Richard Huskey

@richardhuskey

Associate Professor, UC Davis. PI, Cognitive Communication Science Lab. Associate Editor, Journal of Communication. Ski bum at heart.

1,773 Following2,244 Followers290 Posts

Bluesky

Is this your profile?Claim it and richardhuskey.dirty.social is yours.Claim profile

Richard Huskey@richardhuskey· 3d

7/n Curiosity matters, too. People who take joy in learning something new deliberately choose uncertain books, and are more likely to enjoy that choice. Thrill-seekers also explore uncertain books, but in a more random way.

Curiosity modulates exploration patterns in people’s book selection. A) Boxplots of cross-validated parameter estimates of the temperature parameter τ (left box plot) and the information bonus parameter
(right box plot) for each participant in the experimental dataset. Parameter estimates for each participant are shown as points displayed for each box plot and are horizontally jittered for improved visual interpretation. B) The partial regression plot of the logarithm of
regressed on joyous exploration; C) logarithm of τ regressed on thrill seeking; and D) logarithm of τ regressed on joyous exploration. The solid line indicates the estimated regression line, and the shaded area around the regression line represents the 95% CI of regression coefficient estimates. E) Line plot of reading enjoyment by the number of books that have been previously read in the experimental dataset, for participants high (darker colored line) and low (lighter colored line) in joyous exploration. Mean estimates are plotted as points, and 95% CI are plotted as the vertical lines. F) The regression models of subsequent book enjoyment regressed on exploration distance for participants high (darker colored line) and low (lighter colored line) in joyous exploration.

Richard Huskey@richardhuskey· 3d

6/n We found that people generalize from their past reads. Liking or disliking one type of book provides information about how much a future book will be liked or disliked. People also seek out books precisely because they don't know what they'll get.

Signatures of learning and reward generalization in real-world and experimental data. Top row A–C): Real-world Amazon dataset. Bottom row D–F): Controlled experiment dataset. A, D) Average reading enjoyment increases as a function of the number of books read. B, E) The exploration distance, measured as the Euclidean distance between semantic embedding vectors of consecutive book choices, by the number of books that have been read. C, F) The relationship between immediate reward and subsequent exploration: higher ratings for the preceding book are associated with smaller exploration distances in the subsequent choice. The points in the line plot indicate the mean estimates, and the vertical lines indicate the 99% CI. Note: distances are in arbitrary units, as Euclidean distances calculated within their respective spaces: a 384D semantic embedding space for the Amazon dataset and a 15 × 15 spatial grid for the experimental dataset.

Signatures of directed exploration in the real-world data, and model prediction accuracy in both real-world and experiment data. A) Exploration distance by the number of ratings for subsequent book choices. Mean estimates were plotted as points, and 99% CI were plotted as vertical lines. B) Bivariate distributions of exploration distance and the variance of ratings for subsequent book choices. Darker colors indicate higher probability density. The solid line is plotted as the regression line of variance of ratings regressing on exploration distance. The predictive accuracies for the GP and the UCB models for each subject are plotted for real-world dataset (C) and experimental data (D). Bar height indicates a mean estimate for the predictive accuracy (measured as pseudo-
⁠) for all subjects in the dataset, and the vertical line indicates the 99% CI of the predictive accuracy.

Richard Huskey@richardhuskey· 3d

5/n We ran two different studies. The first examined more than 2 million real-world choices from nearly 35,000 readers while a second experiment removed recommendations, bestseller lists, and other people's reviews.

The semantic space of books in the real-world and experimental book selection environment. A) Real-world books are represented in a semantic embedding space. A sample of 10,000 books was plotted in a 2D space, which is t-SNE transformed from a 384D semantic embedding space. Books are naturally clustered by their genre in this semantic embedding space, and together constitute a patchy book foraging environment. B) Readers make sequential book selection trajectories in the semantic book space. Highlighted points depict selected books, with ratings corresponding to the scale shown in Panel E, while faint background points represent books available but not selected. The arrows connecting two points denote the sequential order of consecutive book choices. C) Probability density plot of the distribution of people’s exploration distance (solid line curve), which is measured as the Euclidean distance between the semantic embedding vectors of consecutive book choices. The dotted line curve denotes the null distribution of exploration distance, which assumes people randomly select books. D) The experimental book selection landscape. A total of 15 × 15 options were arranged in a grid and presented for participants to make book selections. Each point encodes a book option, and each point is visually encoded to indicate the genre of the book. Book options were arranged so that semantically similar books are placed close to each other. E) Participants completed a total of 15 trials of a click-read-rate task, where they clicked one option from the grid, read the synopsis of the book, and then rated their reading enjoyment. F) Participants’ sequential book selection trajectories in the experiment.

