About

Frank Keil is the Charles C. and Dorathea S. Dilley Professor of Psychology and Linguistics at Yale University. He earned his Ph.D. in 1977 from the University of Pennsylvania. His research broadly focuses on how children and adults construct causal interpretations of the world around them and how these interpretations compare to other ways of tracking information. Keil's work investigates how humans cognitively reduce the enormous causal complexity of the world into more manageable forms, akin to how image processing software compresses information to handle storage and computation requirements. He studies the nature of these coarse causal gists and what they do and do not capture about real-world causal relations, as well as how individuals recognize their own knowledge gaps and how they construe and access knowledge in other minds when filling in those gaps. His research is informed by a developmental perspective, exploring how early competencies form a foundation for more sophisticated causal understanding later in life. Keil's work also considers the interplay between domain-specific and domain-general processes, such as between folkbiology and folkphysics, and addresses how people naturally construct explanatory accounts of various phenomena, a concept he refers to as folk science.

Research topics

• Computer Science
• Neuroscience
• Cognitive science
• Social psychology
• Psychology
• Data science
• Developmental psychology
• World Wide Web

Selected publications

• Developing intellectual humility: questions, dilemmas, and future directions

  Current Psychology · 2026-03-17

  articleOpen access

  This article presents an overview and critique of current interdisciplinary research on the nature and development of intellectual humility (IH), with the aim of systematically outlining currently debated open questions. We focus on four specific areas of research: (1) theoretical questions regarding the nature of IH, (2) issues with the measurement of IH in development, (3) existing research on the development of IH and related socio-cognitive abilities, and (4) interventions to increase IH in children and adolescents. We critically review the existing empirical and theoretical literature in these areas, identify and articulate open questions, and map out directions for future research that follow from these questions. The main theoretical issues we identify concern the distinction between different features of IH (i.e., internal vs. external, self- vs. other-directed) and their relation to each other as well as the distinction between IH as a prescriptive virtue as opposed to a descriptive character trait. As we will demonstrate, taking seriously the notion of IH as a virtue raises crucial questions for its empirical study in the contexts of measurement, development, and potential interventions.

  PublisherOA PDFDOI

• Euclid preparation. Cosmology Likelihood for Observables in Euclid (CLOE). 2. Code implementation

  HAL (Le Centre pour la Communication Scientifique Directe) · 2026-03-26

  preprint

  International audience

  Publisher

• Illusions of Understanding in the Sciences

  2026-01-03 · 12 citations

  articleOpen accessSenior author

  Scientists seek to understand the causes of observed phenomena. Beliefs that they have succeeded are based on understanding that is rarely or possibly never complete, and varies in depth and quality. Most often scientists believe they understand more than they do, making their belief an illusion. This illusion then persists in explanations scientists provide in print, in talks, or in discussions. The illusion that a scientist has a valid and complete explanation tends to be magnified when the data is well described by mathematical and computer simulation models due to the precision of such models, and their ability to predict well; prediction does not imply causality, but gives the illusion that it does. The first part of this essay supports the case for the universality of partial and incomplete levels of understanding by showing the difficulty of reaching a deep level of understanding for even a simple analysis and model that most scientists use and believe they understand: linear regression. The second part highlights some implications of the existence of many levels of understanding and explanation, and their use by scientists for design, testing, analysis, and theory development. It discusses the way that deduction and induction depend on the levels of understanding and the implications of the illusion that a scientist’s understanding is deep. It makes a case that the many incomplete levels of understanding affect, often unwittingly, the ways scientists design experiments, test theories, comprehend, communicate, and teach.

