People talk a lot about lab-grown meat but to me one of the most fascinating methodological advances I’ve seen recently is lab-grown conversations.
Conversation
People talk to each other in all kinds of ways for all kinds of reasons - imagine trying to study real conversations and keep track of everyone’s context and what level of irony they were on!
1
39
So instead you just engineer conversations between strangers in a lab, on the topic of answers to trivia questions (to minimize context and maximize legibility), and generalize that to ALL human conversation.
1
1
31
For example, if these lab-grown conversations don’t generate much more accurate answers to the trivia questions, it means conversations in general are not very useful.
1
1
29
The best part is they don’t even have to study the lab-grown conversations at all - they can just study the trivia question answers and answers to survey questions about confidence.
1
24
Constructed objects are very science-friendly because they’re much more likely to pop out a replicable regularity than their more complex real-world equivalents. It’s really a brilliant strategy: narrow, then generalize.
2
4
46
Replying to
it’s a narrow road when you’re hyper-agreeable and conflict-averse but also think certain methodologies are importantly wrong, this is an experiment to see if I get in trouble
2
3
Show replies


