Revealing human inductive biases for category learning by simulating cultural transmission

We explored people’s inductive biases in category learning—that is, the factors that make learning category structures easy or hard—using iterated learning. This method uses the responses of one participant to train the next, simulating cultural transmission and converging on category structures that people find easy to learn. We applied this method to four different stimulus sets, varying in the identifiability of their underlying dimensions. The results of iterated learning provide an unusually clear picture of people’s inductive biases. The category structures that emerge often correspond to a linear boundary on a single dimension, when such a dimension can be identified. However, other kinds of category structures also appear, depending on the nature of the stimuli. The results from this single experiment are consistent with previous empirical findings that were gleaned from decades of research into human category learning.     

The evolution of frequency distributions: Relating regularization to inductive biases through iterated learning

The regularization of linguistic structures by learners has played a key role in arguments for strong innate constraints on language acquisition, and has important implications for language evolution. However, relating the inductive biases of learners to regularization behavior in laboratory tasks can be challenging without a formal model. In this paper we explore how regular linguistic structures can emerge from language evolution by iterated learning, in which one person’s linguistic output is used to generate the linguistic input provided to the next person. We use a model of iterated learning with Bayesian agents to show that this process can result in regularization when learners have the appropriate inductive biases. We then present three experiments demonstrating that simulating the process of language evolution in the laboratory can reveal biases towards regularization that might not otherwise be obvious, allowing weak biases to have strong effects. The results of these experiments suggest that people tend to regularize inconsistent word-meaning mappings, and that even a weak bias towards regularization can allow regular languages to be produced via language evolution by iterated learning.     

Iterated learning: Intergenerational knowledge transmission reveals inductive biases

Cultural transmission of information plays a central role in shaping human knowledge. Some of the most complex knowledge that people acquire, such as languages or cultural norms, can only be learned from other people, who themselves learned from previous generations. The prevalence of this process of “iterated learning” as a mode of cultural transmission raises the question of how it affects the information being transmitted. Analyses of iterated learning utilizing the assumption that the learners are Bayesian agents predict that this process should converge to an equilibrium that reflects the inductive biases of the learners. An experiment in iterated function learning with human participants confirmed this prediction, providing insight into the consequences of intergenerational knowledge transmission and a method for discovering the inductive biases that guide human inferences.     

Language evolution by iterated learning with bayesian agents

Languages are transmitted from person to person and generation to generation via a process of iterated learning: people learn a language from other people who once learned that language themselves. We analyze the consequences of iterated learning for learning algorithms based on the principles of Bayesian inference, assuming that learners compute a posterior distribution over languages by combining a prior (representing their inductive biases) with the evidence provided by linguistic data. We show that when learners sample languages from this posterior distribution, iterated learning converges to a distribution over languages that is determined entirely by the prior. Under these conditions, iterated learning is a form of Gibbs sampling, a widely-used Markov chain Monte Carlo algorithm. The consequences of iterated learning are more complicated when learners choose the language with maximum posterior probability, being affected by both the prior of the learners and the amount of information transmitted between generations. We show that in this case, iterated learning corresponds to another statistical inference algorithm, a variant of the expectation-maximization (EM) algorithm. These results clarify the role of iterated learning in explanations of linguistic universals and provide a formal connection between constraints on language acquisition and the languages that come to be spoken, suggesting that information transmitted via iterated learning will ultimately come to mirror the minds of the learners.     

Language Evolution Can Be Shaped by the Structure of the World

Human languages vary in many ways but also show striking cross‐linguistic universals. Why do these universals exist? Recent theoretical results demonstrate that Bayesian learners transmitting language to each other through iterated learning will converge on a distribution of languages that depends only on their prior biases about language and the quantity of data transmitted at each point; the structure of the world being communicated about plays no role (Griffiths & Kalish, 2005 , 2007 ). We revisit these findings and show that when certain assumptions about the relationship between language and the world are abandoned, learners will converge to languages that depend on the structure of the world as well as their prior biases. These theoretical results are supported with a series of experiments showing that when human learners acquire language through iterated learning, the ultimate structure of those languages is shaped by the structure of the meanings to be communicated.     

