Convention record: SPSP2025-pre-conference
Conference Records 2025-02-20:
Morning sessions
Conference Name: SPSP pre-conference, Computational psy
Location: Denver
Session name: Keynote session 1
Topic: Theory-driven modeling
Time: 8:30 am
Speaker: Robert Hawkins(https://rdhawkins.com/)
Details and Comments
Keywords: social interaction
- Why computational models
- Hidden assumptions/fancy analysis/theory buildings
- Social interaction meets Shanon's model: Signal = informativity + cost
- Social interaction → feedback channel
- ⬆️ social cognition ⇄ group interactions
Comments:
- pre-print paper: https://arxiv.org/pdf/1912.07199, they used cos distance to describe the similarity with CLSlized texts
Session name: Data Blitz 1
Topic: Conversational Launch Pads
Time:
Speaker: Emma Templeton
Details and Comments
Keywords: interpersonal communication
- Input: raw conversations
- output: topic clusters → she calls them as launch pads
Topic: Facial Stereotyping Relies on a Dynamic and High-Dimensional Trait Space
Time:
Speaker: Gabriel Fajardo
Details and Comments
Keywords: facial trait
- Inputs: images and traits spaces (labels)
- Models: RCN (recurrent connectionist networks)
- Outputs: traits
Comments: is it possible to conduct a reverse analysis?
Topic: Environmental variability makes cultural transmission more model-based
Time:
Speaker: Xavier Roberts-Gaal
Details and Comments
Keywords: cultural transmission
- Check whether environmental variability moderate the optimal way to cultural transmission
Comments: check the work
Session name: Keynote session 2
Topic: Data-driven discovery
Time:
Speaker: Madalina Vlasceanu
Details and Comments
Keywords: climate cognition
- Psychological distance matters
- study1: Individual climate cognition (value) datasets
- study2: collective actions
- study3: votes
Comments: traditional data exploration
Session name: Data Blitz 2
Topic: Uncertainty, NLP, and Wisdom: The effect of uncertainty appraisal on meta-cognition
Time:
Speaker: Niyati Kachhiyapatel
Details and Comments
Keywords: uncertainty
- data are collected from existed project
Comments: the usage of NLP is unknown → they used cosine similarity
Topic: Moralization has increased during the social media era
Time:
Speaker: Curtis Puryear
Details and Comments
Keywords: Moralization
- inputs: data scrapped from X
- Model: BERT pre-trained for identify moralization
- outputs: moralization index? (similarity)
Comments:
- check threat amplification theory
- how about using time series transformer model to dig more
Topic: Lifestyle Polarization on a College Campus: Do Liberals and Conservatives Behave Differently in Everyday Life?
Time:
Speaker: Sanaz Talaifar
Details and Comments
Keywords: lifestyle polarization
- perceived lifestyle
Comments:
- interesting, but how computational
Session name: Keynote3
Topic: Analysis of Bias
Time:
Speaker: Gandalf Nicolas
Details and Comments
Keywords: LLMs
- RQ: missing evaluation on multiple dimensional stereotypes
- Inputs: prompts
- Models: ChatGPT 3.5 turbo
- Outputs: outputs by chatGPT
- Analysis on bias: by using text embedding analysis SBERT
- K-means: clustering
- Compared with human LLM rated people as more positive -> limitation of training data (due to safeguards)
- Prevalence vs. Valence
- Prevalence matters
Comments:
- Oh, he is co-researching with Fiske
Session name: Data Blitz 3
Topic: People who share encounters with racism are silenced online by humans and machines
Time:
Speaker: Cinoo Lee
Details and Comments
- flagged posts vs. non-flagged posts
- Key features: social identity threat/tone policing
Topic: Networks of beliefs: An integrative theory of individual-and social-level belief dynamics
Time:
Speaker: Jonas Dalege
Details and Comments
Keywords: belief dynamics
Comments: definitions are mathematically correct?
Afternoon sessions
Conference Name: SPSP pre-conference, media and technology
Location: Denver
Session name: null
Topic: NA
Time: NA
Speaker: NA
Details and Comments
- Almost all studies presented this year are not in my interests
Daily Summary
What's new
- NLP studies become popular this year, however, there is no new techs updated
What should I check
- Those NLP papers focus on the cosine similarity, try some new distance and test again
- Check if it is possible to use OT or persistent homology to describe the feature space of the point clouds