Research article:
"Six Fallacies in Substituting Large Language Models for Human Participants." — (Some of the fallacies mentioned are: "anthropomorphizing AI systems," "treating LLMs as the average human," "substituting model data for human evidence," etc.)

Discussion on Bluesky:

"Ultimately," the paper's abstract notes, "the analysis supports conceptualizing LLMs as pragmatic simulation tools ... rather than as replacements for human participants."

 

when a study says they need to check my Reddit account's history to "verify personality"

 

Yep, there are some kinds of research studies and some software test projects that may want access to social media histories, or streaming media histories, or Amazon purchase histories, or bank account involvement. ... I decline.

 

tbh I don't really care about the watch history of streaming apps but Reddit history that goes back 10yrs is a bit much :emoji_grin:

 

Mine too is about a decade.

Amazon purchase history, 2 decades.

 

Actually I forget if some studies that ask for Amazon history want the purchase history, or the search history. One or the other. – Whatever. I always decline those.

 

Some other studies want Gооgle search histories.
There again, I'd be inclined to decline.

(One of those has been mentioned on this forum before,
in a review post with HIT details, so that confirms I'm not just imagining it.)

 

New Requester Profile!

MSach [A67S1NYX9PDOV] Spanish speech quality test | Evaluación de la calidad de voz en español - $0.40 | PANDA

Unrated

Unrated

Unrated

$4.46 / hour
00:05:23 / completion time

​

Pros:

Cons:

Catchability: Okay (HIT Last Seen: 9 minutes ago)
Qualifications:

To read MSach's full profile check out TurkerView!​

 

I went to the MTurk page and it prompted me to log in.
What the hell is this? Hope it's not the new captcha.