3 ms·
They're arguing that a jump from 1% to 2% of total speech is enough to tell if someone is anxious or depressed: https://imgur.com/a/Eu5sR https://imgur.com/a/Eu
by fny 9y ago
They're arguing that a jump from 1% to 2% of total speech is enough to tell if someone is anxious or depressed: https://imgur.com/a/Eu5sR https://imgur.com/a/Eu5sR
While that might be statistically significant, it's practically useless. How on earth can anyone even perceive this day to day?
Also, there's a lot of potential for the results to be drowned out by bias in the study design and other confounding factors.
I really don't think the evidence supports the hypothesis in any meaningful way.
- dragonwriter 9y ago> While that might be statistically significant, it's practically useless. How on earth can anyone even perceive this day to day? An always-on, always-on-hand AI assistant might do it fairly easily.
- buttscicles 9y agoAlso Facebook, Twitter, Tumblr, etc. Sites and apps with lots of user generated text
- nerfhammer 9y agodifference of 1% to 2% = twice as often. Suggests trying to use those words half as frequently. If you see yourself using one of them, flip a coin and leave it be if it lands on tails or reconsider it if it lands on heads. (insert usual caveats about direction of causality here)