Large language models (LLMs) excel at generating empathic responses in text-based conversations. But, how reliably do they judge the nuances of empathic communication? We investigate this question by comparing how experts, crowdworkers, and LLMs annotate empathic communication across four evaluative frameworks drawn from psychology, natural language processing, and communications applied to 200 real-world conversations where one speaker shares a personal problem and the other offers support. Drawing on 3,150 expert annotations, 2,844 crowd annotations, and 3,150 LLM annotations, we assess inter-rater reliability between these three annotator groups. We find that expert agreement is high but varies across the frameworks' sub-components depending on their clarity, complexity, and subjectivity. We show that expert agreement offers a more informative benchmark for contextualizing LLM performance than standard classification metrics. Across all four frameworks, LLMs consistently approach this expert level benchmark and exceed the reliability of crowdworkers. These results demonstrate how LLMs, when validated on specific tasks with appropriate benchmarks, can support transparency and oversight in emotionally sensitive applications including their use as conversational companions.

Humans can be notoriously imperfect evaluators. They are often biased, unreliable, and unfit to define "ground truth." Yet, given the surging need to produce large amounts of training data in educational applications using AI, traditional inter-rater reliability (IRR) metrics like Cohen's kappa remain central to validating labeled data. IRR remains a cornerstone of many machine learning pipelines for educational data. Take, for example, the classification of tutors' moves in dialogues or labeling open responses in machine-graded assessments. This position paper argues that overreliance on human IRR as a gatekeeper for annotation quality hampers progress in classifying data in ways that are valid and predictive in relation to improving learning. To address this issue, we highlight five examples of complementary evaluation methods, such as multi-label annotation schemes, expert-based approaches, and close-the-loop validity. We argue that these approaches are in a better position to produce training data and subsequent models that produce improved student learning and more actionable insights than IRR approaches alone. We also emphasize the importance of external validity, for example, by establishing a procedure of validating tutor moves and demonstrating that it works across many categories of tutor actions (e.g., providing hints). We call on the field to rethink annotation quality and ground truth--prioritizing validity and educational impact over consensus alone.

In this study, we investigate the use of large language models (LLMs), specifically ChatGPT, for structured deductive qualitative coding. While most current research emphasizes inductive coding applications, we address the underexplored potential of LLMs to perform deductive classification tasks aligned with established human-coded schemes. Using the Comparative Agendas Project (CAP) Master Codebook, we classified U.S. Supreme Court case summaries into 21 major policy domains. We tested four intervention methods: zero-shot, few-shot, definition-based, and a novel Step-by-Step Task Decomposition strategy, across repeated samples. Performance was evaluated using standard classification metrics (accuracy, F1-score, Cohen's kappa, Krippendorff's alpha), and construct validity was assessed using chi-squared tests and Cramer's V. Chi-squared and effect size analyses confirmed that intervention strategies significantly influenced classification behavior, with Cramer's V values ranging from 0.359 to 0.613, indicating moderate to strong shifts in classification patterns. The Step-by-Step Task Decomposition strategy achieved the strongest reliability (accuracy = 0.775, kappa = 0.744, alpha = 0.746), achieving thresholds for substantial agreement. Despite the semantic ambiguity within case summaries, ChatGPT displayed stable agreement across samples, including high F1 scores in low-support subclasses. These findings demonstrate that with targeted, custom-tailored interventions, LLMs can achieve reliability levels suitable for integration into rigorous qualitative coding workflows.

Meningeal lymphatic vessels (MLVs) are responsible for the drainage of waste products from the human brain. An impairment in their functionality has been associated with aging as well as brain disorders like multiple sclerosis and Alzheimer's disease. However, MLVs have only recently been described for the first time in magnetic resonance imaging (MRI), and their ramified structure renders manual segmentation particularly difficult. Further, as there is no consistent notion of their appearance, human-annotated MLV structures contain a high inter-rater variability that most automatic segmentation methods cannot take into account. In this work, we propose a new rater-aware training scheme for the popular nnU-Net model, and we explore rater-based ensembling strategies for accurate and consistent segmentation of MLVs. This enables us to boost nnU-Net's performance while obtaining explicit predictions in different annotation styles and a rater-based uncertainty estimation. Our final model, MLV$^2$-Net, achieves a Dice similarity coefficient of 0.806 with respect to the human reference standard. The model further matches the human inter-rater reliability and replicates age-related associations with MLV volume.

Machine Learning models have many potentially beneficial applications in education settings, but a key barrier to their development is securing enough data to train these models. Labelling educational data has traditionally relied on highly skilled raters using complex, multi-class rubrics, making the process expensive and difficult to scale. An alternative, more scalable approach could be to use non-expert crowdworkers to evaluate student work, however, maintaining sufficiently high levels of accuracy and inter-rater reliability when using non-expert workers is challenging. This paper reports on two experiments investigating using non-expert crowdworkers and comparative judgement to evaluate complex student data. Crowdworkers were hired to evaluate student responses to open-ended reading comprehension questions. Crowdworkers were randomly assigned to one of two conditions: the control, where they were asked to decide whether answers were correct or incorrect (i.e., a categorical judgement), or the treatment, where they were shown the same question and answers, but were instead asked to decide which of two candidate answers was more correct (i.e., a comparative/preference-based judgement). We found that using comparative judgement substantially improved inter-rater reliability on both tasks. These results are in-line with well-established literature on the benefits of comparative judgement in the field of educational assessment, as well as with recent trends in artificial intelligence research, where comparative judgement is becoming the preferred method for providing human feedback on model outputs when working with non-expert crowdworkers. However, to our knowledge, these results are novel and important in demonstrating the beneficial effects of using the combination of comparative judgement and crowdworkers to evaluate educational data.

