Evaluating large language models (LLMs) is crucial for both assessing their capabilities and identifying safety or robustness issues prior to deployment. Reliably measuring abstract and complex phenomena such as 'safety' and 'robustness' requires strong construct validity, that is, having measures that represent what matters to the phenomenon. With a team of 29 expert reviewers, we conduct a systematic review of 445 LLM benchmarks from leading conferences in natural language processing and machine learning. Across the reviewed articles, we find patterns related to the measured phenomena, tasks, and scoring metrics which undermine the validity of the resulting claims. To address these shortcomings, we provide eight key recommendations and detailed actionable guidance to researchers and practitioners in developing LLM benchmarks.

While the capabilities and utility of AI systems have advanced, rigorous norms for evaluating these systems have lagged. Grand claims, such as models achieving general reasoning capabilities, are supported with model performance on narrow benchmarks, like performance on graduate-level exam questions, which provide a limited and potentially misleading assessment. We provide a structured approach for reasoning about the types of evaluative claims that can be made given the available evidence. For instance, our framework helps determine whether performance on a mathematical benchmark is an indication of the ability to solve problems on math tests or instead indicates a broader ability to reason. Our framework is well-suited for the contemporary paradigm in machine learning, where various stakeholders provide measurements and evaluations that downstream users use to validate their claims and decisions. At the same time, our framework also informs the construction of evaluations designed to speak to the validity of the relevant claims. By leveraging psychometrics' breakdown of validity, evaluations can prioritize the most critical facets for a given claim, improving empirical utility and decision-making efficacy. We illustrate our framework through detailed case studies of vision and language model evaluations, highlighting how explicitly considering validity strengthens the connection between evaluation evidence and the claims being made.

Medical large language models (LLMs) research often makes bold claims, from encoding clinical knowledge to reasoning like a physician. These claims are usually backed by evaluation on competitive benchmarks; a tradition inherited from mainstream machine learning. But how do we separate real progress from a leaderboard flex? Medical LLM benchmarks, much like those in other fields, are arbitrarily constructed using medical licensing exam questions. For these benchmarks to truly measure progress, they must accurately capture the real-world tasks they aim to represent. In this position paper, we argue that medical LLM benchmarks should (and indeed can) be empirically evaluated for their construct validity. In the psychological testing literature, "construct validity" refers to the ability of a test to measure an underlying "construct", that is the actual conceptual target of evaluation. By drawing an analogy between LLM benchmarks and psychological tests, we explain how frameworks from this field can provide empirical foundations for validating benchmarks. To put these ideas into practice, we use real-world clinical data in proof-of-concept experiments to evaluate popular medical LLM benchmarks and report significant gaps in their construct validity. Finally, we outline a vision for a new ecosystem of medical LLM evaluation centered around the creation of valid benchmarks.

As AI systems appear to exhibit ever-increasing capability and generality, assessing their true potential and safety becomes paramount. This paper contends that the prevalent evaluation methods for these systems are fundamentally inadequate, heightening the risks and potential hazards associated with AI. I argue that a reformation is required in the way we evaluate AI systems and that we should look towards cognitive sciences for inspiration in our approaches, which have a longstanding tradition of assessing general intelligence across diverse species. We will identify some of the difficulties that need to be overcome when applying cognitively-inspired approaches to general-purpose AI systems and also analyse the emerging area of "Evals". The paper concludes by identifying promising research pathways that could refine AI evaluation, advancing it towards a rigorous scientific domain that contributes to the development of safe AI systems.

Customizable 3D avatar-based facial expression stimuli may improve user engagement in behavioral biomarker discovery and therapeutic intervention for autism, Alzheimer's disease, facial palsy, and more. However, there is a lack of customizable avatar-based stimuli with Facial Action Coding System (FACS) action unit (AU) labels. Therefore, this study focuses on (1) FACS-labeled, customizable avatar-based expression stimuli for maintaining subjects' engagement, (2) learning-based measurements that quantify subjects' facial responses to such stimuli, and (3) validation of constructs represented by stimulus-measurement pairs. We propose Customizable Avatars with Dynamic Facial Action Coded Expressions (CADyFACE) labeled with AUs by a certified FACS expert. To measure subjects' AUs in response to CADyFACE, we propose a novel Beta-guided Correlation and Multi-task Expression learning neural network (BeCoME-Net) for multi-label AU detection. The beta-guided correlation loss encourages feature correlation with AUs while discouraging correlation with subject identities for improved generalization. We train BeCoME-Net for unilateral and bilateral AU detection and compare with state-of-the-art approaches. To assess construct validity of CADyFACE and BeCoME-Net, twenty healthy adult volunteers complete expression recognition and mimicry tasks in an online feasibility study while webcam-based eye-tracking and video are collected. We test validity of multiple constructs, including face preference during recognition and AUs during mimicry.