Richard Huskey@richardhuskey· 3d

🧵In a bookstore, what section do you head to? The one you know you'll love? The one you're not so sure about? Algorithms help us make this choice. But what algorithms are our brains running when we choose for ourselves? In a study published in @pnasnexus.org, we tested this doi.org/10.1093/pnas...

Interior of a narrow, arched-ceiling bookstore or library lined with tall shelves of books. Handwritten signatures and notes cover the ceiling, while a large wooden beam spans the room and an open doorway at the far end lets in bright daylight, with several people visible inside and outside.

Photo: parretl via Pixabay https://pixabay.com/photos/library-old-book-book-store-books-934285/

Richard Huskey@richardhuskey· Jun 24

5/7 During EEG, we see widespread dynamic connectivity across the cortex. This connectivity is metastable; different parts of the brain become more or less synchronized over time. These metastable patterns are particularly clear in frequency bands associated with memory and cognitive control.

Functional connectivity analysis for the Theta band. The brain’s global functional connectivity network for boredom (A), flow (B), and anxiety (C) that demonstrates the spatial distribution of the dominant nodes and connections within the identified functional network. Node size indicates its hub score after centrality analysis, and the color line between nodes denotes the T-stat of the significant connections (n = 42). Significant functional connections were identified using cluster-based analysis (p < 0.025, cluster-level p < 0.05). The dFC analysis across the three experiences indicates the temporal activity of the identified functional networks: Root-Mean Square (RMS) (D), absolute (E) and variance (F) of Phase Difference Differential (PDD), and Structure Connectivity Rank (SCR) (G). In the violin plot, scatter points display the individual data point values, the horizontal line represents the mean, the white circle indicates the median, and the bar around the median denotes the interquartile range (IQR, q3-q1). All multiple comparisons are corrected using the post hoc Tukey tests (* p < 0.05, ** p < 0.01, *** p < 0.001).

Functional connectivity analysis for the Alpha band. The brain’s global functional connectivity network for boredom (A), flow (B), and anxiety (C) that demonstrates the spatial distribution of the dominant nodes and connections within the identified functional network. Node size indicates its hub score after centrality analysis, and the color line between nodes denotes the T-stat of the significant connections (n = 42). Significant functional connections were identified using cluster-based analysis (p < 0.025, cluster-level p < 0.05). The dFC analysis across the three experiences indicates the temporal activity of the identified functional networks: Root-Mean Square (RMS) (D), absolute (E) and variance (F) of Phase Difference Differential (PDD), and Structure Connectivity Rank (SCR) (G). In the violin plot, scatter points display the individual data point values, the horizontal line represents the mean, the white circle indicates the median, and the bar around the median denotes the interquartile range (IQR, q3-q1). All multiple comparisons are corrected using the post hoc Tukey tests (* p < 0.05, ** p < 0.01, *** p < 0.001).

Richard Huskey@richardhuskey· Jun 24

4/7 With fMRI, we find the brain has high entropy during flow. This may reflect unpredictability of flow-tasks. We also find that during flow, decreasing entropy is associated with increased flow experience. This might reflect mastery, or creativity, although it is still too soon to tell.

Regression of centered flow scores against entropy rate across the 3 conditions. In this analysis, the intercept represents the mean value of global entropy rate at the mean value of the flow scores.

Richard Huskey@richardhuskey· Jun 24

2/7 We know there's no flow neuron, no flow brain region, and no specific flow network. After all, how could something as experientially complex as flow reduce to just part of the brain?