Distance metrics can only serve as proxy for similarity in information retrieval of similar instances. Learning a good similarity function from human annotations improves the quality of retrievals. This work uses deep metric learning to learn these user-defined similarity functions from few annotations for a large football trajectory dataset. We adapt an entropy-based active learning method with recent work from triplet mining to collect easy-to-answer but still informative annotations from human participants and use them to train a deep convolutional network that generalizes to unseen samples. Our user study shows that our approach improves the quality of the information retrieval compared to a previous deep metric learning approach that relies on a Siamese network. Specifically, we shed light on the strengths and weaknesses of passive sampling heuristics and active learners alike by analyzing the participants' response efficacy. To this end, we collect accuracy, algorithmic time complexity, the participants' fatigue and time-to-response, qualitative self-assessment and statements, as well as the effects of mixed-expertise annotators and their consistency on model performance and transfer-learning.

Since the inception of crowdsourcing, aggregation has been a common strategy for dealing with unreliable data. Aggregate ratings are more reliable than individual ones. However, many natural language processing (NLP) applications that rely on aggregate ratings only report the reliability of individual ratings, which is the incorrect unit of analysis. In these instances, the data reliability is under-reported, and a proposed k-rater reliability (kRR) should be used as the correct data reliability for aggregated datasets. It is a multi-rater generalization of inter-rater reliability (IRR). We conducted two replications of the WordSim-353 benchmark, and present empirical, analytical, and bootstrap-based methods for computing kRR on WordSim-353. These methods produce very similar results. We hope this discussion will nudge researchers to report kRR in addition to IRR.

Radiologists today play a key role in making diagnostic decisions and labeling images for training A.I. algorithms. Low inter-reader reliability (IRR) can be seen between experts when interpreting challenging cases. While teams-based decisions are known to outperform individual decisions, inter-personal biases often creep up in group interactions which limit non-dominant participants from expressing true opinions. To overcome the dual problems of low consensus and inter-personal bias, we explored a solution modeled on biological swarms of bees. Two separate cohorts; three radiologists and five radiology residents collaborated on a digital swarm platform in real time and in a blinded fashion, grading meniscal lesions on knee MR exams. These consensus votes were benchmarked against clinical (arthroscopy) and radiological (senior-most radiologist) observations. The IRR of the consensus votes was compared to the IRR of the majority and most confident votes of the two cohorts.The radiologist cohort saw an improvement of 23% in IRR of swarm votes over majority vote. Similar improvement of 23% in IRR in 3-resident swarm votes over majority vote, was observed. The 5-resident swarm had an even higher improvement of 32% in IRR over majority vote. Swarm consensus votes also improved specificity by up to 50%. The swarm consensus votes outperformed individual and majority vote decisions in both the radiologists and resident cohorts. The 5-resident swarm had higher IRR than 3-resident swarm indicating positive effect of increased swarm size. The attending and resident swarms also outperformed predictions from a state-of-the-art A.I. algorithm. Utilizing a digital swarm platform improved agreement and allows participants to express judgement free intent, resulting in superior clinical performance and robust A.I. training labels.

Medical toxicology is the clinical specialty that treats the toxic effects of substances, be it an overdose, a medication error, or a scorpion sting. The volume of toxicological knowledge and research has, as with other medical specialties, outstripped the ability of the individual clinician to entirely master and stay current with it. The application of machine learning techniques to medical toxicology is challenging because initial treatment decisions are often based on a few pieces of textual data and rely heavily on prior knowledge. ML techniques often do not represent knowledge in a way that is transparent for the physician, raising barriers to usability. Rule-based systems and decision tree learning are more transparent approaches, but often generalize poorly and require expert curation to implement and maintain. Here, we construct a probabilistic logic network to represent a portion of the knowledge base of a medical toxicologist. Our approach transparently mimics the knowledge representation and clinical decision-making of practicing clinicians. The software, dubbed Tak, performs comparably to humans on straightforward cases and intermediate difficulty cases, but is outperformed by humans on challenging clinical cases. Tak outperforms a decision tree classifier at all levels of difficulty. Probabilistic logic provides one form of explainable artificial intelligence that may be more acceptable for use in healthcare, if it can achieve acceptable levels of performance.

Saliency maps are a popular approach to creating post-hoc explanations of image classifier outputs. These methods produce estimates of the relevance of each pixel to the classification output score, which can be displayed as a saliency map that highlights important pixels. Despite a proliferation of such methods, little effort has been made to quantify how good these saliency maps are at capturing the true relevance of the pixels to the classifier output (i.e. their "fidelity"). We therefore investigate existing metrics for evaluating the fidelity of saliency methods (i.e. saliency metrics). We find that there is little consistency in the literature in how such metrics are calculated, and show that such inconsistencies can have a significant effect on the measured fidelity. Further, we apply measures of reliability developed in the psychometric testing literature to assess the consistency of saliency metrics when applied to individual saliency maps. Our results show that saliency metrics can be statistically unreliable and inconsistent, indicating that comparative rankings between saliency methods generated using such metrics can be untrustworthy.