Large language models can infer psychological dispositions of social media users.

Large language models (LLMs) demonstrate increasingly human-like abilities across a wide variety of tasks. In this paper, we investigate whether LLMs like ChatGPT can accurately infer the psychological dispositions of social media users and whether their ability to do so varies across socio-demographic groups. Specifically, we test whether GPT-3.5 and GPT-4 can derive the Big Five personality traits from users' Facebook status updates in a zero-shot learning scenario. Our results show an average correlation of r = 0.29 ( range = [ 0.22 , 0.33 ] ) between LLM-inferred and self-reported trait scores-a level of accuracy that is similar to that of supervised machine learning models specifically trained to infer personality. Our findings also highlight heterogeneity in the accuracy of personality inferences across different age groups and gender categories: predictions were found to be more accurate for women and younger individuals on several traits, suggesting a potential bias stemming from the underlying training data or differences in online self-expression. The ability of LLMs to infer psychological dispositions from user-generated text has the potential to democratize access to cheap and scalable psychometric assessments for both researchers and practitioners. On the one hand, this democratization might facilitate large-scale research of high ecological validity and spark innovation in personalized services. On the other hand, it also raises ethical concerns regarding user privacy and self-determination, highlighting the need for stringent ethical frameworks and regulation.

Using machine learning to predict student retention from socio-demographic characteristics and app-based engagement metrics.

Student attrition poses a major challenge to academic institutions, funding bodies and students. With the rise of Big Data and predictive analytics, a growing body of work in higher education research has demonstrated the feasibility of predicting student dropout from readily available macro-level (e.g., socio-demographics or early performance metrics) and micro-level data (e.g., logins to learning management systems). Yet, the existing work has largely overlooked a critical meso-level element of student success known to drive retention: students' experience at university and their social embeddedness within their cohort. In partnership with a mobile application that facilitates communication between students and universities, we collected both (1) institutional macro-level data and (2) behavioral micro and meso-level engagement data (e.g., the quantity and quality of interactions with university services and events as well as with other students) to predict dropout after the first semester. Analyzing the records of 50,095 students from four US universities and community colleges, we demonstrate that the combined macro and meso-level data can predict dropout with high levels of predictive performance (average AUC across linear and non-linear models = 78%; max AUC = 88%). Behavioral engagement variables representing students' experience at university (e.g., network centrality, app engagement, event ratings) were found to add incremental predictive power beyond institutional variables (e.g., GPA or ethnicity). Finally, we highlight the generalizability of our results by showing that models trained on one university can predict retention at another university with reasonably high levels of predictive performance.

Regional personality differences predict variation in early COVID-19 infections and mobility patterns indicative of social distancing.

The early stages of the COVID-19 pandemic revealed stark regional variation in the spread of the virus. While previous research has highlighted the impact of regional differences in sociodemographic and economic factors, we argue that regional differences in social and compliance behaviors-the very behaviors through which the virus is transmitted-are critical drivers of the spread of COVID-19, particularly in the early stages of the pandemic. Combining self-reported personality data that capture individual differences in these behaviors (3.5 million people) with COVID-19 prevalence and mortality rates as well as behavioral mobility observations (29 million people) in the United States and Germany, we show that regional personality differences can help explain the early transmission of COVID-19; this is true even after controlling for a wide array of important sociodemographic, economic, and pandemic-related factors. We use specification curve analyses to test the effects of regional personality in a robust and unbiased way. The results indicate that in the early stages of COVID-19, Openness to experience acted as a risk factor, while Neuroticism acted as a protective factor. The findings also highlight the complexity of the pandemic by showing that the effects of regional personality can differ (a) across countries (Extraversion), (b) over time (Openness), and (c) from those previously observed at the individual level (Agreeableness and Conscientiousness). Taken together, our findings support the importance of regional personality differences in the early spread of COVID-19, but they also caution against oversimplified answers to phenomena as complex as a global pandemic. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Depression predictions from GPS-based mobility do not generalize well to large demographically heterogeneous samples.

Depression is one of the most common mental health issues in the United States, affecting the lives of millions of people suffering from it as well as those close to them. Recent advances in research on mobile sensing technologies and machine learning have suggested that a person's depression can be passively measured by observing patterns in people's mobility behaviors. However, the majority of work in this area has relied on highly homogeneous samples, most frequently college students. In this study, we analyse over 57 million GPS data points to show that the same procedure that leads to high prediction accuracy in a homogeneous student sample (N = 57; AUC = 0.82), leads to accuracies only slightly higher than chance in a U.S.-wide sample that is heterogeneous in its socio-demographic composition as well as mobility patterns (N = 5,262; AUC = 0.57). This pattern holds across three different modelling approaches which consider both linear and non-linear relationships. Further analyses suggest that the prediction accuracy is low across different socio-demographic groups, and that training the models on more homogeneous subsamples does not substantially improve prediction accuracy. Overall, the findings highlight the challenge of applying mobility-based predictions of depression at scale.