To optimize its performance, a competitive team, such as a soccer team, must maintain a delicate balance between organization and disorganization. On the one hand, the team should maintain organized patterns of behavior to maximize the cooperation between its members. On the other hand, the team’s behavior should be disordered enough to mislead its opponent and to maintain enough degrees of freedom. In this paper, we have analyzed this dynamic in the context of soccer games and examined whether it is correlated with the team’s performance. We measured the organization associated with the behavior of a soccer team through the Tsallis entropy of ball passes between the players. Analyzing data taken from the English Premier League (2015/2016), we show that the team’s position at the end of the season is correlated with the team’s entropy as measured with a super-additive entropy index. Moreover, the entropy score of a team significantly contributes to the prediction of the team’s position at the end of the season beyond the prediction gained by the team’s position at the end of the previous season.

Depression accounts for a major share of global disability‐adjusted life‐years (DALYs). Diagnosis typically requires a psychiatrist or lengthy self‐assessments, which can be challenging for symptomatic individuals. Developing reliable, noninvasive, and accessible detection methods is a healthcare priority. Voice analysis offers a promising approach for early depression detection, potentially improving treatment access and reducing costs. This paper presents a novel pipeline for depression detection that addresses several critical challenges in the field, including data imbalance, label quality, and model generalizability. Our study utilizes a high‐quality, high‐depression‐prevalence dataset collected from a specialized chronic pain clinic, enabling robust depression detection even with a limited sample size. We obtained a lift in the accuracy of up to 15% over the 50–50 baseline in our 52‐patient dataset using a 3‐fold cross‐validation test (which means the train set is n = 34, std 2.8%, p ‐value 0.01). We further show that combining voice‐only acoustic features with a single self‐report question (subject unit of distress [SUDs]) significantly improves predictive accuracy. While relying on SUDs is not always good practice, our data collection setting lacked incentives to misrepresent depression status; SUDs were highly reliable, giving 86% accuracy; adding acoustic features raises it to 92%, exceeding the stand‐alone potential of SUDs with a p ‐value 0.1. Further data collection will enhance accuracy, supporting a rapid, noninvasive depression detection method that overcomes clinical barriers. These findings offer a promising tool for early depression detection across clinical settings.

This paper studies students' engagement in e-learning environments in which students work independently and solve problems without external supervision. We propose a new method to infer engagement patterns of users in such self-directed environments. We view engagement as a continuous process in time, measured along carefully chosen axes that are derived from student data in the system. We use a non-supervised learning approach (Principal Component Analysis) to to choose our axes. We construct a trajectory of user activity by projecting the user's statistics along the selected PCs at regular time intervals. This approach is applied to a popular e-learning software for K12 math education that is used by thousands of students worldwide. We identify cohorts of users according to the way their trajectory changes over time (e.g. monotone up, down, constant, etc). Each of the cohorts exhibits distinct behavioral dynamics and differed substantially in the amount of time users spent in the e-learning system. Specifically, one cohort included students that dropped out of the system after choosing very difficult problems that they were not able to complete, while another cohort included students users that chose more diverse problems and stayed longer in the system. In future work these results can be used by teachers or intelligent tutors to track students' engagement in the system and decide whether and how to intervene.

There are certain contexts, where we would like to analyze the behavior of small interacting systems, such as sports teams. While large interacting systems have drawn much attention in the past years, let it be physical systems of interacting particles or social networks, small systems are short of appropriate quantitative modeling and measurement tools. We propose a simple procedure for analyzing a small system through the degree in which its behavior at different granularity levels (e.g., dyads) non-linearly diverges from the simple additive behavior of its sub-units. For example, we may model the behavior of a soccer team by measuring the extent to which the behavior changes when we move from individual players to dyads, triads, and so on. In this paper, we address the challenge of modeling small systems in terms of measuring divergence from additivity at different granularity levels of the system. We present and develop a measure for quantifying divergence from additivity through what we term a Relative Entropy Lattice, and illustrate its benefits in modeling the behavior of a specific small system, a soccer team, using data from the English Premier League. Our method has practical implications too, such as allowing the coach to identify “hidden” weak spots in the team's behavior.

Machine translation (MT) is being used by millions of people daily, and therefore evaluating the quality of such systems is an important task. While human expert evaluation of MT output remains the most accurate method, it is not scalable by any means. Automatic procedures that perform the task of Machine Translation Quality Estimation (MT-QE) are typically trained on a large corpus of source–target sentence pairs, which are labeled with human judgment scores. Furthermore, the test set is typically drawn from the same distribution as the train. However, recently, interest in low-resource and unsupervised MT-QE has gained momentum. In this paper, we define and study a further restriction of the unsupervised MT-QE setting that we call oblivious MT-QE. Besides having no access no human judgment scores, the algorithm has no access to the test text’s distribution. We propose an oblivious MT-QE system based on a new notion of sentence cohesiveness that we introduce. We tested our system on standard competition datasets for various language pairs. In all cases, the performance of our system was comparable to the performance of the non-oblivious baseline system provided by the competition organizers. Our results suggest that reasonable MT-QE can be carried out even in the restrictive oblivious setting.