The effectiveness of feedback in multiple-cue probability learning

How effective are different types of feedback in helping us to learn multiple contingencies? This article attempts to resolve a paradox whereby, in comparison to simple outcome feedback, additional feedback either fails to enhance or is actually detrimental to performance in nonmetric multiple-cue probability learning (MCPL), while in contrast the majority of studies of metric MCPL reveal improvements at least with some forms of feedback. In three experiments we demonstrate that if feedback assists participants to infer cue polarity then it can in fact be effective in nonmetric MCPL. Participants appeared to use cue polarity information to adopt a linear judgement strategy, even though the environment was nonlinear. The results reconcile the paradoxical contrast between metric and nonmetric MCPL and support previous findings of people's tendency to assume linearity and additivity in probabilistic cue learning.     

The effects of cognitive feedback and task complexity in multiple-cue probability learning

Subjects learned three multiple-cue probability learning (MCPL) tasks of varying complexity under conditions of feedforward and cognitive feedback. Statistical information about the task was presented in the feedforward condition; in the cognitive feedback condition, subjects were given statistical information about their own performance in addition to information about the task. Task complexity was manipulated by varying the function forms relating the values of the cues to the values of the criterion: The least complex task contained three positive linear function forms, the intermediate task contained three inverted U-shaped function forms, and there was one negative linear function form, one inverted U-shaped function form, and one U-shaped function form in the most complex task. The hypothesis that cognitive feedback is more effective than feedforward for the more complex task used here was not confirmed: There were no differences between the two groups with regard to r_a (achievement), G (knowledge), or R_s,² (cognitive control). The level of performance on all tasks, including the complex nonlinear tasks, was very high under both conditions, however, an unexpected and important finding.     

Missing information in multiple-cue probability learning

In a multiple-cue probability learning task, participants learned to use six discrete symptoms (i.e., cues) to diagnose which of three possible flu strains a hypothetical patient suffered from. For some patients, information regarding the status of certain symptoms was not available. Various possible ways in which the missing cue information might be processed were distinguished and tested in a series of three experiments (Ns = 80, 109, and 61). The results suggest that the judged probability of the outcome variable (i.e., flu strain) was assessed by "filling in" the missing cue information with a mean value based on previous observations. The predictions of other methods of processing missing cue information are inconsistent with the data.     

Choice strategies in multiple-cue probability learning

Choice strategies for selecting among outcomes in multiple-cue probability learning were investigated using a simulated medical diagnosis task. Expected choice probabilities (the proportion of times each outcome was selected given each cue pattern) under alternative choice strategies were constructed from corresponding observed judged probabilities (of each outcome given each cue pattern) and compared with observed choice probabilities. Most of the participants were inferred to have responded by using a deterministic strategy, in which the outcome with the higher judged probability is consistently chosen, rather than a probabilistic strategy, in which an outcome is chosen with a probability equal to its judged probability. Extended practice in the learning environment did not affect choice strategy selection, contrary to reports from previous studies, results of which may instead be attributable to changes with practice in the variability and extremity of the perceived probabilities on which the choices were based.     

The role of working memory capacity in multiple-cue probability learning

Multiple-cue probability learning (MCPL) involves learning to predict a criterion when outcome feedback is provided for multiple cues. A great deal of research suggests that working memory capacity (WMC) is involved in a wide range of tasks that draw on higher level cognitive processes. In three experiments, we examined the role of WMC in MCPL by introducing measures of working memory capacity, as well as other task manipulations. While individual differences in WMC positively predicted performance in some kinds of multiple-cue tasks, performance on other tasks was entirely unrelated to these differences. Performance on tasks that contained negative cues was correlated with working memory capacity, as well as measures of explicit knowledge obtained in the learning process. When the relevant cues predicted positively, however, WMC became irrelevant. The results are discussed in terms of controlled and automatic processes in learning and judgement.     

