Publications
Abstract Although authorship is central to the career advancement of many ecologists, the processes used to assign authorship credit on papers are not always clear, ethical, or equitable. We developed the concept of authorship climate (perceptions of fairness related to procedures, communication practices, and outcomes of authorship decisions) and measured it using a survey of early‐career scientists from four fields across U.S. Ph.D.‐granting institutions. The results pointed to important differences in authorship climate perceptions across fields, career stages, and demographics. This paper provides an overview of the results, especially for the field of biology that encompasses ecology, and provides recommendations for improving authorship climate for ecologists.
Language models can use verifiable rewards to improve at a wide variety of reasoning tasks. However, both parametric (e.g. RLVR) and non-parametric (e.g. prompt optimization) approaches to doing so typically require hundreds of training samples and thousands of model rollouts, making them expensive in the best case and intractable in the worst. To address this challenge, we introduce Contrastive Reflection (CORE), a non-parametric learning algorithm that compares past reasoning traces to generate insights: short natural-language descriptions of reasoning strategies and constraints that capture differences between successful and unsuccessful problem attempts. Across four reasoning tasks, we demonstrate that CORE enables more rapid improvement than both parametric (GRPO) and non-parametric (GEPA, episodic RAG, and MemRL) methods, while using fewer rollouts. Under fixed rollout budgets with as few as five training samples, CORE achieves the strongest performance in most task-data regimes. Finally, we highlight how CORE is substantially more context-efficient than non-parametric baselines, requiring fewer prompt tokens while storing learned knowledge as compact, interpretable natural-language insights. Our results therefore suggest that distilling contrasts between successful and unsuccessful reasoning traces into abstract and useful insights can provide a more efficient and interpretable route to model self-improvement than weight updates, prompt optimization, or direct reuse of stored reasoning traces.
Through building models of student thinking, researchers have identified quantitative reasoning as a foundation for students' mathematical development. This has generated a need to support teachers' capacity to teach for such reasoning. In this paper, we discuss a perspective on supporting prospective and practicing teachers in constructing meanings that foreground quantitative reasoning and research-based models of student thinking. We call this perspective competing meanings and describe it as a three-phase cognitive process. First is a perturbation of an extant meaning resulting from its enactment. Second is the construction of an alternative meaning to reconcile the perturbation. Third are acts of reflection to compare extant and alternative meanings. We introduce the competing meanings perspective and illustrate it with examples. We also discuss theoretical resources that inform this perspective, including Piagetian constructs, quantitative reasoning, and mathematical knowledge for teaching.
We examined potential processes by which children decide to make hard as opposed to easy choices to accomplish a goal. Five- to 7-year olds (N = 100, 52 girls, 61% White, 7% Hispanic or Latine, 5% Asian or Asian American, 5% Black or African American, 2% Native American, American Indian, or Alaskan Native, 13% Mixed Race, 7% not reporting) constructed a gear machine by choosing between easy and hard building instructions. Children chose to take on more difficult instructions based on their accuracy on previously difficult choices (OR = 3.60). Accuracy on easy choices did not relate to their subsequent decisions. These results suggest that children monitor performance on more challenging decisions and use this information to control their decision making about future efforts adaptively.
This research paper describes an exploratory study on the effectiveness of Chat Debugging: troubleshooting malfunctioning analog circuits on breadboards and printed circuit boards (PCB) by undergraduates through conversations with public-domain large language models (LLMs). Through thematic analysis of students' voluntarily shared chat logs when debugging pre-determined buggy circuits under exam and time pressure, we discovered multimodal usage patterns by students and considerable domain knowledge and sensible debugging suggestions offered by off-the-shelf LLMs. Meanwhile, we also identified major gaps in LLM technologies and students' skills during human-AI collaborative debugging, such as LLMs' limitations in 2D/3D image-based reasoning, unjustified tone of confidence, and students' deficits in fundamental concepts and critical thinking.
This study develops and validates a human-centered approach to measuring small-group, multidimensional engagement in a game-based Computer-Supported Collaborative Learning (CSCL) environment. Grounded in sociocultural perspectives, the proposed framework integrates multimodal data sources, including discourse, interaction logs, and linguistic indicators, to assess behavioral, collaborative, and disciplinary engagement through theory-informed human interpretation. Human-annotated engagement ratings serve as ground truth for evaluating whether supervised machine learning models can approximate multimodal interpretive judgments, establishing a pathway toward scalable assessment. Three supervised learning algorithms were compared to examine model performance across engagement dimensions, and multiple regression analysis was conducted to investigate associations between quest-level average engagement and group learning outcomes. Results indicate that average collaborative and disciplinary engagement ratings positively predict group performance, whereas behavioral engagement alone does not significantly relate to learning outcomes. Qualitative analysis of group responses further illustrates these associations. Methodologically and theoretically, the study advances multimodal learning analytics by proposing a generalizable, human-centered workflow that integrates multimodal interpretation with model-assisted scalability, emphasizing the essential role of human interpretation in producing valid and contextually grounded engagement assessments in collaborative learning environments.
Many college students experience introductory physics as psychologically threatening. In a preregistered RCT, we applied the biopsychosocial model of challenge-threat to describe patterns of threat in introductory physics and test whether a 5-d mindfulness training could reduce threat and increase engagement among undergraduates. Course-wide surveys (N = 954) screened students for the RCT and revealed roughly half of students experience psychological threat. Students identified with systemically excluded groups were more likely to experience psychological threat while systemically advantaged students were more likely to experience psychological challenge in their introductory physics course. In the RCT (N = 149), mindfulness training reduced psychological threat and fostered greater engagement in introductory physics, and, consistent with our theory of change, mindfulness training was associated with greater physics engagement through reductions in threat. The results demonstrate that mindfulness can help students manage stress more effectively by reducing psychological threat and fostering engagement in contexts like introductory physics.
Cognitive effort, defined as the relationship between cognitive load and task performance, offers insight into how individuals efficiently allocate mental resources during cognitively demanding activities. This metric is crucial in high-stakes public health and clinical training, where unmanaged cognitive overload has been linked to medical errors and workforce burnout. This study aims to examine whether cognitive effort varies systematically across task segments and whether it can be estimated at the individual level using brain signal data and machine learning. Method: Functional near-infrared spectroscopy (fNIRS) data were collected from 16 participants during a structured digital cognitive task comprising four sequential segments separated by short and long rest intervals. Cognitive effort was defined through relative neural efficiency and relative neural involvement, which combined measures of prefrontal hemodynamic activity with task performance. The analysis followed a two-stage approach. First, a segment-level group analysis assessed whether cognitive effort differed significantly across predefined task segments, thereby confirming that the task structure produced meaningful changes in cognitive demand. Second, participant-independent machine learning models predicted task performance from brain signal features. These predicted performance scores were then combined with neural measures to estimate cognitive effort at the individual level. Results: First, statistical analysis showed significant differences in cognitive effort across the four task segments. This confirms that even small changes in the assessment structure impact on the collective cognitive efficiency of the trainees. Then, we used machine learning on fNIRS data to predict individual performance scores. We found that the effort calculated from these predicted scores was nearly identical to that calculated from the actual scores. This result suggests that our effort metric is strongly based on brain signals. Conclusion: The findings demonstrate the feasibility of estimating cognitive effort from brain signals using artificial intelligence at both group and individual levels. Public Health Implications: Estimation of cognitive effort from digital task data may support scalable monitoring of cognitive workload and mental fatigue in technology-mediated environments, which complements subjective assessment methods used in public and digital health research.


