Asking the right questions about the psychology of human inquiry: Nine open challenges

Coenen, Anna; Nelson, Jonathan D.; Gureckis, Todd M.

doi:10.3758/s13423-018-1470-5

Cited by 78 publications

(88 citation statements)

References 225 publications

(292 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Klayman and Ha's account has since been broadly adopted to explain PTS in adult learning (see Coenen, Nelson, & Gureckis, 2018), and similar arguments have been successfully applied to other instances of PTS beyond the 2-4-6 task (e.g., Oaksford & Chater, 1994). Navarro and Perfors (2011) have also extended this argument by demonstrating that PTS is a 'near-optimal' learning strategy in contexts where correct hypotheses are 'sparse,' and provide further justification for learners' intuitive assumption of sparsity.…”

Section: Pts As a Means Of Generating Evidencementioning

confidence: 99%

“…Conclusion: Relationship to Truth The claim that we are 'intuitive scientists' in our exploratory learning is well established as part of the account of human inquiry (Coenen et al, 2018). However, the fact that self-directed learners often choose to conduct repeated positive tests of their hypotheses, rather than (apparently) more informative interventions has historically complicated this claim.…”

Section: The Search For Invariance (Si) Hypothesismentioning

confidence: 99%

See 1 more Smart Citation

The Search for Invariance: Repeated Positive Testing Serves the Goals of Causal Learning

Lapidow

Walker

2020

Language and Concept Acquisition From Infancy Through Childhood

View full text Add to dashboard Cite

To appear in Childers, J.B., Graham, S.A. & Namy, L. (Eds). (2019). Learning Language and Concepts from Multiple Examples in Infancy and Childhood What is invariant does not emerge unequivocally except with a flux. The essentials become evident in the context of changing nonessentials.-James Gibson, 1979 Human learners are intuitively exploratory: We acquire new knowledge from the outcomes of our actions. However, in order for exploration to support learning, at least some of these actions must serve to evaluate our existing knowledge. Despite this need for informative 'hypothesis testing' in everyday learning, decades of research examining self-directed experimentation suggests that learners rarely choose informative tests. That is, instead of selecting actions to test whether their current hypothesis is correct, both children and adults tend to prefer 'positive tests': actions that will produce an effect assuming their current hypothesis is correct (see Klayman, 1995; Zimmerman, 2007). To illustrate, suppose you drop an ice cube on the floor and it shatters. As a learner, you might form an initial hypothesis that 'impact with an unyielding surface causes ice to shatter.' This hypothesis is also a causal explanation for your observation: indicating how one variable (X) makes a difference to the state of another variable (Y). According to traditional interpretations of Popper's (1959) falsificationist approach, testing this hypothesis would require disconfirming its alternatives. That is, assessing whether 'X is the cause of Y,' requires 'negative tests,' or actions to determine whether Y occurs in the absence of X. Here, since Y is 'shattering' and X is 'impacting an unyielding surface,' you should drop an ice cube on a yielding surface (not-X), like rubber or cotton, to determine whether it will shatter. However, learners rarely choose this kind of disconfirming action during their exploration. Instead, they are much more likely to repeat the initial observation: e.g., to pick up another ice-cube and drop it on the same surface, or a similar one. This tendency to generate multiple positive examples is a puzzling characteristic of self-directed learning, since it does not initially appear to be informative. After all, these repeated demonstrations often produce the same evidence, and do not distinguish between the current hypothesis (i.e., impacting an unyielding surface) and potential alternatives (e.g., impacting any surface at a particular speed), since they are consistent with both. Why then, do self-directed learners consistently and repeatedly conduct positive tests? In this chapter, we propose a novel answer to this question: the Search for Invarinace (SI) hypothesis, which suggests that observing muliple, positive examples may facilitate learning by allowing us to assess the invariance of our causal theories. That is, by repeatedly activating a hypothesized cause and checking if its anticipated effect occurs, positive tests generate information about the degree to which this relationship holds across tim...

show abstract

Section: Pts As a Means Of Generating Evidencementioning

confidence: 99%

Section: The Search For Invariance (Si) Hypothesismentioning

confidence: 99%

The Search for Invariance: Repeated Positive Testing Serves the Goals of Causal Learning

Lapidow

Walker

2020

Language and Concept Acquisition From Infancy Through Childhood

View full text Add to dashboard Cite

show abstract

“…The main challenge for AML method is to predict which question a human will ask from the given context. A number of recent studies have discussed this challenge [33][34][35]. Rothe et al in [33] proposed a model that predicts what questions human learners will ask and can creatively generate novel questions that did not exist in the training data.…”

Section: Prediction Of the Next Questionmentioning

confidence: 99%

“…Their work in [34] showed that human can accurately evaluate question quality by using the Bayesian ideal observers. In the most recent review [35], authors highlight and discuss nine challenges about the psychology of human inquiry.…”

Section: Prediction Of the Next Questionmentioning

confidence: 99%

Human Active Learning

Radwan¹

2019

Active Learning - Beyond the Future

View full text Add to dashboard Cite

Active machine learning (AML) is a popular research area in machine learning. It allows selection of the most informative instances in training data of the domain for manual labeling. AML aims to produce a highly accurate classifier using as few labeled instances as possible, thereby minimizing the cost of obtaining labeled data. As machines can learn from experience like humans do, using AML for human category learning may help human learning become more efficient and hence reduce the cost of teaching. This chapter is a review of recent research literature concerning the use of AML technique to enhance human learning and teaching. There are a few studies on the applications of AML to the human category learning domain. The most interesting study was by Castro et al., which showed that humans learn faster with better performance when they can actively select the informative instances from a pool of unlabeled data instead of random sampling. Although AML can facilitate object categorization for humans, there are still many challenges and questions that need to be addressed in the use of AML for modeling human categorization. In this chapter, we will discuss some of these challenges. Figure 5. A sample of images from the dataset used for teaching and test. Columns L1-L4 show different difficulty levels.Active Learning -Beyond the Future 26

show abstract

“…In the past years, a strong interest in human information search and active learning has emerged, with several studies finding beneficial effects of active compared to passive learning (see Coenen, Nelson, & Gureckis, 2018, for a review). For instance, Lagnado and Sloman (2004) found that learners who were given the opportunity to actively intervene on a causal system made more accurate inferences than passive learners who could not freely decide which information to obtain (also see Steyvers, Tenenbaum, Wagenmakers, & Blum, 2003).…”

Section: Active Learningmentioning

confidence: 99%

Learning functions actively

Jones¹,

Schulz²,

Meder³

et al. 2019

Preprint

View full text Add to dashboard Cite

How do people actively learn functional rules, i.e. a mapping of continuous inputs onto a continuous output? We investigate information search behavior in a multiple-featurefunction learning task in which participants either actively select or passively receive observations. We find that participants benefit from actively selecting information, inparticular in their function extrapolation performance. By introducing and comparing different models of active function learning, we find that participants are best described by a non-parametric function learning model that learns about both the underlying function and inputs that are likely to produce high outputs. These results enrich our understandingof active function learning in complex domains.

show abstract

Asking the right questions about the psychology of human inquiry: Nine open challenges

Cited by 78 publications

References 225 publications

The Search for Invariance: Repeated Positive Testing Serves the Goals of Causal Learning

The Search for Invariance: Repeated Positive Testing Serves the Goals of Causal Learning

Human Active Learning

Learning functions actively

Contact Info

Product

Resources

About