Canada

Some linguistic features are more readily learned than others, and are thereby more likely to be maintained in diachronic language change, giving rise to typological universals. Less readily learned features may give rise to typological gaps. We consider an apparent typological gap—that a morphologically superlative determiner (e.g., gleebest in &quot;gleebest of the cows&quot;) with a negative meaning is cross-linguistically unattested—and ask whether it reflects an underlying learning bias. We find 4-year-olds know that such determiners indicate quantity (replicating Wellwood, Gagliardi, &amp; Lidz, 2016), but only when positive (‘most’), but not negative (‘least’). Importantly, the observed bias is not specific to the apparent typological gap: same-age children showed difficulty learning the negative meaning of a non-superlative determiner, though such meanings are attested. The data thus suggest that children are generally biased against negativity, consistent with much prior work on conceptual bias and language learning/processing.

CogSci 2022

&quot;Most&quot; is easy but &quot;least&quot; is hard: Novel determiner learning in 4-year-olds

Some linguistic features are more readily learned than others, and are thereby more likely to be maintained in diachronic language change, giving rise to typological universals. Less readily learned features may give rise to typological gaps. We consider an apparent typological gap—that a morphologically superlative determiner (e.g., gleebest in "gleebest of the cows") with a negative meaning is cross-linguistically unattested—and ask whether it reflects an underlying learning bias. We find 4-year-olds know that such determiners indicate quantity (replicating Wellwood, Gagliardi, & Lidz, 2016), but only when positive (‘most’), but not negative (‘least’). Importantly, the observed bias is not specific to the apparent typological gap: same-age children showed difficulty learning the negative meaning of a non-superlative determiner, though such meanings are attested. The data thus suggest that children are generally biased against negativity, consistent with much prior work on conceptual bias and language learning/processing.

"Most" is easy but "least" is hard: Novel determiner learning in 4-year-olds

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. 





Facebook sends users friend suggestions when they have friends in common with others. However, how do people actually consider mutual friendships when inferring whether two individuals are friends? We examined how people (N=352) rely on numerical factors when judging whether individuals are friends based on their mutual connections. Participants saw two target individuals and their friends within a group, and judged if the targets were friends with each other. Experiment 1 manipulated the number of mutual friends and the number of friends each target had. Experiment 2 manipulated the proportion of these two factors. People were more likely to infer friendships when I) targets had more mutual friends; II) targets had fewer friends each; and III) there was a high proportion of mutual friends, which are all factors that impact the likelihood of targets’ friendship. In sum, people consider relevant numerical information about mutual friends when making friendship judgments.


Inferring friendships from mutual connections

How much information can people gain from being lied to? We propose that people can infer the truth from false messages if two preconditions are met: (1) bigger lies are more costly, and (2) speakers have known, directional deception goals. We tested this with a marble-flipping task in which a judge tried to accurately estimate the number of sampled marbles, while a sender attempted to make the judge over- or underestimate. The sender could produce larger lies about the number of marbles drawn by physically clicking marbles along a lower or higher cost function. We found that judges took into consideration both the senders' goals and costs to correct for bias introduced by senders' lies. Our paradigm allows us to show that a large amount of the variation can be explained by people correcting others' lies based on the lies that they themselves would produce.


Inferring truth from lies

Humans’ ability to attribute mental states to agents has been hypothesized to underpin our unique social behaviors. However, questions remain about the extent to which our representational Theory of Mind (ToM) is shared with non-human primates (NHPs). Here, we present a set of computational models each built to formalize a different representational theory of a foundational ToM component—understanding what others can see—and compare each model’s performance to that of NHPs across a range of previously published perspective-taking experiments. Our results show that multiple competing theories can account for NHPs’ perspective-taking abilities, including both human-like ToM and less complex mentalistic theories, but not simpler, non-mentalistic theories. This work supports the idea that NHPs may reason about others’ mental states when assessing their visual perspectives, and provides promising avenues for future work using computational modeling to determine if and how NHPs represent more complex mental states (e.g., ignorance, belief).

Multiple representational theories explain non-human primate perspective-taking: Evidence from computational modeling

Other people's mental states---what they want, what they know, and how they combine the two to act---are structured by the experiences that they've had. In line with this, we propose that inferences about other people's experiences are a central, but often neglected, aspect of human Theory of Mind. We explore this idea by presenting and testing a computational model that jointly infers others' desires, knowledge, and experience. We find that, by focusing inferences on others' experience, our model can make richer inferences about other's knowledge than would be otherwise possible. Our model quantitatively fits participant judgments on two experiments above an and beyond an alternative model. Overall, our work extends the richness of human Theory of Mind judgements that can be formalized as Bayesian inference over a generative model.