When Extremists Win: Cultural Transmission Via Iterated Learning When Populations Are Heterogeneous

How does the process of information transmission affect the cultural or linguistic products that emerge? This question is often studied experimentally and computationally via iterated learning, a procedure in which participants learn from previous participants in a chain. Iterated learning is a powerful tool because, when all participants share the same priors, the stationary distributions of the iterated learning chains reveal those priors. In many situations, however, it is unreasonable to assume that all participants share the same prior beliefs. We present four simulation studies and one experiment demonstrating that when the population of learners is heterogeneous, the behavior of an iterated learning chain can be unpredictable and is often systematically distorted by the learners with the most extreme biases. This results in group‐level outcomes that reflect neither the behavior of any individuals within the population nor the overall population average. We discuss implications for the use of iterated learning as a methodological tool as well as for the processes that might have shaped cultural and linguistic evolution in the real world.     

The Conditional Reasoning Measurement System for Aggression: An Overview

The Conditional Reasoning Measurement System is described. This procedure focuses on how people solve what on the surface appear to be inductive reasoning problems. The true intent of the problems is to determine if solutions based on implicit biases are logically attractive to a respondent. In this article, we focus on the types of implicit biases that underlie aggressive individuals' attempts to justify aggressive behavior. People who consistently select solutions based on these types of biases are scored as being potentially aggressive because they are cognitively prepared to rationalize aggression. Scores on the Conditional Reasoning Test for Aggression (CRT-A) have been shown to have acceptable psychometric properties and an average, uncorrected validity of .44 against behavorial criteria (in 10 studies).     

A Bayesian Model of Biases in Artificial Language Learning: The Case of a Word‐Order Universal

In this article, we develop a hierarchical Bayesian model of learning in a general type of artificial language‐learning experiment in which learners are exposed to a mixture of grammars representing the variation present in real learners’ input, particularly at times of language change. The modeling goal is to formalize and quantify hypothesized learning biases. The test case is an experiment ( Culbertson, Smolensky, & Legendre, 2012 ) targeting the learning of word‐order patterns in the nominal domain. The model identifies internal biases of the experimental participants, providing evidence that learners impose (possibly arbitrary) properties on the grammars they learn, potentially resulting in the cross‐linguistic regularities known as typological universals. Learners exposed to mixtures of artificial grammars tended to shift those mixtures in certain ways rather than others; the model reveals how learners’ inferences are systematically affected by specific prior biases. These biases are in line with a typological generalization—Greenberg's Universal 18—which bans a particular word‐order pattern relating nouns, adjectives, and numerals.     

Induction, overhypothesis, and the origin of abstract knowledge. Evidence from 9-month-old infants

Human cognition relies on the ability to extract generalizable knowledge from limited evidence. One type of inductive learning, overhypothesis formation, allows learners to make inferences that take them beyond the limits of direct experience, leading to the creation of abstract knowledge. The developmental roots of this ability have yet to be investigated. We report three experiments examining whether 9-month-old infants are capable of forming overhypotheses. Our results show that when given evidence about a few objects in some category, infants formed a second-order generalization about categories in general. These findings provide evidence that infants possess a powerful mechanism for inductive learning-a mechanism that may be applied to many domains and that can account for the development of many inductive biases later on.     

Incorporating prior biases innetwork models of conceptual rule learning

A series of simulations is reported in which extant formal categorization models are applied to human rule-learning data (Salatas & Bourne, 1974). These data show that there are clear differences in the ease with which humans learn rules, with the conjunctive the easiest and the biconditional the hardest. The original ALCOVE model (an exemplar-based model), a configuralcue model, and two-layer backpropagation models did not fit the rule-learning data. ALCOVE successfully fit the data, however, when prior biases observed in human rule learning were implemented into weights of the network. Thus, current empirical learning models may not fare well in situations in which learners enter the concept-formation situation with preconceived biases regarding the kinds of concepts that are possible, but such biases might nevertheless be captured within these models. By incorporating preexperimental biases, ALCOVE may hold promise as a comprehensive category-learning model.     

Atypical birdsong and artificial languages provide insights into how communication systems are shaped by learning,use, and transmission

In this article, I argue that a comparative approach focusing on the cognitive capacities and behavioral mechanisms that underlie vocal learning in songbirds and humans can provide valuable insights into the evolutionary origins of language. The experimental approaches I discuss use abnormal song and atypical linguistic input to study the processes of individual learning, social interaction, and cultural transmission. Atypical input places increased learning and communicative pressure on learners, so exploring how they respond to this type of input provides a particularly clear picture of the biases and constraints at work during learning and use. Furthermore, simulating the cultural transmission of these unnatural communication systems in the laboratory informs us about how learning and social biases influence the structure of communication systems in the long run. Findings based on these methods suggest fundamental similarities in the basic social–cognitive mechanisms underlying vocal learning in birds and humans, and continuing research promises insights into the uniquely human mechanisms and into how human cognition and social behavior interact, and ultimately impact on the evolution of language.     

