.json): CC BY 4.0. Use it anywhere, including commercially; credit "conversation-topics by Jack Ships" with a link to this repo.

---

Made by Jack, who also runs randomtopics.app, a free random topic generator built on these lists.