Working in groups can help Republicans and Democrats agree on controversial content moderation online

A new study by Professor Damon Centola and alum Douglas Guilbeault explored how content moderators can reach consensus on classifying controversial material online, including inflammatory, offensive, or hateful images.

University of Pennsylvania

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

Over half of Americans believe tech companies should take action to restrict extremely violent content on their platforms, according to Pew data, yet even trained content moderators consistently disagree in their decisions for how to classify hate speech and offensive images. A new study by Annenberg School for Communication Professor Damon Centola and Stanford University Assistant Professor Douglas Guilbeault (Ph.D. ‘20) has identified a key mechanism to aid content moderators, even those across the political aisle, in reaching consensus on classifying controversial material online: working in teams.

In an experiment involving over 600 participants with diverse political views, Centola and Guilbeault found that content moderators who classified controversial social media content in groups reached near-perfect agreement on what should remain online. Those who worked alone showed only 38% agreement by the end of the experiment. 

“Morally controversial content, such as offensive and hateful images on social media, is especially challenging to categorize, given widespread disagreement in how people interpret and evaluate this content,” Centola says. “Yet, recent large-scale analyses of classification patterns over social media suggest that separate populations, such as Democrats and Republicans, can reach surprising levels of agreement in the categorization of inflammatory content like fake news and hate speech, despite considerable differences in their moral reasoning and worldview. We wanted to know why.”

Centola and Guilbeault had a hunch that a phenomenon called “structural synchronization” might be behind this agreement across partisan divides. “Structural synchronization is a process in which interacting in social networks can filter individual variation and lead separate networks to arrive at highly similar classifications of controversial content,” says Centola.

To test this hypothesis, Centola and Guilbeault designed a large-scale experiment in which 620 participants evaluated a curated set of controversial social media images. The images ranged from depictions of interpersonal conflicts, such as bullying or domestic violence, to images involving militaristic violence, including armed conflict and terrorism. The participant sample reflected a politically diverse group, with nearly half identifying as Democrats (49.6%), followed by Independents (28.3%), Republicans (20.7%), and a small fraction identifying with another party. All participants had prior experience as content moderators.

Participants were told that the images came from Facebook posts and that Facebook had requested their help in determining whether each image should be removed. Participants were randomly assigned to one of two conditions: the individual condition, in which they classified images alone, and the network condition, in which participants classified images while interacting in structured networks of 50 people. 

In each round of the experiment, participants in the network condition were paired with a partner and assigned one of two roles: “speaker” or “hearer.” The speaker was shown a set of three randomly selected images, with one of the images highlighted. The speaker was given thirty seconds to assign a violation tag to the image or select the "Do Not Remove" option. The hearer then attempted to identify the image based on the tag. Correct matches resulted in cash rewards, and mismatches resulted in both participants losing money. Once all pairs completed a round, a new round would begin with everyone in the network being randomly paired again. Participants had no information about the decisions of others outside their immediate partner.

In analyzing the experiment's results, Centola and Guilbeault found substantial differences between the two conditions. In the individual condition, participants showed only 38% agreement about which images should remain on Facebook. Their judgments varied widely, and partisan differences were pronounced: Democrats and Republicans agreed on classifications for only about 30% of the images. In contrast, the network condition classified controversial social media content with near-perfect agreement across each of the eight networks. Even more striking was the reduction in political disagreement. Networked teams reduced partisan gaps by 23 percentage points. By the end of the task, Democrats and Republicans within the same networks agreed on most content classifications.

The researchers also explored participants’ emotional responses during the experiment. Those in the network condition reported significantly more positive feelings about the task and lower emotional stress compared with individuals assessing content alone. 

“Our findings suggest that collaboration may not only encourage collective decision-making but also reduce the psychological strain associated with evaluating violent or disturbing material,” Centola says. “At a time when tech companies and policymakers are grappling with how to manage harmful content online, the study provides compelling evidence that structured social interaction can strengthen both accuracy and agreement in moderation decisions. This is a hopeful sign for creating systems that support more consistent, transparent, and psychologically healthy content moderation.”

