Helen Schmidt
@helen-schmidt.bsky.social
PhD candidate @ Temple studying emotions & social cognition
NIH F31 fellow • ggplot enthusiast • she/her • https://helenschmidt.info/
NIH F31 fellow • ggplot enthusiast • she/her • https://helenschmidt.info/
🎉 The upshot: Even casual "small talk" has deep structure! Conversations follow predictable patterns as people jointly navigate semantic space to build connections.
Check out our full paper & data here: osf.io/69yw2/
#CogSci #Conversation #NLP
Check out our full paper & data here: osf.io/69yw2/
#CogSci #Conversation #NLP
June 21, 2025 at 9:56 PM
🎉 The upshot: Even casual "small talk" has deep structure! Conversations follow predictable patterns as people jointly navigate semantic space to build connections.
Check out our full paper & data here: osf.io/69yw2/
#CogSci #Conversation #NLP
Check out our full paper & data here: osf.io/69yw2/
#CogSci #Conversation #NLP
🔍 Why does this matter? It suggests strangers aren't just chatting randomly – they're systematically exploring "topic space" together, starting with safe common ground before venturing into more personal territory.
Think of it as conversational foraging!
Think of it as conversational foraging!
June 21, 2025 at 9:56 PM
🔍 Why does this matter? It suggests strangers aren't just chatting randomly – they're systematically exploring "topic space" together, starting with safe common ground before venturing into more personal territory.
Think of it as conversational foraging!
Think of it as conversational foraging!
🌊 Key finding #2: Conversations have NESTED specificity! Both within individual topics AND across entire conversations, people start general and get increasingly specific – like waves within waves – demonstrating fractal conversation structure. 🌀
June 21, 2025 at 9:56 PM
🌊 Key finding #2: Conversations have NESTED specificity! Both within individual topics AND across entire conversations, people start general and get increasingly specific – like waves within waves – demonstrating fractal conversation structure. 🌀
🎯 Key finding #1: Topics aren't random! We found systematic patterns in how people transition between conversation themes. Some topics cluster at conversation beginnings (introductions), others toward the end (politics, relationships).
June 21, 2025 at 9:56 PM
🎯 Key finding #1: Topics aren't random! We found systematic patterns in how people transition between conversation themes. Some topics cluster at conversation beginnings (introductions), others toward the end (politics, relationships).
🔬 Our approach: We had 1,505 people annotate topic shifts in stranger conversations from the CANDOR corpus. Then we used semantic embeddings to track how meaning changes throughout conversations.
The validation was solid – annotated topic shifts = sharp drops in semantic similarity!
The validation was solid – annotated topic shifts = sharp drops in semantic similarity!
June 21, 2025 at 9:56 PM
🔬 Our approach: We had 1,505 people annotate topic shifts in stranger conversations from the CANDOR corpus. Then we used semantic embeddings to track how meaning changes throughout conversations.
The validation was solid – annotated topic shifts = sharp drops in semantic similarity!
The validation was solid – annotated topic shifts = sharp drops in semantic similarity!
🤔 The big question: When people chat, are they just randomly jumping between topics? Or is there actually a method to the conversational madness?
We suspected conversations have a fractal-like structure – getting more specific at multiple nested levels.
We suspected conversations have a fractal-like structure – getting more specific at multiple nested levels.
June 21, 2025 at 9:56 PM
🤔 The big question: When people chat, are they just randomly jumping between topics? Or is there actually a method to the conversational madness?
We suspected conversations have a fractal-like structure – getting more specific at multiple nested levels.
We suspected conversations have a fractal-like structure – getting more specific at multiple nested levels.
#30DayMapChallenge Day 5: A Journey 🏃🏼♀️
Such a fun challenge to recreate the route from my first time running the Cherry Blossom 10 Mile race in D.C.!
Such a fun challenge to recreate the route from my first time running the Cherry Blossom 10 Mile race in D.C.!
November 17, 2024 at 9:01 PM
#30DayMapChallenge Day 5: A Journey 🏃🏼♀️
Such a fun challenge to recreate the route from my first time running the Cherry Blossom 10 Mile race in D.C.!
Such a fun challenge to recreate the route from my first time running the Cherry Blossom 10 Mile race in D.C.!
Day 3: Polygons 🧱
Thoroughly enjoyed creating my own Lego bricks!
Thoroughly enjoyed creating my own Lego bricks!
November 3, 2024 at 10:19 PM
Day 3: Polygons 🧱
Thoroughly enjoyed creating my own Lego bricks!
Thoroughly enjoyed creating my own Lego bricks!
Day 2: Lines 📏
Ancient geoglyphs mapped in modern day R
Ancient geoglyphs mapped in modern day R
November 3, 2024 at 10:06 PM
Day 2: Lines 📏
Ancient geoglyphs mapped in modern day R
Ancient geoglyphs mapped in modern day R
Day 1: Points 📍
November 3, 2024 at 10:05 PM
Day 1: Points 📍