Study: ChatGPT's random number picks reveal human-like biases, not uniform distribution
By
adunk
6d ago· 5 min readenCode
100/100
Golden Brown
Bagelometer↗
If you only eat one bagel today, this is the bagel.
Score100TypeanalysisSentimentneutral
Summary
This research project investigates how ChatGPT (gpt-4.1) behaves when asked to "pick a random number between 1 and 100" over 10,000 trials. Unlike a true random uniform distribution, the LLM exhibits predictable biases, clustering around certain numbers (37, 73, 42, 69) and avoiding round numbers — mirroring human psychological patterns in random number generation. The project characterizes the distribution against a uniform baseline, revealing that LLMs trained on human text inherit human-like non-random behaviors.
Key quotes
· 4 pulledIf you ask a person to 'pick a random number between 1 and 100', they are remarkably predictable.
Answers cluster on 37 and 73, on 'messy' numbers, and on memes like 42 and 69, while round numbers are quietly avoided.
A true random generator would instead produce a flat, uniform distribution.
Does an LLM, which is trained on human text, behave like a fair die, or does it inherit human biases?
When asked to pick a random number between 1 and 100, ChatGPT does not follow a random uniform distribution - exmergo/research-chatgpt-guesses-between-1-and-100