Modes of executive control in sequence learning: from stimulus-based to plan-based control

The authors argue that human sequential learning is often but not always characterized by a shift from stimulus- to plan-based action control. To diagnose this shift, they manipulated the frequency of 1st-order transitions in a repeated manual left-right sequence, assuming that performance is sensitive to frequency-induced biases under stimulus- but not plan-based control. Indeed, frequency biases tended to disappear with practice, but only for explicit learners. This tendency was facilitated by visual-verbal target stimuli, response-contingent sounds, and intentional instructions and hampered by auditory (but not visual) noise. Findings are interpreted within an event-coding model of action control, which holds that plans for sequences of discrete actions are coded phonetically, integrating order and relative timing. The model distinguishes between plan acquisition, linked to explicit knowledge, and plan execution, linked to the action control mode.     

Language as shaped by the brain

It is widely assumed that human learning and the structure of human languages are intimately related. This relationship is frequently suggested to derive from a language-specific biological endowment, which encodes universal, but communicatively arbitrary, principles of language structure (a Universal Grammar or UG). How might such a UG have evolved? We argue that UG could not have arisen either by biological adaptation or non-adaptationist genetic processes, resulting in a logical problem of language evolution. Specifically, as the processes of language change are much more rapid than processes of genetic change, language constitutes a "moving target" both over time and across different human populations, and, hence, cannot provide a stable environment to which language genes could have adapted. We conclude that a biologically determined UG is not evolutionarily viable. Instead, the original motivation for UG--the mesh between learners and languages--arises because language has been shaped to fit the human brain, rather than vice versa. Following Darwin, we view language itself as a complex and interdependent "organism," which evolves under selectional pressures from human learning and processing mechanisms. That is, languages themselves are shaped by severe selectional pressure from each generation of language users and learners. This suggests that apparently arbitrary aspects of linguistic structure may result from general learning and processing biases deriving from the structure of thought processes, perceptuo-motor factors, cognitive limitations, and pragmatics.     

Ontogenetic constraints on neural and behavioral plasticity: evidence from imprinting and face processing.

This paper addresses the extent and limits on brain plasticity during development through the detailed study of imprinting in the domestic chick and the development of face processing in human infants. In both of these systems, evidence for constraints on plasticity is reviewed. The first source of constraint comes from the basic architecture of learning mechanisms that support plasticity. With regard to the chick, a specific "Hebbian" model based on the known neural circuitry of the region of the brain involved is presented and discussed. In human infants, a more abstract model inspired by cortical circuitry is mentioned. The second source of constraint comes from biases on the nature of the stimuli selected for attention by the young organism. Both in the chick and the human there is evidence for a subcortical brain system which orients their attention toward conspecifics, and particularly to their faces. It is argued that these systems tutor, or bias the input to, the more plastic learning systems.     

Hypotheses and Inductive Predictions

This paper studies the use of hypotheses schemes in generatinginductive predictions. After discussing Carnap–Hintikka inductive logic,hypotheses schemes are defined and illustrated with two partitions. Onepartition results in the Carnapian continuum of inductive methods, the otherresults in predictions typical for hasty generalization. Following theseexamples I argue that choosing a partition comes down to making inductiveassumptions on patterns in the data, and that by choosing appropriately anyinductive assumption can be made. Further considerations on partitions makeclear that they do not suggest any solution to the problem of induction.Hypotheses schemes provide the tools for making inductive assumptions, but theyalso reveal the need for such assumptions.     

Phonological Concept Learning

Linguistic and non‐linguistic pattern learning have been studied separately, but we argue for a comparative approach. Analogous inductive problems arise in phonological and visual pattern learning. Evidence from three experiments shows that human learners can solve them in analogous ways, and that human performance in both cases can be captured by the same models. We test GMECCS (Gradual Maximum Entropy with a Conjunctive Constraint Schema), an implementation of the Configural Cue Model (Gluck & Bower, 1988a ) in a Maximum Entropy phonotactic‐learning framework (Goldwater & Johnson, 2003 ; Hayes & Wilson, 2008 ) with a single free parameter, against the alternative hypothesis that learners seek featurally simple algebraic rules (“rule‐seeking”). We study the full typology of patterns introduced by Shepard, Hovland, and Jenkins ( 1961 ) (“SHJ”), instantiated as both phonotactic patterns and visual analogs, using unsupervised training. Unlike SHJ, Experiments 1 and 2 found that both phonotactic and visual patterns that depended on fewer features could be more difficult than those that depended on more features, as predicted by GMECCS but not by rule‐seeking. GMECCS also correctly predicted performance differences between stimulus subclasses within each pattern. A third experiment tried supervised training (which can facilitate rule‐seeking in visual learning) to elicit simple rule‐seeking phonotactic learning, but cue‐based behavior persisted. We conclude that similar cue‐based cognitive processes are available for phonological and visual concept learning, and hence that studying either kind of learning can lead to significant insights about the other.     