As Large Language Models and Natural Language Processing (NLP) technology rapidly develop and spread into daily life, it becomes crucial to anticipate how their use could harm people. One problem that has received a lot of attention in recent years is that this technology has displayed harmful biases, from generating derogatory stereotypes to producing disparate outcomes for different social groups. Although a lot of effort has been invested in assessing and mitigating these biases, our methods of measuring the biases of NLP models have serious problems and it is often unclear what they actually measure. In this paper, we provide an interdisciplinary approach to discussing the issue of NLP model bias by adopting the lens of psychometrics -- a field specialized in the measurement of concepts like bias that are not directly observable. In particular, we will explore two central notions from psychometrics, the construct validity and the reliability of measurement tools, and discuss how they can be applied in the context of measuring model bias. Our goal is to provide NLP practitioners with methodological tools for designing better bias measures, and to inspire them more generally to explore tools from psychometrics when working on bias measurement tools.

The advent of large language models (LLMs) has revolutionized natural language processing, enabling the generation of coherent and contextually relevant human-like text. As LLMs increasingly power conversational agents used by the general public world-wide, the synthetic personality embedded in these models, by virtue of training on large amounts of human data, is becoming increasingly important. Since personality is a key factor determining the effectiveness of communication, we present a comprehensive method for administering and validating personality tests on widely-used LLMs, as well as for shaping personality in the generated text of such LLMs. Applying this method, we found: 1) personality measurements in the outputs of some LLMs under specific prompting configurations are reliable and valid; 2) evidence of reliability and validity of synthetic LLM personality is stronger for larger and instruction fine-tuned models; and 3) personality in LLM outputs can be shaped along desired dimensions to mimic specific human personality profiles. We discuss application and ethical implications of the measurement and shaping method, in particular regarding responsible AI.

The case experience of anesthesiologists is one of the leading causes of accidental dural punctures and failed epidurals - the most common complications of epidural analgesia used for pain relief during delivery. We designed a bimanual haptic simulator to train anesthesiologists and optimize epidural analgesia skill acquisition. We present an assessment study conducted with 22 anesthesiologists of different competency levels from several Israeli hospitals. Our simulator emulates the forces applied to the epidural (Touhy) needle, held by one hand, and those applied to the Loss of Resistance (LOR) syringe, held by the other one. The resistance is calculated based on a model of the epidural region layers parameterized by the weight of the patient. We measured the movements of both haptic devices and quantified the results' rate (success, failed epidurals, and dural punctures), insertion strategies, and the participants' answers to questionnaires about their perception of the simulation realism. We demonstrated good construct validity by showing that the simulator can distinguish between real-life novices and experts. Face and content validity were examined by studying users' impressions regarding the simulator's realism and fulfillment of purpose. We found differences in strategies between different level anesthesiologists, and suggest trainee-based instruction in advanced training stages.

Text embedding models from Natural Language Processing can map text data (e.g. words, sentences, documents) to supposedly meaningful numerical representations (a.k.a. text embeddings). While such models are increasingly applied in social science research, one important issue is often not addressed: the extent to which these embeddings are valid representations of constructs relevant for social science research. We therefore propose the use of the classic construct validity framework to evaluate the validity of text embeddings. We show how this framework can be adapted to the opaque and high-dimensional nature of text embeddings, with application to survey questions. We include several popular text embedding methods (e.g. fastText, GloVe, BERT, Sentence-BERT, Universal Sentence Encoder) in our construct validity analyses. We find evidence of convergent and discriminant validity in some cases. We also show that embeddings can be used to predict respondent's answers to completely new survey questions. Furthermore, BERT-based embedding techniques and the Universal Sentence Encoder provide more valid representations of survey questions than do others. Our results thus highlight the necessity to examine the construct validity of text embeddings before deploying them in social science research.