It travels through time... differently

See samplr package: lucas-castillo.github.io/samplr/

with @pleonv.bsky.social, Johanna Falbén, Nick Chater and @asanborn.bsky.social. Thank you!
(8/8)

If you use randomness in cognitive models: if you were using it as a catch-all for unexplained variance then keep at it, but if your model postulates that people use random draws then consider the time frame used and if lower than 2s consider autocorrelated noise instead. (7/8)

If you like to be creative, explore the world, make good choices, be protected from agents exploiting patterns in your behavior: this is good news! We now know that the strategy of behaving randomly is available to humans and doesn't need much time, which is useful in all these domains.

(6/8)

We found people's sequences are random if 2-4s elapse between items! 🤯🤯

In the experiment we ran we asked people to do the task at two different speeds and so we could test whether it's time that matters (as in the weather) or number of items (as in card shuffles). The answer: time ⌛⌛⌛(5/8)

We analysed previous data from experiments asking people to generate sequences at random, and did our own experiment: we looked at altered versions of the sequences where we skipped some items (thinned sequences). This way we could evaluate sequences at different delays between items. (4/8)

We thought: actual random stuff isn't random instantaneously (the weather is unpredictable some time from now; a deck of cards needs a few shuffles). Maybe we haven't given people enough time? (3/8)

Being able to generate randomness would be quite useful (to avoid others taking advantage of patterns in your behavior, to be creative, to explore your environment...) -- BUT! Research on human random generation says people cannot do this (2/8)