“Structural Synchronization in the Classification of Controversial Content” was authored by Douglas Guilbeault and Damon Centola and published in Journal of Social Computing.

Funding
This research was funded by Facebook’s Content Moderation Research Award. Facebook played no role in the design of this study, nor in the collection and analysis of the data. The sample of participants in the study was drawn from individuals who work in content moderation across a range of public social media platforms.

---

Journal

Journal of Social Computing

DOI

10.23919/JSC.2025.0024 

Method of Research

Experimental study

Subject of Research

People

Article Title

Structural Synchronization in Classification of Controversial Content

Article Publication Date

17-Dec-2025

 

Study uncovers disrupted brain balance in alcohol dependence

Experiments show how dependence on alcohol rewires the brain's "go" and "stop drinking" signaling systems, with implications for preventing relapse.

Scripps Research Institute

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Orexin, one of the two signals the researchers studied, originates in neurons (green) in the brain’s hypothalamus (left) that connect with a part of the thalamus called the pPVT (right). 

view more 

Credit: Scripps Research

LA JOLLA, CA—A new study by Scripps Research reveals that alcohol dependence disrupts two signaling pathways in a stress-related part of the brain—and offers insights on developing drugs to treat this condition.

The research, conducted in animal models and published in Frontiers in Pharmacology on November 26, 2025, helps explain why people with alcohol use disorder (AUD) struggle to stay sober, especially under stress.

"We think that alcohol dependence changes these systems, and that's why individuals are prone to seek out alcohol even if they've gone without it for some time,” says senior author Rémi Martin-Fardon, an associate professor in the Department of Translational Medicine.

AUD, a condition in which someone cannot control their drinking despite the harm it causes, affects nearly 28 million Americans. FDA-approved medications like naltrexone, which reduces cravings, exist but have significant drawbacks, including nausea, and they do not work for everyone.

This recent study points to potential new treatments by uncovering molecular changes in two signaling systems—orexin and dynorphin—and the unexpected effects of blocking them. Inhibiting either signal individually reduced relapse-like behavior but, interestingly, blocking both cancelled this protective effect.

Once someone loses control of their drinking, regaining it can be a lifelong struggle. Research has demonstrated that stress increases the likelihood someone will become dependent on alcohol and will relapse after they try to stop. But the relationship goes two ways, because drinking itself activates the body’s stress response systems. 

Among the systems affected is orexin-dynorphin signaling. These two neuropeptides are released by the same neurons, which originate in the brain’s hypothalamus, the region that coordinates the release of chemical signals. In normal brains, they have opposing effects: Orexin, discovered by scientists at Scripps Research and another team in 1998, acts as a "go" signal and promotes drug-seeking behavior. Meanwhile, dynorphin serves as the "stop" signal. In AUD, excessive drinking appears to alter dynorphin signaling so it produces the many diverse and unpleasant feelings that accompany withdrawal and motivate continued drinking. 

Martin-Fardon's group has previously studied this relationship in cocaine addiction. This time, in experiments led by postdoctoral researcher Francisco Flores-Ramirez with research assistant Glenn Pascasio, they turned to alcohol.

The team zeroed in on a small region within the brain’s thalamus called the posterior paraventricular nucleus of the thalamus, or pPVT, a stress-processing hub that receives orexin and dynorphin signals. Martin-Fardon’s earlier work indicated that the pPVT is key to stress-triggered relapse-like behavior.

In this study, they found telling changes in gene expression. In alcohol-dependent rats, the hypothalamus appeared to be ramping up production of both the orexin "go" and dynorphin "stop" signals. But the pPVT's ability to receive them was skewed. Its neurons expressed fewer receptors for orexin, but more for dynorphin.