The role of working memory capacity in multiple-cue probability learning

Multiple-cue probability learning (MCPL) involves learning to predict a criterion when outcome feedback is provided for multiple cues. A great deal of research suggests that working memory capacity (WMC) is involved in a wide range of tasks that draw on higher level cognitive processes. In three experiments, we examined the role of WMC in MCPL by introducing measures of working memory capacity, as well as other task manipulations. While individual differences in WMC positively predicted performance in some kinds of multiple-cue tasks, performance on other tasks was entirely unrelated to these differences. Performance on tasks that contained negative cues was correlated with working memory capacity, as well as measures of explicit knowledge obtained in the learning process. When the relevant cues predicted positively, however, WMC became irrelevant. The results are discussed in terms of controlled and automatic processes in learning and judgement.     

Note on subjects' hypotheses in multiple-cue probability learning

Subjects' hypotheses in multiple-cue probability learning was studied by trial-by-trial verbal reports in tasks with linear and nonlinear cue-criterion relations. In accordance with earlier findings, the results of the present experiment show that subjects learned linear tasks rapidly, while nonlinear tasks were not learned at all. The subjects' verbal hypotheses show that the subjects were concerned with rules for combining cue values, rather than with cue-criterion functions. This explains why the subjects fail to learn nonlinear multiple-cue tasks. It also shows that, although the subjects seem to learn multiple-cue tasks through a hypothesis testing process of the same kind as in single-cue probability tasks, the hypotheses tested in multiple-cue tasks are different from those tested in single-cue tasks. Finally, the present results suggest that the picture of the subjects' cognitive models of multiple-cue probability learning tasks provided by the customary multiple regression analysis may be fundamentally misleading in that it suggests that the subjects learn the validity of each cue rather than a combination rule.     

The influence of formal,substantive, and contextual task properties on the relative effectiveness of different forms of feedback in multiple-cue probability learning tasks

Previous research shows that when participants are given information that allows them to compare task properties with properties of their cognitive system (cognitive feedback) they reach higher levels of achievement in multiple-cue probability learning (MCPL) tasks than when they are shown the correct answer after each trial (outcome feedback). It was hypothesized that this finding was in part due to the use of neutral task content in nearly all previous research, for neutral content fails to provide information about task properties in the outcome feedback condition. A study was conducted to compare the level of achievement reached with cognitive and outcome feedback under three conditions which varied the congruence between task properties implied by task content and actual task properties. The study supported the above hypothesis. When task content provided no task information, the level of achievement with cognitive feedback was higher than that reached with outcome feedback, even with perfect task predictability. When the task information provided by task content was congruent with actual task properties, however, the level of achievement with outcome feedback was as high as that with cognitive feedback. And when the task information provided by task content was incongruent with actual task properties, the level of achievement with cognitive feedback was as low as that with outcome feedback. Two additional experiments suggest that the effectiveness of cognitive feedback depends on contextual task properties, such as the number of blocks of trials and the credibility of the feedback. The results indicate that the relative effectiveness of outcome and cognitive feedback depends on formal, substantive, and contextual task properties.     

Note on detection of cue intercorrelation in multiple-cue probability learning

Abstract.— After completion of learning tasks with two different levels of cue intercorrelation (r _u=0.00 and 0.80) the detection of cue intercorrelation was tested with either a reproduction or a recognition method in a 2×2 factorial design. The reproduction method yielded an almost perfect matching between the reproduced intercorrelations and the intercorrelations of the tasks, while the recognition method resulted in systematically lower subjective intercorrelations. For the recognition method the subjective intercorrelation was found to vary with the criterion of recognition, so that the stricter the criterion, the higher the subjective intercorrelation.     

An hypothesis generation model for judgment in nonmetric multiple-cue probability learning

The purpose of this paper is to present a two-phase hypothesis generation model to describe behavior in multiple-cue probability learning tasks with nonmetric cues. The model assumes that on each trial the subject generates two sets of hypotheses: (a) a hypothesis concerning which cue dimension (or pattern) will lead to a correct prediction on that trial and (b) a hypothesis concerning which response will be correct given the cue dimension attended to on that trial.Five-hundred-twelve subjects were assigned to 20 groups in a binary choice task involving two binary cue dimensions. Each group observed cues which differed in validity. Analysis of the data indicated that subjects attend to both cue dimensions in making judgments even when one cue has zero validity. A test of the fit of the observed data to the asymptotic response proportions predicted by the model indicated a reasonable fit.     

