The current study examined the impact of experiencing either a guided play or direct instruction learning environment on causal learning, curiosity, and divergent thinking. Forty-three children (age 4- to 6-years) participated in an online experiment in which they completed a causal learning task in either guided play or direct instruction condition. Children also completed measures of curiosity and divergent thinking and a second causal learning free exploration task after which they were tested on their causal learning. It was predicted that children in the guided play condition would perform better than children in the direct instruction condition on a test of causal learning because guided play provides a balance of child agency and adult guidance that is optimal for supporting learning. Contrary to the hypothesis, children in the direct instruction condition performed better on the test of causal learning. This finding is likely due to the cognitive demands placed on children in the guided play condition. These demands were likely the result of completing the task in an online environment, and the current study has implications for learning and conducting research online. Based on prior research, it was also predicted that children in the guided play condition would outperform children in the direct instruction condition on measures of curiosity and divergent thinking, and that curiosity would also predict children’s causal learning. There were no effects of condition on either curiosity or divergent thinking, but curiosity did predict children’s scores on the test of causal learning. This finding suggests that curiosity is a powerful driver of children’s learning and deserves further investigation. / Psychology

In many circumstances, predictions elicited from induced classification models are useful to a certain extent, as such predictions provide insight into what the future may hold. Such models, in and of themselves, hold little value beyond making such predictions, as they are unable to inform their user as to how to change a predicted outcome. Consider, for example, a health care domain where a classification model has been induced to learn the mapping from patient characteristics to disease outcome. A patient may want to know how to lessen their probability of developing such a disease. In this document, four different approaches to inverse classification, the process of turning predictions into prescriptions by working backwards through an induced classification model to optimize for a particular outcome of interest, are explored. The first study develops an inverse classification framework, which is created to produce instance-specific, real-world feasible recommendations that optimally improve the probability of a good outcome, while being as classifier-permissive as possible. Real-world feasible recommendations are obtained by imposition of constraints that specify which features can be optimized over and accounts for user-specific preferences. Assumptions are made as to the differentiability of the classification function, permitting the use of classifiers with exploitable gradient information, such as support vector machines (SVMs) and logistic regression. Our results show that the framework produces real-world recommendations that successfully reduce the probability of a negative outcome. In the second study, we further relax our assumptions as to the differentiability of the classifier, allowing virtually any classification function to be used. Correspondingly, we adjust our optimization methodology. To such an end, three heuristic-based optimization methods are devised. Furthermore, non-linear (quadratic) relationships between feature changes and so-called cost, which accounts for user preferences, are explored. The results suggest that non-differentiable classifiers, such as random forests, can be successfully navigated using the specified framework and updated, heuristic-based optimization methodology. Furthermore, findings suggest that regularizers, encouraging sparse solutions, should be used when quadratic/non-linear cost-change relationships are specified. The third study takes a longitudinal approach to the problem, exploring the effects of applying the inverse classification process to instances across time. Furthermore, we explore the use of added temporal linkages, in the form of features representing past predicted outcome probability (i.e., risk), on the inverse classification results. We further explore and propose a solution to a missing data subproblem that frequently arises in longitudinal data settings. In the fourth and final study, a causal formulation of the inverse classification framework is provided and explored. The formulation encompasses a Gaussian Process-based method of inducing causal classifiers, which is subsequently leveraged when the inverse classification process is applied. Furthermore, exploration of the addition of certain dependencies is explored. The results suggest the importance of including such dependencies and the benefits of taking a causal approach to the problem.

Using an allergist task, Uengoer, Lotz and Pearce (2013) found that in a design A+/AX+/BY+/CY-, the blocked cue X was indicated to cause the outcome to a greater extent than the uncorrelated cue Y. This finding has been termed “the redundancy effect” by Pearce and Jones (2015). According to Vogel and Wagner (2017), the redundancy effect “presents a serious challenge for those theories of conditioning that compute learning through a global error-term” (p. 119). One such theory is the Rescorla-Wagner (1972) model, which predicts the opposite result, that Y will have a stronger association with the outcome than X. This thesis explored the basis of the redundancy effect in human causal learning. Evidence from Chapter 2 suggested that the redundancy effect was unlikely to have been due to differences in attention between X and Y. Chapter 3 explored whether differences in participants’ certainty about the causal status of X and of Y contributed to the redundancy effect. Manipulations aimed at disambiguating the effects that X had on the outcome, including outcome-additivity training and low outcome rate, resulted in lower ratings for this cue and a smaller redundancy effect. However, the redundancy effect was still significant with both manipulations, suggesting that while participants’ uncertainty about the causal status of X contributed to it, there may have been other factors. Chapter 4 investigated whether another factor was a lack of inhibition for cue C. In a scenario where inhibition was more plausible than in an allergist task, a negative correlation between causal ratings for C and for Y, and a positive correlation between ratings for C and the magnitude of the redundancy effect, were found. In addition, establishing C as inhibitory resulted in a smaller redundancy effect than establishing C as neutral. Overall, findings of this thesis suggest that the redundancy effect in human causal learning is the result of participants’ uncertainty about the causal status of X, and a lack of inhibition for C. Further work is recommended to explore whether combining manipulations targeting X and Y would reverse the redundancy effect, whether effects of outcome additivity and outcome rate on X are the result of participants’ uncertainty about this cue, and the extent to which participants rely on single versus summed error.

No description available.

No description available.

No description available.