Modes of coordination based on phase alignment. VIS, visual network; SMT, somatomotor network; DAT, dorsal attention network; VAT, ventral attention network; LBC, limbic network; FPA, frontoparietal; DMN, default mode network; SC, subcortical; CB, cerebellum.

Neurosynth meta-analysis of the coordination modes. The psychological concepts associated with the brain regions in each coordination mode are presented here in a word-cloud format. Word size is proportional to the region’s contribution to the phase alignment of the mode.

Richard Huskey@richardhuskey· Apr 18

Last day of #SANS2026! Today, @rachaelkee.bsky.social will present her poster introducing Inoxity. It’s an open source iOS app that allows for high-throughput collection of EMA and bio-behavioral data. What can you do with high-throughput data? Check her preprint! doi.org/10.33767/osf...

Research poster by Kee, Madgula, and Huskey introducing Inoxity, an open-source iOS platform for high-throughput data collection. Inoxity integrates passive sleep tracking, Apple Screen Time capture, and EMA surveys to study sleep–media dynamics in real-world settings. An N=1 proof-of-concept shows strong agreement with Apple Health sleep stage data.

Richard Huskey@richardhuskey· Apr 16

Also today at #SANS2026, Ziyu Zhao investigates the “brain rot” hypothesis. When studied rigorously, in large multi-national samples, we find a whole lot of nothing. Conclusion: previously observed alarming results might be driven by sampling bias and/or measurement error (P1-C-20)

Richard Huskey@richardhuskey· Apr 16

We’re at #SANS2026! Today, Valerie Klein will share her undergraduate capstone project using drift diffusion models to understand how mental health status influences affective media selection. Find her at P1-F-32, and check out the preprint, here doi.org/10.21203/rs....

Research poster by Klein, Gong, Eden, and Huskey using the drift-diffusion model to examine how depression, anxiety, and loneliness shape media selection in 313 undergraduates. Key finding: anxiety and depression are linked to adaptive, affect-regulating selection; loneliness is linked to maladaptive use.

Richard Huskey@richardhuskey· Mar 27

Psychometric profiles
Six-panel figure comparing Bot v7.3 (top, red) with a human participant (bottom, blue) on three reaction-time markers: QQ plot, SD-versus-mean RT, and autocorrelation. The bot and human show similarly right-skewed RT distributions and positive SD–mean relationships, but the bot’s QQ plot bends more sharply at the upper tail and its autocorrelation stays positive across many lags. The human shows a strong lag-1 autocorrelation that drops close to zero afterward, while the bot shows more sustained serial dependence.

Marker distributions across humans with bot overlaid
Four histograms show human distributions in light blue with a red vertical line marking the bot’s value for A) QQ correlation, B) skewness, C) SD–mean slope, and D) mean autocorrelation. The bot falls well inside the human range for QQ correlation, skewness, and SD–mean slope, usually near the middle of the distribution. For mean autocorrelation, the bot’s value is shifted to the high end, indicating more trial-to-trial dependence than most humans.

RT distributions
Two-panel figure of reaction-time histograms. Left panel overlays all human trials (blue, N=208,104) and bot trials (red, N=273): both peak around 400–600 ms, but the bot has a broader spread, including more very fast responses and a heavier slow tail extending past 1,000 ms. Right panel breaks the bot RTs down by flanker condition: congruent, neutral, and incongruent distributions overlap heavily, with incongruent trials tending slower overall and all conditions showing occasional very slow responses.

Richard Huskey@richardhuskey· Mar 27

1/n New preprint with Ziyu Zhao, @dougaparry.bsky.social, & @jacobtfisher.online Can an AI bot complete a live online reaction-time task & produce data that passes as human? We built an autonomous bot to take the Attention Network Test (ANT) in real time Preprint: doi.org/10.31234/osf...

Screenshot of a manuscript title page. Title: “An AI agent can complete the Attention Network Test with human-like behavioral signatures: Implications for the bot-or-not debate.” Authors: Richard Huskey, Ziyu Zhao, Douglas A. Parry, and Jacob T. Fisher, with university affiliations listed below. The abstract says an autonomous AI agent completed the Attention Network Test in real time and produced mostly human-like behavioral data. Across seven code revisions, the bot achieved attention network scores within published human norms, 95.8% accuracy, and reaction-time patterns showing positive skew and trial-to-trial autocorrelation. Compared with 796 human participants, the bot fell within the human range on several measures but showed elevated autocorrelation and a bimodal reaction-time distribution due to intermittent detection failures. The paper argues this makes simple bot-vs-human detection harder in online reaction-time studies.

