Canada

Being able to learn from small amounts of data is a key characteristic of human intelligence, but exactly how small? In this paper, we introduce a novel experimental paradigm that allows us to examine classification in an extremely data-scarce setting, asking whether humans can learn more categories than they have exemplars (i.e., can humans do &quot;less-than-one shot&#39;&#39; learning?). An experiment conducted using this paradigm reveals that people are capable of learning in such settings, and provides several insights into underlying mechanisms. First, people can accurately infer and represent high-dimensional feature spaces from very little data. Second, having inferred the relevant spaces, people use a form of prototype-based categorization (as opposed to exemplar-based) to make categorical inferences. Finally, systematic, machine-learnable patterns in responses indicate that people may have efficient inductive biases for dealing with this class of data-scarce problems.

CogSci 2022

Can Humans Do Less-Than-One-Shot Learning?

Being able to learn from small amounts of data is a key characteristic of human intelligence, but exactly how small? In this paper, we introduce a novel experimental paradigm that allows us to examine classification in an extremely data-scarce setting, asking whether humans can learn more categories than they have exemplars (i.e., can humans do "less-than-one shot'' learning?). An experiment conducted using this paradigm reveals that people are capable of learning in such settings, and provides several insights into underlying mechanisms. First, people can accurately infer and represent high-dimensional feature spaces from very little data. Second, having inferred the relevant spaces, people use a form of prototype-based categorization (as opposed to exemplar-based) to make categorical inferences. Finally, systematic, machine-learnable patterns in responses indicate that people may have efficient inductive biases for dealing with this class of data-scarce problems.

technical paper

Join us for CogSci2022! Our 44th Annual Conference will be centered around the theme of diversity, with an accompanying range of keynotes and invited speakers who themselves highlight the diversity of our field. Our three symposia are Cognition across Diverse Populations, Cognition within Diverse Societies, and Global Challenges for Cognitive Science. The conference is being offered in a fully hybrid environment for onsite and online attendees.
 
 [**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/07/Cogsci-2022-PAAG-Detailed-July-12final-1.pdf)
 [**The Detailed Program can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/08/DeArmond-CogSci2022-program-v113-final.pdf)

This conference requires a registration to attend. Please use the link below to secure your ticket. The website will be available to all registered attendees on July 25th. 

[Register to Attend](https://cognitivesciencesociety.org/registration/)

Registration Is Required

The 44th Annual Meeting of the Cognitive Science Society highlighted research on the theme Cognitive Diversity in addition to the full breadth and diversity of research topics offered by the society’s membership. 





Paradigms using the free recall of word lists have furthered our understanding of the organizational structure of memory by elucidating the role of contextual associations on memory search. We adapted the traditional word list-learning paradigm to investigate whether emphasizing contextual associations between items influences subsequent retrieval. Specifically, we introduced a review period between encoding and recall of word lists where items were repeated to highlight either the temporal or semantic associations at encoding. We found that temporal review led to stronger temporal clustering compared to a semantic or control review, and semantic review led to stronger semantic clustering compared to a temporal or control review. Moreover, participants recalled more list items when semantic associations were emphasized, with the degree of semantic clustering at recall predicting memory performance. These results demonstrate that emphasizing contextual associations during a repeated viewing after initial encoding can affect subsequent memory organization and recall.

Emphasizing associations from encoding affects free recall at retrieval

The free recall of triadic comparisons, a task used in clinical settings, presents a unique analysis challenge for many memory models, because learning occurs incidentally and items are presented multiple times in triads. To account for this design, we extend the SIMPLE (Brown et al., 2007) model of memory, which assumes to-be-remembered items are stored as separate logarithmically-compressed temporal traces. The ability to retrieve these traces depends on the acuity of memory probes and the semantic similarity between the items represented by the traces. We applied this model to a real-world clinical data set including healthy controls, people with mild cognitive impairment (MCI), and people with Alzheimer’s dementia. We found that people with MCI had lower acuity than healthy controls, but both groups placed roughly equal weight on temporal and semantic cues. People with dementia had both lower acuity and placed much more weight on temporal cues than semantic cues.

A model of free recall for multiple encounters of semantically related stimuli with an application to understanding cognitive impairment

Retrieval practice of information through testing has been shown to improve learning. So has studying examples. In this paper, we address inconsistencies in the literature concerning which of these two approaches is best. We test the hypothesis that learning depends on what is being learned; whereas practice emphasizes memorization, studying examples allows for selectivity of encoding, resulting in different information being learned. Accordingly, we predicted that practice will improve learning in situations that emphasize memorization (such as learning facts or simple associations), whereas studying examples will improve learning in situations where there are multiple pieces of information available and selectivity is necessary (such as when learning skills or procedures). We report evidence from a laboratory study using naturalistic materials showing results consistent with these predictions.

Learning depends on knowledge: The benefits of retrieval practice vary for facts and skills

Skill learning usually unfolds exponentially — we improve rapidly early on, and then performance levels off. However, we do not know whether people’s representations of skill learning accurately reflect this fact. Here, we asked people to predict the learning trajectory for a novel visuomotor task, “Lollitoss.” First, we established that skill learning unfolds exponentially on Lollitoss (Exp. 1). Across two experiments probing people’s trial-by-trial predictions of learning in Lollitoss using direct performance (Exp. 2a) and likelihood estimates (Exp. 2b), we found that people accurately represent the learning curve as exponential. However, we also found systematic errors - people think individuals start out better, make less errors, and learn slower in the task than in reality. Taken together, we find that people are surprisingly accurate at representing the overall shape of learning, but misestimate certain features, like the rate of learning, which may potentially have downstream effects on self-directed learning.

Thinking about doing: Representations of skill learning

We study the effect of Dutch proficiency on immigrants’ labour market performance, savings and financial wealth in the Netherlands. Different from past research, we had participants (N=659) take a language proficiency test apart from self-reported assessments, and measured participants’ IQ, patience, saving intention, risk aversion, self-control, temporal focus, etc. to better control for individual characteristics. Immigrants’ labour market performance and financial wealth were initially surveyed in 2016, and then again in 2020-2021. We find that Dutch proficiency affects immigrants’ earnings (employment probabilities; income; hourly wages) in 2016 and predicts participants’ earnings in 2021 even after controlling for the baseline in 2016, individual characteristics and demographic information. Furthermore, the results for the first time reveal that language proficiency can also predict immigrants’ current and future savings and financial wealth. Importantly, using an instrumental variables approach we show that language proficiency has a causal effect. Our findings have important theoretical and policy implications.

The value of host-country language: The effect of Dutch language proficiency on immigrants’ income, savings and financial wealth in the Netherlands

Over the last 50 years, there have been efforts on behalf of the US government to simplify public legal documents for the benefit of society at large . However, there has been no systematic evaluation of how effective these efforts--collectively referred to as the ``plain language movement''--have been. Here we report the results of a large-scale longitudinal corpus analysis (n$\approx$225 million words), in which we compare every law passed by congress between 1951 to 2009 (as well as concurrent resolutions and proclamations), with a comparably sized sample of English texts from four different genres published during the same time period. We find that laws remain laden with features associated with processing difficulty--including center-embedding, passive voice, low-frequency jargon and capitalization--relative to each of the four baseline genres of English, and that the prevalence of these features has not meaningfully declined since the onset of the plain language movement (in some cases, their prevalence has increased). These findings suggest that top-down efforts to simplify legal language have thus far remained largely ineffectual, despite the apparent tractability of these changes, raising and informing difficult questions of law and public policy.

So much for plain language: An analysis of the accessibility of United States federal laws (1951-2009)

What is the function of immature vocalizing in early learning environments? Previous work on infants in the US indicates that prelinguistic vocalizations elicit caregiver speech which is simplified in its linguistic structure. However, there is substantial cross-cultural variation in the extent to which children’s vocalizations elicit responses from caregivers. In the current study we ask whether children’s vocalizations elicit similar changes in their immediate caregivers’ speech structure across two cultural sites with differing perspectives on how to interact with infants and young children. Here we compare Tseltal Mayan and US caregivers’ verbal responses to their children’s vocalizations. Similar to findings from US dyads, we found that children from the Tseltal community regulate the statistical structure of caregivers’ speech simply by vocalizing. Following the interaction burst hypothesis, where clusters of child-adult contingent response alternations facilitate learning from limited input, we reveal a stable source of information facilitating language learning within ongoing interaction.

Immature vocalizations simplify the speech of Tseltal Mayan and US caregivers

Human language permits us to call to mind objects, events, and ideas that we cannot witness directly. This capacity requires that one links words not only to their referents, but to mental representations of those referents. Together with the recognition that words are used intentionally for communication, this link constitutes ‘verbal reference.’ Although the development of verbal reference is a pivotal achievement, questions concerning its origins remain. To address this gap, we investigate infants’ ability to establish a representation of an object that is hidden from view based on language input and to learn its name.

Semantic priming supports infants’ ability to represent and name unseen objects

Recent work has raised a question about whether adult language learners take advantage of indirect negative evidence (here, statistical preemption) while learning a new language. Statistical preemption predicts that exposure to conventional formulations results in better recognition that unconventional formulations are unacceptable. In a preregistered study, 61 undergraduates enrolled in Spanish classes were exposed to instances of conventional constructions in Spanish for 3 days to determine whether the exposure would bring their responses to unconventional formulations into closer alignment with those of native Spanish speakers. Judgment data confirms an effect of statistical preemption: students showed an increased recognition of the fact that unconventional (unwitnessed) formulations were unacceptable. These results are consistent with the idea that learning a new language is, in large part, learning which formulations to avoid: learning what not to say.

Language learners’ unacceptability judgments improve with repeated exposure to acceptable sentences

Children's early language skill has been linked to later educational outcomes, making it important to measure early language accurately. Parent-reported instruments such as the Communicative Development Inventories (CDIs) have been shown to provide reliable and valid measures of children's aggregate early language skill. However, CDIs contain hundreds of vocabulary items, some of which may not be heard (and thus learned) equally often by children of varying backgrounds. This study used a database of American English CDIs to identify words demonstrating strong bias for particular demographic groups of children, on dimensions of sex (male vs. female), race (white vs. non-white), and maternal education (high vs. low). For each dimension, we identified dozens of strongly biased items, and showed that eliminating even some of these items can reduce the magnitude of differences between groups. Additionally, we investigated how well the relative frequency of words spoken to young girls vs. boys predicted sex-based word learning bias, and discuss possible sources of demographic differences in early word learning.

Can Humans Do Less-Than-One-Shot Learning?

Downloads

Next from CogSci 2022

Emphasizing associations from encoding affects free recall at retrieval

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES