Charles F. Gettys

An evaluation of human act generation performance

Abstract Act generation is important in decision making in ill-defined problems where the subject must synthesize actions that might solve the problem. Two experiments explored human abilities to generate actions which might solve ill-defined decision problems. Subjects were given unlimited time to suggest as many solutions as possible. Their suggestions were compared to a hierarchical structural model of the actions developed by the experimenters that could be taken to solve the problem. Although subjects were capable of generating several actions that might be worth taking, their suggestions were far from complete. The second experiment replicated and extended these results by introducing instructions to generate quality actions. It employed a structural model derived from cluster analysis, an improved utility estimation technique, and offered substantial monetary incentives for quality or quantity of actions generated. This study confirmed the general conclusion that subjects fail to generate important high-utility actions. The implications of this result are discussed in respect to decision analysis in which a complete structural model of the decision problem is highly desirable.

Hypothesis plausibility and hypothesis generation

Abstract A model of the process of generating hypotheses in an inference task is developed. First, a hypothesis retrieval process is assumed which consists of a directed recursive memory search of long-term memory. This memory search is assumed to be triggered by a second plausibility estimation process which assesses the plausibility of hypotheses currently held by the decision maker. Research was conducted to examine these and other assumptions of this model. The assumption that new hypotheses are generated when information renders the currently held hypotheses less probable was supported by the results. A hypothesized plausibility threshold process was also studied. The results suggested that new hypotheses are introduced only if their plausibility is high enough to make them active competitors for the most likely hypothesis.

Plausibility assessments in hypothesis generation

Abstract Our principle goal was to examine one aspect of predicisional behavior: how decision makers evaluate their feelings of certainty toward specified and unspecified hypotheses. Our experimental task permitted a comparison of subjective to veridical values. Subjects generally were overconfident for specified hypotheses and conservative for unspecified hypotheses. Subjects also expressed more confidence with specified hypotheses when evaluating them individually than as an aggregate. An explanation for our results may be that subjects underestimated the importance of unspecified hypotheses because they had difficulty recalling them.

Divergent thinking in act generation

Abstract Act generation is a process used in decision making and problem solving to create possible actions that might solve a problem. An act generation experiment was conducted using undergraduates and advanced graduate students. Large performance differences were observed between the two groups in act generation performance and divergent thinking ability. Divergent thinking ability, as measured by Guilfords Alternate Uses test, was approximately twice as high for the graduate subjects as compared to the undergraduates, and similar differences were observed in act generation performance. The correlation between divergent thinking scores and two act generation performance measures were 0.43 and 0.49 for the combined groups. These results suggests that divergent thinking ability is an important predictor of act generation performance.

A Partition of Small Group Predecision Performance into Informational and Social Components.

Abstract New theoretical and methodological techniques for partitioning and identifying the sources of performance differences between groups and individuals in hypothesis and act generation tasks are presented in two experiments. Experiment 1 presents a two-component model which separates group performance into informational and social components. The model proposes that the pooling of information in an interacting group (the information component) is mediated by the social factors (e.g., level of arousal, cohesiveness) which are present in a given situation (the social component). Interacting groups were found to be inferior to nominal groups in a hypothesis generation task. Thus, in Experiment 1, the social component was found to have a negative effect and the information component was found to be positive. Experiment 2 further partitions the social component into a social information component which accounts for the additional information which becomes available as a result of group interaction and a social, noninformational component which consists of purely social factors. The social information component estimates the synergistic effect of group interaction on information retrieval. The social informational component was estimated by including a group of subjects who exchanged ideas (information) via computers but had no social interaction. The “information exchange” group was found to be somewhat superior to a nominal group in an act generation task, and both of these groups were superior to an interacting group. Experiment 2 illustrates that even when the social, noninformational component has a negative effect on the informational component, the social information component may have a positive effect.

Inferring user expertise for adaptive interfaces

A technique based on two heuristic rules for inferring expertise is demonstrated by inferring user expertise in word-processing tasks. The heuristic rules were translated into practice by examining command frequencies and requests for on-line help from the 12 participants in the study who were engaged in personal word-processing tasks. These variables were found to be related to word-processing expertise. A scoring rule derived from these variables ranged from 71% to 87% correct in predicting the expertise of the user. The application of this technique to adaptive interfaces that incorporate estimates of user expertise is discussed.

Predicting Individual Differences in Generation of Hypotheses

An experiment was performed to determine the extent to which individual differences in hypothesis generation could be predicted. Measures of achievement, general mental ability, and information were included with the Alternate Uses test as predictors of hypothesis-generation performance. The Alternate Uses test was the best predictor of performance. Several variants of the Alternate Uses test were also employed to isolate the components of hypothesis generation. Two components were involved: retrieval of implicit dimensions of the objects and retrieval of uses when the dimensions are explicitly provided. The latter component was much more important. The results suggest that good hypothesis generators have skills that enable them effectively to retrieve information stored in memory.

SOME OBSERVATIONS AND MEASUREMENTS OF THE PANUM PHENOMENON.

The angular separation between the “binocular” and the “monocular” line of Panum’s limiting case was systematically varied under conditions in which the changes in seen relative depth could be quantified. Stereoscopic, equidistant, and anomalous depth localizations were seen. A criterion of variability of depth localization was utilized to differentiate the mechanism operative in determining the seen depth. When stereopsis is clearly present, depth in Panum’s limiting case is predictable and reveals a one-to-one relationship with the angle of lateral separation of the stereoscopic stimuli, i.e., the odd line cooperates in free binocular vision with both of the paired lines to give “true” stereoscopic depth. The range of angular separation over which Panum’s limiting case will give rise to stereoscopic depth is increased by free eye movements well beyond the usually reported limits of Panum’s retinal areas.

Alternative Response and Scoring Methods for Multiple Choice Items: An Empirical Study of Probabilistic and Ordinal Response Modes.

Binary, probability, and ordinal scoring proce dures for multiple-choice items were examined. In a situation where true scores were experimentally controlled by the manipulation of partial informa tion, it was found that both the probability and or dinal scoring systems were more reliable than the binary scoring method. A second experiment using vocabulary items and standard reliability estimation procedures also showed higher reliability for the two partial information scoring methods relative to binary scoring.

Multiple-stage probabilistic information processing

Abstract Human performance on hierarchical inferences may be suboptimum. As a means of improving this performance, it is suggested that a hierarchical inference task be decomposed into a series of simpler intuitive inferences linking each stage in the hierarchy and that assessments of these inferences be formally combined using the optimum hierarchical model. The results of laboratory experiments with this procedure show that it can improve performance over a completely intuitive approach. A case study in which the procedure was applied to a real world inference problem is also described and some practical difficulties are identified.