Richard Huskey@richardhuskey· Mar 12

How do we deal with rich, multilevel, and multimodal data? In a new preprint, @rachaelkee.bsky.social and I sketch an answer! For all the details, and the preprint link, see Rachael’s thread. Be sure to give her a follow, especially if you’re interested in neuroscience, sleep, and media!

Three-panel infographic illustrating Marr’s Levels of analysis in a study of repeated exposure to crime media.
(a) At the top left, a diagram labeled “Marr’s Levels” shows three stacked boxes: 1. Computation (why / problem) in orange, 2. Algorithm (what / rules) in coral, and 3. Implementation (how / physical) in dark red.
(b) At the top right, an illustration of a person looking at a smartphone surrounded by social media icons, a warning symbol, and a clock, labeled “Repeated exposure to crime media.”
(c) The bottom row maps each level to research methods: Computation shows a person with checklist icons and the text “EMA/ESM surveys, in-app surveys, pop-up questionnaires.” Algorithm shows a smartphone social media interface with the text “Digital trace data, smartphone usage logs, screen time APIs.” Implementation shows a person interacting with a smartwatch and the text “Wearable biometric sensors (e.g., smartwatches, Oura ring), consumer-grade EEG (e.g., MUSE headbands), eyetracking devices (e.g., Gazepoint GP3).”

Richard Huskey@richardhuskey· Feb 14

Today @rachaelkee.bsky.social successfully defended her QE! She submitted two papers: (1) a positioning statement introducing what she is calling “high-throughput communication science” and (2) a proposal for the first-ever demonstration of this research agenda. Way to go, Rachael! 🎉🎉🎉

Richard Huskey@richardhuskey· Feb 11

7/11 As symptoms shift, preferences shift. ↑ Depression and anxiety: preferences shifts from high- to low-arousal movies. ↑ Loneliness: stronger preference for negative movies. These effects emerge specifically for drift rate, but *not* caution (a) or bias (Z).

PHQ-9, depression: A negative interaction between depression and arousal on drift rate. As depression increases, preference for high-arousal movies decreases and reverses toward low-arousal content at moderate symptom levels. Other parameters (boundary, bias, non-decision time) show no credible effects.

GAD-7, anxiety: A similar negative interaction between anxiety and arousal on drift rate. Higher anxiety predicts reduced preference for high-arousal movies, with reversal at higher symptom levels. A small effect appears on decision bias for valence, but drift rate shows the primary effect.

UCLA-L, loneliness: A negative interaction between loneliness and valence on drift rate. As loneliness increases, preference for negatively valenced movies strengthens. No credible effects on other decision parameters.

Richard Huskey@richardhuskey· Feb 11

6/11 First, we replicate our prior work (doi.org/10.1093/joc/...). People show preferential evidence accumulation (v) for negatively valenced and high-arousal movies. Differences in valence and arousal also reduce decision boundary. People are less cautious when options differ affectively.

Four posterior probability distribution plots from a hierarchical Bayesian drift diffusion model.

(A) Drift rate: negative valence shows a negative effect (preference for negative movies), and high arousal shows a positive effect (preference for high-arousal movies). Both effects are credibly different from zero.

(B) Decision boundary: both valence and arousal differences reduce boundary separation, indicating less cautious decisions.

(C) Decision bias: effects centered near zero, showing no credible bias shifts.

(D) Non-decision time: small reductions when valence or arousal differ.

Vertical dashed lines mark zero for reference.

Richard Huskey@richardhuskey· Feb 11

5/11 We analyzed choices with the drift diffusion model (DDM). It decomposes decisions into: • Drift rate (v): preferential evidence accumulation • Boundary (a): caution • Bias (Z): starting preference • Non-decision time (T): perceptual/motor time We expected mental health moderates drift rate.

