Great interview! Can Adam share more about why text-based conversations don't work? I can totally see this; text is alienating and so much more open to interpretation. He says that about 20% end in conflict. Is there anything special in these 20% in terms of words used or the type of people who participate or any other factors?
1. Have you heard of Poe's Law? The idea is that online you can't parody a perspective in text alone that won't be understood by some readers as an honest defense of that perspective. That is, we lack the rest of the context to help us decode what people's intent actually is when they use text online: https://en.wikipedia.org/wiki/Poe%27s_law
2. I see dynamic conversations less as a "spaceship programmed to fly somewhere" and more as "steering a car". You need fast reflexes. You need to quickly jump in and say "wait wait wait, what are you saying here?". On the other hand, text is asymmetric in the time & attention the writer gives, versus the time & attention the reader gives. When we're in our own heads longer, but not listening longer, we have more room to increase the meta-perception gap.
3. There is a long tradition of skepticism over text. Socrates, in the Phaedrus, gives the following intuition: "writing is kind of like painting: painted figures look alive but say nothing when questioned. Written words seem intelligent, but when someone misunderstands them, the text cannot correct the reader. Only living speech, delivered by someone who knows you and the situation, can adjust and respond." Because text is easier, we can produce more noise, which can be offensive and defenseless.
Just some thoughts! But in general, we haven't seen something so special about those 20% that would make it easy for an AI to classify, for example - though we're trying.
Great interview! Can Adam share more about why text-based conversations don't work? I can totally see this; text is alienating and so much more open to interpretation. He says that about 20% end in conflict. Is there anything special in these 20% in terms of words used or the type of people who participate or any other factors?
Thank you Shreeharsh!
I have a few theories here:
1. Have you heard of Poe's Law? The idea is that online you can't parody a perspective in text alone that won't be understood by some readers as an honest defense of that perspective. That is, we lack the rest of the context to help us decode what people's intent actually is when they use text online: https://en.wikipedia.org/wiki/Poe%27s_law
2. I see dynamic conversations less as a "spaceship programmed to fly somewhere" and more as "steering a car". You need fast reflexes. You need to quickly jump in and say "wait wait wait, what are you saying here?". On the other hand, text is asymmetric in the time & attention the writer gives, versus the time & attention the reader gives. When we're in our own heads longer, but not listening longer, we have more room to increase the meta-perception gap.
3. There is a long tradition of skepticism over text. Socrates, in the Phaedrus, gives the following intuition: "writing is kind of like painting: painted figures look alive but say nothing when questioned. Written words seem intelligent, but when someone misunderstands them, the text cannot correct the reader. Only living speech, delivered by someone who knows you and the situation, can adjust and respond." Because text is easier, we can produce more noise, which can be offensive and defenseless.
Just some thoughts! But in general, we haven't seen something so special about those 20% that would make it easy for an AI to classify, for example - though we're trying.
Thanks for the reply! All of that makes sense and would love to know if you find anything in those 20%.