Insight and strategy in multiple-cue learning

In multiple-cue learning (also known as probabilistic category learning) people acquire information about cue-outcome relations and combine these into predictions or judgments. Previous researchers claimed that people can achieve high levels of performance without explicit knowledge of the task structure or insight into their own judgment policies. It has also been argued that people use a variety of suboptimal strategies to solve such tasks. In three experiments the authors reexamined these conclusions by introducing novel measures of task knowledge and self-insight and using "rolling regression" methods to analyze individual learning. Participants successfully learned a four-cue probabilistic environment and showed accurate knowledge of both the task structure and their own judgment processes. Learning analyses suggested that the apparent use of suboptimal strategies emerges from the incremental tracking of statistical contingencies in the environment.     

The influence of shifting cue validity distributions and group discussion feedback on multiple cue probability learning

The effects of group discussion feedback and shifting cue validities on subjects' achievement, consistency, and matching in a multiple cue probability learning task were investigated in a task of constant system predictability. Students utilized distributions of meaningfully labeled cues of equal or unequal validity to predict final exam performance. Subjects' performance was examined in two conditions in which a shift in cue validities occurred and in two conditions in which the same validities were employed throughout the task. Feedback from group discussion was provided for half the subjects. Results indicated that subject performance was best when equally valid cues were utilized, especially after an anticipated shift in cue validities. Group discussion feedback facilitated performance only when a shift occurred. The results were discussed in the context of dynamic prediction environments.     

The effectiveness of distance learning initiatives in organizations

Today, organizations are increasingly adopting distance learning methods to train and develop their employees. Despite the widespread use of these methods, little research has been done regarding their effectiveness. The present paper reviews current literature on the effectiveness of distance learning methods in terms of employees’ reactions, learning, behavior, and organizational results. Suggestions for future research and practice are also offered.     

The self-choice effect from a multiple-cue perspective

The self-choice effect refers to the fact that self-chosen items are remembered better than experimenter-assigned items (Takahashi, 1991). The present study investigated the hypothesis that (a) response choice involves relational processing as activation of both target and context items, and (b) such activated context items are effective as potential retrieval cues for recall of target items. In the experiment, participants chose (choice condition) or were assigned (force condition) a target to remember for each trial. Prior to free recall of the target items, context words, related new words, or unrelated words were presented in a recognition task as potential retrieval cues. The results of a subsequent free recall test indicated that the incidental cues were more effective in the choice condition than in the force condition. Also, recognition resulted in a greater rate of successfully recognized context words at the cost of increasing falsely recognized related new words in the choice condition in comparison with the force condition. These results indicated that response choice activates context items at encoding, which operate as potential retrieval cues for recall of target items. Such cuing mechanisms operative in the self-choice effect are consistent with the multiple-cue theory proposed by Soraci et al. (1994; see also Soraci et al., 1999) for generative processing.     

The role of variable labels in cue probability learning tasks

Two experiments were conducted to investigate effects of variable labels on cue probability learning performance. Variable labels were manipulated to produce conditions in which the actual statistical structure was congruent or incongruent with that implied by the labels. In Experiment 1 the relevant aspect of statistical structure was the relative magnitude of validity of two cues. In Experiment 2, it was the direction of the slope relating the single cue to the criterion. In terms of task knowledge, congruent labels were better than either incongruent or abstract labels in both tasks, suggesting that congruent labels provide useful information about statistical structure. Prediction consistency with congruent labels was higher than with abstract in both experiments, and marginally better than with incongruent labels in Experiment 2. Compared to abstract labels, incongruent labels led to higher consistency in the two-cue task. In the one-cue task, both incongruent labels and neutral labels (which have content but do not imply any particular statistical structure) increased task knowledge. A proposed Rule Selection and Formation model attributes the advantages of variable labels to better retrieval, a reduced set of strategies, and modification of interpretations of labels.