Conceptual diagram of the drift diffusion model. Two horizontal lines represent upper and lower decision boundaries. A starting point (z) lies between them. Jagged lines show noisy evidence accumulating over time toward one boundary or the other, with slope representing drift rate (v). Reaction time distributions are shown at the top and bottom corresponding to choices at the upper and lower boundaries. Parameters labeled include drift rate (v), boundary (a), bias (z), and non-decision time (t).

Richard Huskey@richardhuskey· Feb 11

4/11 Does mental health shape media choice? To test this, participants completed a two-choice decision task (196 trials), selecting movie summaries that systematically varied in both valence and arousal.

Illustration of the experimental task.

A participant sits at a laptop while pairs of movie summaries appear sequentially on screen. Each trial presents two summaries side-by-side. Participants press one of two keys (Z or M) to select their preferred option. A dashed arrow labeled “Time” indicates the repeated sequence of trials across the task. In total, participants completed 196 two-choice decisions.

Richard Huskey@richardhuskey· Feb 11

3/11 Our sample (n = 313) shows similar patterns. Students report substantial symptoms of depression (PHQ-9), anxiety (GAD-7), and loneliness (UCLA-L).

Four histograms labeled A–D.

(A) Age distribution centered around 20 years, with most participants between 18–22.

(B) PHQ-9 scores spread broadly from 0–25, with many students in the mild-to-moderate range.

(C) GAD-7 scores similarly distributed from 0–20, with substantial representation in moderate ranges.

(D) Loneliness (UCLA-L) scores spanning roughly 20–70, with many students reporting elevated loneliness.

Overall, the sample shows meaningful variation and substantial levels of depression, anxiety, and loneliness.

Richard Huskey@richardhuskey· Dec 3

Tonight’s commute home was epically not bad

Richard Huskey@richardhuskey· Nov 21

This is going to ruin the tour www.pnas.org/doi/10.1073/...

Screenshot of a PNAS research article titled “The potential existential threat of large language models to online survey research.” The article is authored by Sean J. Westwood and edited by James N. Druckman. It notes the submission date (July 9, 2025), acceptance date (September 12, 2025), and publication date (November 20, 2025).

Richard Huskey@richardhuskey· Nov 2

The chapter contains details about day-to-day operations. But my favorite section is the last: "Future Directions and Advice". We point to key articles that guide the lab's organization, and how we think about a life in science. In short: Deep thinking is rarely efficient. But that’s the point.

Screen shot of part of a chapter, with the section heading "Future Directions and Advice. The chapter text reads:

There exists no shortage of excellent advice on how to run a lab, be a good mentor, or succeed as a trainee. Below we share some resources that we’ve found particularly helpful over the years. At the Helm: Leading Your Laboratory (Barker, 2010) provides a comprehensive overview of managing the day-to-day operations of a lab. When new trainees are considering joining the lab, they are often assigned to read our lab manual and expectations documents. They are also encouraged to read Barres’s (2013) reflection on how to pick a graduate advisor. Incidentally, Barres’s article offers great advice on how to be a graduate advisor (Richard re-reads this article every year). In the same way, Raman (2014) offers excellent advice on how to be an advisee. It is also worth noting that what constitutes a good scientific problem changes over time. What is good for a first year graduate student is quite different from that of a postdoc. The same is true for a fledgling lab compared to a well-established one. Thankfully, Alon (2009) offers expert advice on how to pick a good scientific problem. There is always pressure to find efficiencies, but thinking deeply is rarely efficient and lab meetings are imperfect for facilitating deep discussions. Deep thinking that leads to scientific progress flourishes in dyads (Yanai & Lercher, 2024). Therefore, our lab will continue to prioritize (inefficient but essential) one-on-one meetings between PI and trainee, even if that places an upper limit on our lab size. Finally, and drawing on Schwartz (2008), we recognize that science thrives on embracing our “absolute stupidity”—the kind of ignorance that stems from tackling the unknown and asking questions without immediate answers. We will continue to embrace absolute stupidity and wish the same for you, too.

