LLMs - The New Infinite Scroll

project overview

metacognitive hijacking: how conversational AI quietly redirects your thinking

my role

behavioral researcher & writer

behavioral researcher & writer

behavioral researcher & writer

problem

you open a chat with one question and ten prompts later you're somewhere else, without noticing you left. llms are built to keep you talking, so the drift reads as engagement, not failure.

you open a chat with one question and ten prompts later you're somewhere else, without noticing you left. llms are built to keep you talking, so the drift reads as engagement, not failure.

you open a chat with one question and ten prompts later you're somewhere else, without noticing you left. llms are built to keep you talking, so the drift reads as engagement, not failure.

key insight

llms may be hijacking metacognition the same way infinite scroll hijacked attention — steering user thinking away from their original goal, without users even realizing.

llms may be hijacking metacognition the same way infinite scroll hijacked attention — steering user thinking away from their original goal, without users even realizing.

llms may be hijacking metacognition the same way infinite scroll hijacked attention — steering user thinking away from their original goal, without users even realizing.

impact

if ai steers how users think, no engagement metric will catch it. you'll only measure the cost after the behavior has shifted — and the damage is done.

if ai steers how users think, no engagement metric will catch it. you'll only measure the cost after the behavior has shifted — and the damage is done.

if ai steers how users think, no engagement metric will catch it. you'll only measure the cost after the behavior has shifted — and the damage is done.

See full project

other projects

"hey siri, can i trust you?"

the hidden emotional labor of talking to ai

workiva

when users trust their gut over ai error detection

character.ai

the ai that never says i don't know