  PublisherDOI

• Illusions of Understanding in the Sciences

  PsyArXiv (OSF Preprints) · 2026-01-02

  preprintOpen access1st authorCorresponding

  Scientists seek to understand the causes of observed phenomena. Beliefs that they have succeeded are based on understanding that is rarely or possibly never complete, and varies in depth and quality. Most often scientists believe they understand more than they do, making their belief an illusion. This illusion then persists in explanations scientists provide in print, in talks, or in discussions. The illusion that a scientist has a valid and complete explanation tends to be magnified when the data is well described by mathematical and computer simulation models due to the precision of such models, and their ability to predict well; prediction does not imply causality, but gives the illusion that it does. The first part of this essay supports the case for the universality of partial and incomplete levels of understanding by showing the difficulty of reaching a deep level of understanding for even a simple analysis and model that most scientists use and believe they understand: linear regression. The second part highlights some implications of the existence of many levels of understanding and explanation, and their use by scientists for design, testing, analysis, and theory development. It discusses the way that deduction and induction depend on the levels of understanding and the implications of the illusion that a scientist’s understanding is deep. It makes a case that the many incomplete levels of understanding affect, often unwittingly, the ways scientists design experiments, test theories, comprehend, communicate, and teach.

  Publisher

• No privileged link between intentionality and causation: Generalizable effects of agency in language

  Cognition · 2025-07-15 · 1 citations

  article
  PublisherDOI

• Money on my Mind: An Investigation of the Folk Classification of Money

  Synthese Library/Synthese library · 2025-01-01

  book-chapterSenior author
  PublisherDOI

• Euclid preparation. XCVI. Cosmology Likelihood for Observables in Euclid (CLOE). 3. Inference and Forecasts

  ArXiv.org · 2025-10-10

  preprintOpen access

  The Euclid mission aims to measure the positions, shapes, and redshifts of over a billion galaxies to provide unprecedented constraints on the nature of dark matter and dark energy. Achieving this goal requires a continuous reassessment of the mission's scientific performance, particularly in terms of its ability to constrain cosmological parameters, as our understanding of how to model large-scale structure observables improves. In this study, we present the first scientific forecasts using CLOE (Cosmology Likelihood for Observables in Euclid), a dedicated Euclid cosmological pipeline developed to support this endeavour. Using advanced Bayesian inference techniques applied to synthetic Euclid-like data, we sample the posterior distribution of cosmological and nuisance parameters across a variety of cosmological models and Euclid primary probes: cosmic shear, angular photometric galaxy clustering, galaxy-galaxy lensing, and spectroscopic galaxy clustering. We validate the capability of CLOE to produce reliable cosmological forecasts, showcasing Euclid's potential to achieve a figure of merit for the dark energy parameters $w_0$ and $w_a$ exceeding 400 when combining all primary probes. Furthermore, we illustrate the behaviour of the posterior probability distribution of the parameters of interest given different priors and scale cuts. Finally, we emphasise the importance of addressing computational challenges, proposing further exploration of innovative data science techniques to efficiently navigate the Euclid high-dimensional parameter space in upcoming cosmological data releases.

  PublisherOA PDFDOI

• Cosmology Likelihood for Observables in \Euclid (CLOE). 1. Theoretical recipe

  ArXiv.org · 2025-10-10

  preprintOpen access

  As the statistical precision of cosmological measurements increases, the accuracy of the theoretical description of these measurements needs to increase correspondingly in order to infer the underlying cosmology that governs the Universe. To this end, we have created the Cosmology Likelihood for Observables in Euclid (CLOE), which is a novel cosmological parameter inference pipeline developed within the Euclid Consortium to translate measurements and covariances into cosmological parameter constraints. In this first in a series of six papers, we describe the theoretical recipe of this code for the Euclid primary probes. These probes are composed of the photometric 3x2pt observables of cosmic shear, galaxy-galaxy lensing, and galaxy clustering, along with spectroscopic galaxy clustering. We provide this description in both Fourier and configuration space for standard and extended summary statistics, including the wide range of systematic uncertainties that affect them. This includes systematic uncertainties such as intrinsic galaxy alignments, baryonic feedback, photometric and spectroscopic redshift uncertainties, shear calibration uncertainties, sample impurities, photometric and spectroscopic galaxy biases, as well as magnification bias. The theoretical descriptions are further able to accommodate both Gaussian and non-Gaussian likelihoods and extended cosmologies with non-zero curvature, massive neutrinos, evolving dark energy, and simple forms of modified gravity. These theoretical descriptions that underpin CLOE will form a crucial component in revealing the true nature of the Universe with next-generation cosmological surveys such as Euclid.

  PublisherOA PDFDOI

• Euclid preparation. Cosmology Likelihood for Observables in Euclid (CLOE). 6: Impact of systematic uncertainties on the cosmological analysis

  ArXiv.org · 2025-10-11

  preprintOpen access

  Extracting cosmological information from the Euclid galaxy survey will require modelling numerous systematic effects during the inference process. This implies varying a large number of nuisance parameters, which have to be marginalised over before reporting the constraints on the cosmological parameters. This is a delicate process, especially with such a large parameter space, which could result in biased cosmological results. In this work, we study the impact of different choices for modelling systematic effects and prior distribution of nuisance parameters for the final Euclid Data Release, focusing on the 3$\times$2pt analysis for photometric probes and the galaxy power spectrum multipoles for the spectroscopic probes. We explore the effect of intrinsic alignments, linear galaxy bias, magnification bias, multiplicative cosmic shear bias and shifts in the redshift distribution for the photometric probes, as well as the purity of the spectroscopic sample. We find that intrinsic alignment modelling has the most severe impact with a bias up to $6\,σ$ on the Hubble constant $H_0$ if neglected, followed by mis-modelling of the redshift evolution of galaxy bias, yielding up to $1.5\,σ$ on the parameter $S_8\equivσ_8\sqrt{Ω_{\rm m} /0.3}$. Choosing a too optimistic prior for multiplicative bias can also result in biases of the order of $0.7\,σ$ on $S_8$. We also find that the precision on the estimate of the purity of the spectroscopic sample will be an important driver for the constraining power of the galaxy clustering full-shape analysis. These results will help prioritise efforts to improve the modelling and calibration of systematic effects in Euclid.

  PublisherOA PDFDOI

• Euclid preparation. XCVIII. Cosmology Likelihood for Observables in Euclid (CLOE). 5: Extensions beyond the standard modelling of theoretical probes and systematic effects

  ArXiv.org · 2025-10-10

  preprintOpen access

  Euclid is expected to establish new state-of-the-art constraints on extensions beyond the standard LCDM cosmological model by measuring the positions and shapes of billions of galaxies. Specifically, its goal is to shed light on the nature of dark matter and dark energy. Achieving this requires developing and validating advanced statistical tools and theoretical prediction software capable of testing extensions of the LCDM model. In this work, we describe how the Euclid likelihood pipeline, Cosmology Likelihood for Observables in Euclid (CLOE), has been extended to accommodate alternative cosmological models and to refine the theoretical modelling of Euclid primary probes. In particular, we detail modifications made to CLOE to incorporate the magnification bias term into the spectroscopic two-point correlation function of galaxy clustering. Additionally, we explain the adaptations made to CLOE's implementation of Euclid primary photometric probes to account for massive neutrinos and modified gravity extensions. Finally, we present the validation of these CLOE modifications through dedicated forecasts on synthetic Euclid-like data by sampling the full posterior distribution and comparing with the results of previous literature. In conclusion, we have identified in this work several functionalities with regards to beyond-LCDM modelling that could be further improved within CLOE, and outline potential research directions to enhance pipeline efficiency and flexibility through novel inference and machine learning techniques.

  PublisherOA PDFDOI

Recent grants

• The Role of Mechanistic Explanations in Learning about Science and Technology

  NSF · $2.3M · 2016–2022

• NIH Grant R01HD023922

  NIH · $1.8M · 2003

• NIH Grant R37HD023922

  NIH · $2.6M · 2014

Frequent coauthors

• Samuel G. B. Johnson

  University of Waterloo

  42 shared

• Sami R. Yousif

  University of Pennsylvania

  39 shared

• Brent Strickland

  Centre National de la Recherche Scientifique

  36 shared

• Kristi L. Lockhart

  Yale University

  26 shared

• Jonathan F. Kominsky

  Central European University

  24 shared

• Alexander Noyes

  21 shared

• Mark Sheskin

  15 shared

• Yarrow Dunham

  12 shared