Richard Huskey@richardhuskey· Nov 2

🧵 What does it take to build a small, scrappy, and successful communication neuroscience lab? Our lab, @gongxuanjun.bsky.social, @rachaelkee.bsky.social, Allyson Snyder, Ziyu Zhao, and I put out heads together to answer this question. Here's what we came up with: link.springer.com/chapter/10.1...

Cover for the book "Creating Communication and Media Research Labs: A Blueprint for Success". Edited by Chad Edwards, Autumn Edwards, and Patric R. Spence. Published by Palgrave Pivot.

Richard Huskey@richardhuskey· Nov 2

Richard Huskey@richardhuskey· Oct 31

Twenty years of edits later, my lasagna barely resembles my mother-in-law’s original. Philosophically speaking, am I still making _her_ lasagna?

Stylized illustration of the Ship of Theseus in the style of ancient Greek pottery. A black longship with white-and-black striped sails glides across a reddish-brown sea. Rows of oars extend from the hull, and several sleek black fish leap alongside the ship.

Richard Huskey@richardhuskey· Oct 24

What’s it like inside your head?

My dog named Turner peers up from inside his metal crate. The photo is slightly blurry and tightly cropped, making his face appear pressed against the bars and his expression look sad and longing. The angle and focus give the image an anxious and urgent feeling.

Richard Huskey@richardhuskey· Sep 26

Perfect email? Or perfectest email?

Screenshot of an email. The email begins with:

“25-Sep-2025

Dear Dr. Richard Huskey,”

Then, after the date and salutation, there are several line breaks where the body should be, and then an email signature. The email has no body. It does not ask for anything to be read, or done.

Richard Huskey@richardhuskey· Sep 23

Ask me how I know the school year started today.

Richard Huskey@richardhuskey· Sep 15

📱One scroll can change your day. What guides your scrolling? I’m keynoting at the CityU Hong Kong International Conference on AI-Empowered Communication and Global Innovation. I’ll share our lab’s NeuroAI for sequential media selection. If you’re there, say hi! www.cityu.edu.hk/com/Page.asp...

flyer for “AIxCOM – International Conference on AI-Empowered Communication and Global Innovation.” Host: Department of Media and Communication, City University of Hong Kong. Dates: 18–19 September 2025. Venue: Benjamin & Anny Kwok Conference Room, 4/F Yip Kit Chuen Building, CityUHK.

Highlighted bullets in yellow: Cutting-edge AI issues across GenAI, VR, Games, and Neuroscience. Flagship communication journal editors. Elected and former ICA presidents. Industry leaders from Amazon, Meta, Google, and Alibaba.

Three columns of circular headshots with names, titles, and affiliations labeled “Keynote Speakers”:
• Ingrid Bachmann – ICA President Elect, Pontificia Universidad Católica de Chile.
• Peng Hwa Ang – Former ICA President, Nanyang Technological University.
• Michael Latzer – University of Zurich.
• Junhao Hong – State University of New York at Buffalo.
• Natalie Pang – National University of Singapore.
• Nicole Ellison – Editor in Chief, Journal of Computer-Mediated Communication.
• Jörg Matthes – Editor in Chief, Communication Theory, University of Vienna.
• Richard Huskey – Associate Editor, Journal of Communication, UC Davis.
• Nuri Kim – Associate Editor, Human Communication Research.
• Tammy Lin – Associate Editor, Journal of Computer-Mediated Communication, NCCU Taiwan.
Right column lists editors and scholars from China: Xiaohong Guan (CAS academician, IEEE Fellow), Zhengrong Hu (News and Communication Research), Yan Sui (Modern Communication), Taofu Zhang (Journalism University), Hailong Liu (International Press), Changfeng Chen (Global Media Journal, Chinese).

Row of “Industry Leaders” headshots: George Chen (The Asia Group, CityUHK), Bryan Ha (AWS/Amazon), Yannick Mahé (Meta Greater China), Raymond Tsang (Alibaba Cloud), Di Yao (Google Shanghai), Grace Shao (AI Proem).

Richard Huskey@richardhuskey· Sep 11

Very excited to welcome our newest lab member, Jocelyn (Yachen) Xie! Jocelyn is broadly interested in cognition and media narratives, especially in how moral judgment and decision-making affect people’s preferences and evaluations of media content and characters cogcommscience.com/members/