Reimagining the machine learning life cycle to improve educational outcomes of students.

Machine learning (ML) techniques are increasingly prevalent in education, from their use in predicting student dropout to assisting in university admissions and facilitating the rise of massive open online courses (MOOCs). Given the rapid growth of these novel uses, there is a pressing need to investigate how ML techniques support long-standing education principles and goals. In this work, we shed light on this complex landscape drawing on qualitative insights from interviews with education experts. These interviews comprise in-depth evaluations of ML for education (ML4Ed) papers published in preeminent applied ML conferences over the past decade. Our central research goal is to critically examine how the stated or implied education and societal objectives of these papers are aligned with the ML problems they tackle. That is, to what extent does the technical problem formulation, objectives, approach, and interpretation of results align with the education problem at hand? We find that a cross-disciplinary gap exists and is particularly salient in two parts of the ML life cycle: the formulation of an ML problem from education goals and the translation of predictions to interventions. We use these insights to propose an extended ML life cycle, which may also apply to the use of ML in other domains. Our work joins a growing number of meta-analytical studies across education and ML research as well as critical analyses of the societal impact of ML. Specifically, it fills a gap between the prevailing technical understanding of machine learning and the perspective of education researchers working with students and in policy.

Simplicial closure and higher-order link prediction.

Networks provide a powerful formalism for modeling complex systems by using a model of pairwise interactions. But much of the structure within these systems involves interactions that take place among more than two nodes at once-for example, communication within a group rather than person to person, collaboration among a team rather than a pair of coauthors, or biological interaction between a set of molecules rather than just two. Such higher-order interactions are ubiquitous, but their empirical study has received limited attention, and little is known about possible organizational principles of such structures. Here we study the temporal evolution of 19 datasets with explicit accounting for higher-order interactions. We show that there is a rich variety of structure in our datasets but datasets from the same system types have consistent patterns of higher-order structure. Furthermore, we find that tie strength and edge density are competing positive indicators of higher-order organization, and these trends are consistent across interactions involving differing numbers of nodes. To systematically further the study of theories for such higher-order structures, we propose higher-order link prediction as a benchmark problem to assess models and algorithms that predict higher-order structure. We find a fundamental difference from traditional pairwise link prediction, with a greater role for local rather than long-range information in predicting the appearance of new interactions.

Systematic review and meta-analysis of depression, anxiety, and suicidal ideation among Ph.D. students.

University administrators and mental health clinicians have raised concerns about depression and anxiety among Ph.D. students, yet no study has systematically synthesized the available evidence in this area. After searching the literature for studies reporting on depression, anxiety, and/or suicidal ideation among Ph.D. students, we included 32 articles. Among 16 studies reporting the prevalence of clinically significant symptoms of depression across 23,469 Ph.D. students, the pooled estimate of the proportion of students with depression was 0.24 (95% confidence interval [CI], 0.18-0.31; I2 = 98.75%). In a meta-analysis of the nine studies reporting the prevalence of clinically significant symptoms of anxiety across 15,626 students, the estimated proportion of students with anxiety was 0.17 (95% CI, 0.12-0.23; I2 = 98.05%). We conclude that depression and anxiety are highly prevalent among Ph.D. students. Data limitations precluded our ability to obtain a pooled estimate of suicidal ideation prevalence. Programs that systematically monitor and promote the mental health of Ph.D. students are urgently needed.

Breast Cancer Screening, Incidence, and Mortality Across US Counties.

IMPORTANCE: Screening mammography rates vary considerably by location in the United States, providing a natural opportunity to investigate the associations of screening with breast cancer incidence and mortality, which are subjects of debate. OBJECTIVE: To examine the associations between rates of modern screening mammography and the incidence of breast cancer, mortality from breast cancer, and tumor size. DESIGN, SETTING, AND PARTICIPANTS: An ecological study of 16 million women 40 years or older who resided in 547 counties reporting to the Surveillance, Epidemiology, and End Results cancer registries during the year 2000. Of these women, 53,207 were diagnosed with breast cancer that year and followed up for the next 10 years. The study covered the period January 1, 2000, to December 31, 2010, and the analysis was performed between April 2013 and March 2015. EXPOSURES: Extent of screening in each county, assessed as the percentage of included women who received a screening mammogram in the prior 2 years. MAIN OUTCOMES AND MEASURES: Breast cancer incidence in 2000 and incidence-based breast cancer mortality during the 10-year follow-up. Incidence and mortality were calculated for each county and age adjusted to the US population. RESULTS: Across US counties, there was a positive correlation between the extent of screening and breast cancer incidence (weighted r = 0.54; P < .001) but not with breast cancer mortality (weighted r = 0.00; P = .98). An absolute increase of 10 percentage points in the extent of screening was accompanied by 16% more breast cancer diagnoses (relative rate [RR], 1.16; 95% CI, 1.13-1.19) but no significant change in breast cancer deaths (RR, 1.01; 95% CI, 0.96-1.06). In an analysis stratified by tumor size, we found that more screening was strongly associated with an increased incidence of small breast cancers (≤2 cm) but not with a decreased incidence of larger breast cancers (>2 cm). An increase of 10 percentage points in screening was associated with a 25% increase in the incidence of small breast cancers (RR, 1.25; 95% CI, 1.18-1.32) and a 7% increase in the incidence of larger breast cancers (RR, 1.07; 95% CI, 1.02-1.12). CONCLUSIONS AND RELEVANCE: When analyzed at the county level, the clearest result of mammography screening is the diagnosis of additional small cancers. Furthermore, there is no concomitant decline in the detection of larger cancers, which might explain the absence of any significant difference in the overall rate of death from the disease. Together, these findings suggest widespread overdiagnosis.