Vincent Traag

@vtraag.bsky.social

675 Followers 412 Following

Senior researcher at CWTS, Leiden University. Computational social science, science studies, modelling & networks.

Overzicht van schattingen van percentage stemmen voor partijen. De schattingen zijn gemaakt zonder rekening te houden met zwevende kiezers en vergeleken met schattingen als we dat wel doen. De onzekerheid is zichtbaar groter indien er wel rekening wordt gehouden met zwevende kiezers.
Illustration of the potential collider in the early-career setback effect based on a simulation. We show quality and MCFR values for near-hit (in blue) and near-miss (in red) applicants from a simulation. Simulated applicants who reapply later are opaque, while those who do not are translucent. The marginal distributions of quality and MFCR are similar for funded and unfunded applicants (shown as translucent filled areas in the marginal plots). However, in our simulation, unfunded applicants only reapply when they have a high MFCR, while funded applicants reapply irrespective of their MFCR. The marginal distribution of the MFCR clearly shows that the MFCR for unfunded applicants is higher than for funded applicants who reapply (shown as a solid line in the marginal plots). Due to some association between quality and citations in our simulation, there is also a smaller but visible difference in the marginal distribution for quality between funded and unfunded applicants who reapply. This illustrates that results similar to an early-career setback effect may appear due to a collider bias, and need not represent a causal effect. Moreover, it also shows that the "conservative removal" procedure does not properly address this collider bias, because there are simply not as many funded applicants as unfunded applicants with equally high MFCRs.
The early-career setback effect replicates but does not seem to generalise. In (a) we show that funded applicants have a higher MFCR in all periods. In (b) we show that among near-hit/miss applicants, the "Between" MFCR is higher for unfunded applicants while the MFCR is nearly the same across all other periods. In (c) we show that the Bayesian model suggests a positive effect of funding on later MFCRs. We illustrate this effect using the mean inferred quality as a baseline, and we display the distribution of the MFCR of 30 publications (i.e. using the mean over 30 publications). The inset shows the distribution of the estimated effect coefficient.
The latent quality affects whether a researcher applies for funding, the review score (s)he receives and his or her citation impact. The funding decision at time t affects the decision to apply, the review score and the citation impact at time t + 1. The effect of time is modelled using a survival analysis approach and also considers right censoring of the observations. Field and affiliation effects in the funding decision are not illustrated here for simplification, but are included in the model.
The Matthew effect replicates and generalises. In (a) we show that funded applicants are more likely to receive later-career funding than unfunded applicants, both overall and limited to near-hit/miss applications. In (b) we show that previously funded applicants are more likely to submit later applications than unfunded applicants. In contrast, (c) shows that prior funding does not increase review scores, if anything, the effect is even slightly negative. The illustrated effects are calculated using the mean inferred quality as a baseline. The insets show the posterior distributions of the estimated effect coefficients.
Illustration of the Matthew effect and the early-career setback effect. An academic, say Alice, first applies for early-career funding in 2015. She received early-career funding, and goes on to reapply for later-career funding in 2020. According to the Matthew effect, the chances of Alice receiving later-career funding is higher when she received early-career funding. Alice showed a high Mean Field Citation Rate (MFCR) before receiving her early-career funding, and similarly a high MFCR in between the early-career and later-career applications. According to the early-career setback effect, had she instead not received funding, her "Between" MFCR would have been higher. We not only study the "Between" MFCR, we also study the MFCR after the early-career application ("Post (early)") and after the later-career application ("Post (later)").
Gallery of various network visualisations.
This displays the density of sentiment of posts containing scientific and untrustworthy content from scientist users. The sentiment is based on VADER, which ranges from −1 (most extreme negative) to 1 (most extreme positive), with 0 as the neutral point. This metric allows us to assign a single unidimensional measure of emotions expressed in the posts based on the emojis and the text.
Plot of the level of scientific sources shared by individual users as a function of their number of followers. For each user, we calculate the average proportion of scientific content shared. We consider only posts that contain either scientific sources or untrustworthy sources. We group users in 35 bins based on their follower count, and we compute the average level of scientific sharing by aggregating across the users within each bin.
Statistics of the publications. By selecting a subset of publications shared on Twitter by at least 100 users, we identified a total of 7, 905 publications and divided them into two equally sized groups that have reliably, respectively, unreliably used publications.
Network of publications co-shared by reliable and unreliable users. To examine how scientific publications are discussed on Twitter in more detail, we construct the network of publications co-tweeted by the same user. Specifically, each node represents a scientific article, and edges connect articles that have been tweeted by the same user. Each node represents a scientific publication, and an edge is established between two nodes if both articles were tweeted by at least one user. The resulting network consists of N = 2,084 nodes and E = 53,757 edges. For clarity of the visualisation, only 5,000 edges are displayed. Node size represents the number of tweets mentioning the corresponding article. Colour values are z-score normalised, and for each node, they represent the level of reliable activity of the users sharing the article, ranging from more unreliable users (more violet and closer to −1) to more reliable users (more yellow and closer to 1). This shows a cluster of publications predominantly shared by users who also share unreliable sources.
Pairwise overlap percentages between reliability categories. Each row shows the percentage of users who shared content (i.e. at least one tweet) in that row who also share content (i.e. at least one tweet) in the category in the column. Note that percentages do not add up to 100% because users can share multiple types of content.
Mapping the interaction between science and misinformation in COVID-19 tweets. Publication from @luzuzek.bsky.social, Juan Pablo Bascur, @annabertani.bsky.social, @ricgallotti.bsky.social. This project is supported by European Media and Information Fund. Abstract: During the COVID-19 pandemic, scientific understanding related to the topic evolved rapidly. Along with scientific information being discussed widely, a large circulation of false information, labelled an infodemic by the WHO, emerged. Here, we study the interaction between misinformation and science on Twitter (now X) during the COVID-19 pandemic. We built a comprehensive database of 407M COVID-19 related tweets and classified the reliability of URLs in the tweets based on Media Bias/Fact Check. In addition, we use Altmetric data to see whether a tweet refers to a scientific publication. We find that many users find that many users share both scientific and unreliable content; out of the 1.2M users who share science, also share unreliable content. Publications that are more frequently shared by users who also share unreliable content are more likely to be preprints, slightly more often retracted, have fewer citations, and are published in lower-impact journals on average. Our findings suggest that misinformation is not related to a ``deficit'' of science. In addition, our findings raise some critical questions about certain open science practices and their potential for misuse. Given the fundamental opposition between science and misinformation, our findings highlight the necessity for proactive scientific engagement on social media platforms to counter false narratives during global crises.
Lucila stands in front of slides with two network visualizations of papers highlighting which papers are more often shared by scientists (left) and reliable users (right), with more reliable users/scientists displayed in yellow and less reliable/non-scientists in purple.