Testing domain general learning in an Australian lizard

A key question in cognition is whether animals that are proficient in a specific cognitive domain (domain specific hypothesis), such as spatial learning, are also proficient in other domains (domain general hypothesis) or whether there is a trade-off. Studies testing among these hypotheses are biased towards mammals and birds. To understand constraints on the evolution of cognition more generally, we need broader taxonomic and phylogenetic coverage. We used Australian eastern water skinks (Eulamprus quoyii) with known spatial learning ability in three additional tasks: an instrumental and two discrimination tasks. Under domain specific learning we predicted that lizards that were good at spatial learning would perform less well in the discrimination tasks. Conversely, we predicted that lizards that did not meet our criterion for spatial learning would likewise perform better in discrimination tasks. Lizards with domain general learning should perform approximately equally well (or poorly) in these tasks. Lizards classified as spatial learners performed no differently to non-spatial learners in both the instrumental and discrimination learning tasks. Nevertheless, lizards were proficient in all tasks. Our results reveal two patterns: domain general learning in spatial learners and domain specific learning in non-spatial learners. We suggest that delineating learning into domain general and domain specific may be overly simplistic and we need to instead focus on individual variation in learning ability, which ultimately, is likely to play a key role in fitness. These results, in combination with previously published work on this species, suggests that this species has behavioral flexibility because they are competent across multiple cognitive domains and are capable of reversal learning.     

A Computational Theory of Learning Causal Relationships

I present a cognitive model of the human ability to acquire causal relationships. I report on experimental evidence demonstrating that human learners acquire accurate causal relationships more rapidly when training examples are consistent with a general theory of causality. This article describes a learning process that uses a general theory of causality as background knowledge. The learning process, which I call theory-driven learning (TDL), hypothesizes causal relationships consistent both with observed data and the general theory of causality. TDL accounts for data on both the rate at which human learners acquire causal relationships, and the types of causal relationships they acquire. Experiments with TDL demonstrate the advantage of TDL for acquiring causal relationships over similarity-based approaches to learning: Fewer examples are required to learn an accurate relationship.     

Identifying expectations about the strength of causal relationships

When we try to identify causal relationships, how strong do we expect that relationship to be? Bayesian models of causal induction rely on assumptions regarding people’s a priori beliefs about causal systems, with recent research focusing on people’s expectations about the strength of causes. These expectations are expressed in terms of prior probability distributions. While proposals about the form of such prior distributions have been made previously, many different distributions are possible, making it difficult to test such proposals exhaustively. In Experiment 1 we used iterated learning—a method in which participants make inferences about data generated based on their own responses in previous trials—to estimate participants’ prior beliefs about the strengths of causes. This method produced estimated prior distributions that were quite different from those previously proposed in the literature. Experiment 2 collected a large set of human judgments on the strength of causal relationships to be used as a benchmark for evaluating different models, using stimuli that cover a wider and more systematic set of contingencies than previous research. Using these judgments, we evaluated the predictions of various Bayesian models. The Bayesian model with priors estimated via iterated learning compared favorably against the others. Experiment 3 estimated participants’ prior beliefs concerning different causal systems, revealing key similarities in their expectations across diverse scenarios.     

Are unsupervised neural networks ignorant? Sizing the effect of environmental distributions on unsupervised learning

Learning environmental biases is a rational behavior: by using prior odds, Bayesian networks rapidly became a benchmark in machine learning. Moreover, a growing body of evidence now suggests that humans are using base rate information. Unsupervised connectionist networks are used in computer science for machine learning and in psychology to model human cognition, but it is unclear whether they are sensitive to prior odds. In this paper, we show that hard competitive learners are unable to use environmental biases while recurrent associative memories use frequency of exemplars and categories independently. Hence, it is concluded that recurrent associative memories are more useful than hard competitive networks to model human cognition and have a higher potential in machine learning.