Integrating Experience into Bayesian Theory of Mind

Social learning is essential to survival. It is likely to evolve when it is more
efficient than asocial, trial-and-error learning. The consensus in cultural
evolutionary theory holds that some amount of environmental variability and
uncertainty about the best decisions are necessary for social learning to evolve.
However, current models for the evolution of social learning tend to conflate forms
of uncertainty, and rarely consider different ones in tandem. Moreover, many models
are limited by considering only two possible behaviors and environmental states. Here
we use evolutionary agent-based modeling to identify the complex ways in which
different forms of uncertainty affect social learning. We model a
time-varying environment with dozens of possible behaviors performed by agents
engaging in individual and social learning. We show that ambiguous payoffs, larger
possible decision sets, and shorter agent lifespans sometimes increase social
learning prevalence, as expected. However we also find 
which concrete uncertainty conditions cause evolution to select against social learning.


Some forms of uncertainty may suppress the evolution of social learning

There is much diversity in how people form theories and perspectives about situations. There are many instances today in which people may want others to question their perspectives as a first step to changing them. Aspects of counterfactual thinking such as mutability and availability may be involved in questioning a perspective. We explore this in an exploratory study (n = 80) using ambiguous social judgment scenarios. We performed path analysis on the data to compare four models, three of which are based on counterfactual generation. The best fitting model showed that availability was a mediator for mutability to predict our measure of questioning an initial perspective.

Changing Perspectives: Examining Factors Related To Counterfactual Thinking In Ambiguous Social Judgments

Ownership is pervasive across human societies but invisible to the naked eye, and so people need to infer who owns what. Given its ubiquity, one might expect that people's intuitive theories of ownership would converge. Yet, ownership disputes are common, raising the question of whether these intuitions might actually vary across people. We explored this question using a set of simple, parametrically varied object-transfer events where participants determined how much two agents owned an object. Participants showed strong consensus about ownership in some events, but opposing intuitions in others. Subject-level analyses suggest that these disagreements reflect two overarching intuitive theories, one where intentions are central to ownership, and another where physical possession is prioritized. Our results suggest that people have stable but different intuitive theories of ownership, which stem from their beliefs about the relationship between mental states and possession.


People Have Systematically Different Intuitions about Ownership even in Seemingly Simple Cases

People’s expectations play an important role in their reactions to events. There is often disappointment when events fail to meet expectations and a thrill to having one’s expectations exceeded (e.g. Mellers et al., 1997). Here, we examined how expectations influenced people's reactions to events with serious moral consequences, and their moral decisions concerning such situations. Participants judged pairs of events where a victim experienced a similar harm, but where victims were at different prior risk of being harmed (e.g. someone killed by a drunk driver on a Friday night versus a Saturday morning). People found unexpected harms to be more upsetting and they were more likely to choose to prevent an unexpected harm than an expected one. Further studies explored other factors that might have influenced judgments beyond expectations. This bias may have pernicious moral consequences, potentially leading to reduced concern for victims in most need of help.

Expectations bias moral judgments

Child-directed language (CDL) features words such as doggy, night-night, and tummy that are rarely used in adult-directed language (ADL). Characteristics of CDL variants, such as diminutivization and reduplication, explain why they may be learned and produced earlier by children. However, it is not yet clear how or when children switch to using ADL equivalents—dog, goodnight, stomach. Through analysis of speech transcripts from CHILDES and the Language Development Project corpus, we show that children significantly increase their production of ADL variants across age, with the average CDL-to-ADL transition point at 2.5 years. Many of the linguistic features that distinguish CDL vs. ADL registers (e.g., lexical and syntactic complexity) similarly differentiated the local speech contexts surrounding CDL vs. ADL variants in children’s input. Notably, these differences emerged even in speech that was primarily child-directed. Learners may therefore be able to capitalize on these linguistic cues to support their discovery of register along with context-appropriate CDL/ADL pair use.

From doggy to dog: Developmental shis in children's use of register-specific words

Historically, metaphors have facilitated creative change across multiple disparate domains. Similarly, human adults use metaphors to guide their everyday thinking and reasoning. While previous research in cognitive development has demonstrated that preschoolers understand metaphors, less is known about how preschoolers might learn from metaphors. The current experiments investigate whether preschoolers can use novel metaphors to make additional inferences about artifacts. Experiment 1 demonstrates that both four-year-olds and adults who hear novel positive and negative metaphors (e.g., “Daxes are clouds. Daxes are not suns.”) can form additional inferences based on these metaphor (e.g., that daxes let out water rather than light up). Experiment 2 conceptually replicates this result using a modified paradigm with only positive metaphors (e.g., “Daxes are clouds”). Taken together, these findings suggest that children can not only understand, but also learn from, metaphors. Consequently, metaphors may be a powerful learning mechanism in both adulthood and early childhood.

Premium content

"Most" is easy but "least" is hard: Novel determiner learning in 4-year-olds

Downloads

Next from CogSci 2022

Inferring friendships from mutual connections

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES