skip to main content


Title: Learning to Learn Functions
Humans can learn complex functional relationships between variables from small amounts of data. In doing so, they draw on prior expectations about the form of these relationships. In three experiments, we show that people learn to adjust these expectations through experience, learning about the likely forms of the functions they will encounter. Previous work has used Gaussian processes—a statistical framework that extends Bayesian nonparametric approaches to regression—to model human function learning. We build on this work, modeling the process of learning to learn functions as a form of hierarchical Bayesian inference about the Gaussian process hyperparameters.  more » « less
Award ID(s):
2007278
PAR ID:
10470417
Author(s) / Creator(s):
; ; ; ;
Publisher / Repository:
Wiley
Date Published:
Journal Name:
Cognitive Science
Volume:
47
Issue:
4
ISSN:
0364-0213
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    The goal of item response theoretic (IRT) models is to provide estimates of latent traits from binary observed indicators and at the same time to learn the item response funcitons (IRFs) that map from latent trait to observed response. However, in many cases observed behavior can deviate significantly from the parametric assumptions of traditional IRT models. Nonparametric IRT (NIRT) models overcome these challenges by relaxing assumptions about the form of the IRFs, but standard tools are unable to simultaneously estimate flexible IRFs and recover ability estimates for respondents. We propose a Bayesian nonparametric model that solves this problem by placing Gaussian process priors on the latent functions defining the IRFs. This allows us to simultaneously relax assumptions about the shape of the IRFs while preserving the ability to estimate latent traits. This in turn allows us to easily extend the model to further tasks such as active learning. GPIRT therefore provides a simple and intuitive solution to several longstanding problems in the IRT literature. 
    more » « less
  2. Ab initio microkinetic modeling, parameterized using density functional theory (DFT) energies, is a common tool to quantify reaction rates and analyze reaction mechanisms a priori in heterogeneous catalysis. Such models, however, often have large prediction errors even if they include plausible reaction steps and correctly model the active sites; this is partially due to the intrinsic inaccuracies of the chosen DFT functional. Borrowing concepts from Bayesian calibration theory, we show that transferable data-driven corrections to DFT energies in the form of Gaussian process models trained on single-crystal adsorption calorimetry data can improve the accuracy of microkinetic models substantially. Specifically, we demonstrate that such corrections improve the predictive accuracy of the microkinetic model of the water-gas shift reaction on single-crystal Cu(111) surface by 3 orders of magnitude. We finally show that Gaussian process corrections serve as informed priors in a Bayesian experimental design framework to learn an accurate a posteriori microkinetic model from few kinetic experiments. We posit that these results suggest that even infusing small, related, high-fidelity thermochemistry data, when available, can systematically and substantially improve the predictive accuracy of microkinetic models. 
    more » « less
  3. Many institutions use undergraduate teaching assistants (tutors) in their computing courses to help provide more resources to students. Because of the role tutors play in students' learning experiences, recent work in computing education has begun to explore student-tutor interactions through the tutor's perspective and through direct observation of the interactions. The results suggest that these interactions are cognitively challenging for tutors and may not be as beneficial for students' learning as one might hope. Given that many of these interactions may be unproductive, this work seeks to understand how student expectations of these sessions might be impacting the interactions' effectiveness. We interviewed 15 students in a CS2 course to learn about the expectations and desires that students have when they attend tutoring sessions. Our findings indicate that there is variation in what students consider a desired result from the interaction, that assignment deadlines affect students' expectations and desires for interactions, and that students do not always want what they believe is beneficial for their learning. We discuss implications for instructors and potential guidance for students and tutors to make tutoring sessions more effective. 
    more » « less
  4. Davidesco, Ido (Ed.)
    Reasoning about visual representations in science requires the ability to control one’s attention, inhibit attention to irrelevant or incorrect information, and hold information in mind while manipulating it actively—all aspects of the limited-capacity cognitive system described as humans’ executive functions. This article describes pedagogical intuitions on best practices for how to sequence visual representations among pre-service teachers, adult undergraduates, and middle school children, with learning also tested in the middle school sample. Interestingly, at all ages, most people reported beliefs about teaching others that were different from beliefs about how they would learn. Teaching beliefs were most often that others would learn better from presenting representations one at a time, serially; while learning beliefs were that they themselves would learn best from simultaneous presentations. Students did learn best from simultaneously presented representations of mitosis and meiosis, but only when paired with self-explanation prompts to discuss the relationships between the graphics. These results provide new recommendations for helping students draw connections across visual representations, particularly mitosis and meiosis, and suggest that science educators would benefit from shifting their teaching beliefs to align with beliefs about their own learning from multiple visual representations. 
    more » « less
  5. Humans and autonomous agents often have differing knowledge about the world, the goals they pursue, and the actions they perform. Given these differences, an autonomous agent should be capable of rebelling against a goal when its completion would violate that agent’s preferences and motivations. Prior work on agent rebellion has examined agents that can reject actions leading to harmful consequences. Here we elaborate on a specific justification for rebellion in terms of violated goal expectations. Further, the need for rebellion is not always known in advance. So to rebel correctly and justifiably in response to unforeseen circumstances, an autonomous agent must be able to learn the reasons behind violations of its expectations. This paper provides a novel framework for rebellion within a metacognitive architecture using goal monitoring and model learning, and it includes experimental results showing the efficacy of such rebellion. 
    more » « less