Tessera
Folgen

manushka sondhi 📉

@whatsonian.bsky.social

interested in how causality and a lack thereof guide our understanding of films/TV 🎬 Communication PhD Student @ Michigan State

4 Posts37 Follower119 Folgt
Jacob T. Fisher@jacobtfisher.online · 5mo ago

Hey y'all! Very pleased to announce that we'll be hosting Communication Science Futures again in East Lansing later this year. The conference in 2024 was a blast, and we're looking forward to running it back! Keynote speaker: James Pennebaker Submission deadline: June 5th. → commscifutures.com

Home | Communication Science Futures

Communication Science Futures is a conference aimed at addressing pressing issues and opportunities in the interdisciplinary, social...

commscifutures.com

Mark Thornton@markthornton.bsky.social · 6mo ago

New preprint from Lindsey Tepfer (@ltjaql.bsky.social) and me! We silenced portions of internal monologues in two films to manipulate participants' access to characters' thoughts. Using ISC and RSA, we found that this aligned later neural processing of the narrative & encoding of trait impressions.

Figure 1. Experimental stimuli and paradigm. 1A. Manipulation timeline. Each video features a total of 12 IM moments, half of which were silenced in version 1, and the other half silenced in version 2. This produces two versions of each video where only half of the IMs are audible to the participant. 1B. Voice removal. Background sounds were preserved in both versions, regardless of IM presence. 1C. fMRI paradigm. Participants were randomly assigned a version and watched both videos in a counter-balanced order. 1D. Afterwards, they made trait ratings on each video’s main character in random order before rating them on subsequent traits. Figure 2. Scanner and online participant trait rating results. A. Scanner participants arrive at similar conclusions about characters across versions by the end of each video. B. Trait ratings at the end of the videos are correlated between online and scanner participants. C. The IM manipulation had a significant effect on the trait ratings across the duration of the videos, such that different versions led to different trait impressions. D. Participants who saw different versions changed their ratings for both clip types but changed to a greater degree after seeing the IM segments relative to the NIM. Figure 3. ISC results. Gaining access to the same mental state knowledge in earlier IM clips led to significantly (p < .05, corrected) more aligned neural activity in subsequent NIM clips across a wide swath of the temporal lobe, as well as the frontal pole and SPL. Figure 4. Representational similarity analysis results. 4A. RSA pipeline. Clip-specific patterns of brain activity were correlated across movies within segment type (NIM or IM) to create a correlation distance matrix for each participant. Corresponding trait ratings were partitioned into components specific to each of the two versions of each film or shared across both versions. RDMs based on those traits were used to predict the neural RDMs. The resulting coefficients were sorted into matched (version), unmatched (version), and shared across version, and inference was performed via t-tests across participants. See methods section for further details of this analysis. 4B. Prior trait rating results. 1. While watching IM clips of the same version, the prior trait ratings predict activity in the bilateral A1, STS, STG, and right VLPFC. 2. Prior trait rating predicts activity in these regions again in the same-version NIM clips. 3. Across the two versions, while watching the NIM clips, the prior trait ratings predicted activity in the right occipitotemporal cortex. 4C. Trait updating results. 1. While watching NIM clips, the right STS & MTG predict version-matched trait updating. 2. In the unmatched NIM clips, STS & MTG again predict trait updating. 3. Across the two versions, the precuneus and LOC show pattern similarity for trait updating in participants.
PsyArXivBot@psyarxivbot.bsky.social · 6mo ago

Access to others' internal monologues aligns neural processing of narratives and trait impressions: https://osf.io/7v9md

Rebecca Frazer@beccafraz.bsky.social · 8mo ago

Out now in Scientific Reports! Despite high correlations, ChatGPT models failed to replicate human moral judgments. We propose tests beyond correlation to compare LLM data and human data. With @mattgrizz.bsky.social @andyluttrell.bsky.social @chasmonge.bsky.social www.nature.com/articles/s41...

ChatGPT does not replicate human moral judgments: the importance of examining metrics beyond correlation to assess agreement - Scientific Reports

Scientific Reports - ChatGPT does not replicate human moral judgments: the importance of examining metrics beyond correlation to assess agreement

nature.com