"What that tells us is that just being dependent on alcohol changes the orexin and dynorphin system, and that these changes persist well into abstinence," Flores-Ramirez says.

To reach these conclusions, the team simulated AUD in male rats, which pressed a lever to receive alcohol. After cutting off the alcohol supply, the researchers examined gene expression in both the pPVT and hypothalamus. Some animals had received inhibitors to block orexin or dynorphin signaling in their pPVTs, producing complex and somewhat counterintuitive results. As expected, shutting down the orexin "go" signal reduced stressed rats' attempts to drink alcohol. But blocking dynorphin's "stop" signal also appeared to significantly decrease the relapse-like behavior. When they inhibited both signals together, however, the rats pressed the lever as if they had not received any inhibitor at all.

Because the study focused on a single brain region and relied solely on male animals, the researchers say it's difficult to explain these effects or tie them directly to the changes in gene expression. Still, they sound a note of caution for drug development.

"If you want to combine treatments, you have to be very careful," Martin-Fardon says, though he notes that this strategy may be effective with different inhibitors—perhaps taking advantage of existing drugs or compounds with similar chemistry.

His team is now collaborating with Scripps Research colleagues Edward Roberts and Hugh Rosen, who are developing selective, shorter-acting dynorphin signaling inhibitors intended to provide quick relief. Martin-Fardon is also interested in combining one of these compounds with suvorexant, an insomnia drug that blocks orexin signaling, or a similar drug.   

Authors of the study, “Functional interaction between orexin/dynorphin transmission in the posterior paraventricular nucleus of the thalamus following alcohol dependence: mediation of alcohol-seeking behavior,” include Francisco J. Flores-Ramirez, Glenn Pascasio, and Rémi Martin-Fardon.

This study was supported by funding from the National Institute on Alcohol Abuse and Alcoholism (grant no. AA028549 and AA006420 to RM-F).

---

Journal

Frontiers in Pharmacology

Article Title

Functional interaction between orexin/dynorphin transmission in the posterior paraventricular nucleus of the thalamus following alcohol dependence: mediation of alcohol-seeking behavior

Brain injuries linked with potential risk of suicide, new study finds

The first of its kind study analysed data from over 1.8 million

 adults in the UK across a 20-year period.

University of Birmingham

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

Adults who experience a head injury face a substantially higher risk of attempting suicide compared to those without such injuries, according to the findings from a new UK-based study.

Published in Neurology®, the medical journal of the American Academy of Neurology, the study was led by University of Birmingham researchers. The paper is the first of its kind to examine suicide risk across all types of head injuries in a general population, moving beyond the traditional focus on traumatic brain injuries (TBIs) in military, athletic or hospital settings.

The population-based matched cohort study used nationally represented electronic primary healthcare records from more than 1.8 million adults, linked with Hospital Episode Statistics and Office for National Statistics data.

Researchers found that people with head injuries were 21% more likely to attempt suicide than those without, after analysing data across a 20-year period.
 

Key Findings:

• Researchers found that people with head injuries were 21% more likely to attempt suicide than those without, even after adjusting for age, sex, deprivation, and mental health history.
• The incidence rate was 2.4 per 1,000 person-years in individuals with head injuries, compared to 1.6 per 1,000 person-years in the control group. This translates to an absolute increase in risk of 0.7% (1.3% vs 0.6%), yet the adjusted hazard ratio reported was 21%, and
• Elevated risk was observed across all subgroups, including individuals with no prior mental health conditions, highlighting that head injuries alone are linked to increased psychological vulnerability.

 

Professor Nicola Adderley, Professor of Epidemiology and Real-World Evidence at the University of Birmingham and a lead author of the study, said: “Our findings show that the impact of head injuries are not limited to just physical symptoms or repercussions. They can have profound psychological consequences. Suicide risk assessments should be considered for anyone with a recent head injury, regardless of their mental health history, to improve and safeguard patient outcomes.”

In the UK alone, nearly 6,000 deaths each year are attributed to suicide whilst the number of attempts is significantly higher. The study’s findings showed that the risk of suicide attempt was highest in the first 12 months following a head injury, suggesting a critical window for intervention.

While the risk declined over time, it remained elevated compared to those without head injuries. Researchers also found that social deprivation and a history of mental health conditions further amplified the risk.

While suicide attempts were more common among those with head injuries, the study did not find a significant increase in deaths by suicide after accounting for competing risks such as other causes of death; suggesting that head injuries may lead to more frequent non-fatal attempts.

Researchers are calling for the following changes in healthcare settings:

• Routine suicide risk screening in primary and secondary care settings for patients with head injuries.
   
• Enhanced mental health support, particularly during the first 12 months post-injury; with public awareness campaigns to help families and caregivers recognise warning signs.
   
• The development and testing of suicide risk assessment and prevention strategies for people with head injuries should be investigated, especially within the first 12 months post-head injury and irrespective of mental health history.

 

Professor G. Neil Thomas, Professor of Epidemiology and Research Methods and a lead author of the study, said: “These findings have implications for both clinical practice and health policy; highlighting the urgent need for targeted mental health and wellbeing support.

“The development and testing of robust suicide risk assessment and prevention strategies for people with head injuries should be further investigated; especially within the first 12 months post-head injury and irrespective of mental health history.”

The research utilised data from the Clinical Practice Research Datalink (CPRD), linked with hospital and mortality records, covering a 20-year period (2000–2020). Ethical approval was granted by the Health Research Authority and CPRD Independent Scientific Advisory Committee.

---

Journal

Neurology

Article Title

The risk of suicide attempts after head injury: a matched UK population-based cohort study

Article Publication Date

22-Dec-2025

Exploring why some people may tend to persistently make bad choices

Some people rely on surrounding images and sounds to make decisions more than others. These individuals have a harder time updating their beliefs about these cues when they change to signal risky outcomes, which may lead to poor decision making over time.

Society for Neuroscience

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

When people learn that surrounding visuals and sounds may signify specific choice outcomes, these cues can become guides for decision making. For people with compulsive disorders, addictions, or anxiety, the associations between cues and choice outcomes can eventually promote poor decisions as they come to favor or avoid cues in a more biased manner. Giuseppe di Pellegrino, from the University of Bologna, led a study to explore associative learning and maladaptive decision making in people. 

As reported in their JNeurosci paper, the researchers discovered that some people rely on surrounding cues to make decisions more than others. Furthermore, these individuals may have a harder time updating their beliefs and unlearning these associations when the cues change to signify riskier outcomes. This leads to more disadvantageous decision making that persists over time. 

According to the researchers, this work suggests that some people have stronger cue sensitivity and less of an ability to update their beliefs about cues than others. The researchers aim to continue exploring associative learning in patient populations and probing whether harmful decision patterns—which characterize addictions, compulsive disorders, and anxiety—are more likely in those with heightened sensitivity to visuals and sounds that guide their choices. 

### 

Please contact media@sfn.org for full-text PDF. 

About JNeurosci 

JNeurosci was launched in 1981 as a means to communicate the findings of the highest quality neuroscience research to the growing field. Today, the journal remains committed to publishing cutting-edge neuroscience that will have an immediate and lasting scientific impact, while responding to authors' changing publishing needs, representing breadth of the field and diversity in authorship. 

About The Society for Neuroscience 

The Society for Neuroscience is the world's largest organization of scientists and physicians devoted to understanding the brain and nervous system. The nonprofit organization, founded in 1969, now has nearly 35,000 members in more than 95 countries. 

---

Journal

JNeurosci

DOI

10.1523/JNEUROSCI.1465-25.2025 

Subject of Research

People

Article Title

Reduced Pavlovian Value Updating Alters Decision-Making in Sign-Trackers

Article Publication Date

22-Dec-2025