Gayle Hunt

Using PPT to Analyze Suboptimal Human–Automation Performance

ABSTRACT Diagnostic automation aids are designed to improve human performance by increasing accuracy in event detection tasks. However, human–automation performance has frequently fallen short of expectations, particularly when the aid is highly reliable. In those cases, human–automation performance is often suboptimal, in that a human being augmented with a diagnostic aid does more poorly than the automation itself. Previously, there have been only ambiguous explanations for why this occurs, with few suggestions on how to effectively eliminate suboptimal performance. Fortunately, with the advent of a new general theory of task performance, termed Potential Performance Theory (PPT) by D. Trafimow and S. Rice (2008; 2009), one can now determine exactly why suboptimal performance occurs. Results from the present study reveal that inconsistency is the culprit, rather than just poor strategy selection. Furthermore, PPT allows one to determine exactly how much of the performance decrement is because of inconsistency.

Using Potential Performance Theory to Test Five Hypotheses About Meta-Attribution

ABSTRACT Based on I. Kants (1991) distinction between perfect and imperfect duties and the attribution literature pertaining to that distinction, the authors proposed and tested 5 hypotheses about meta-attribution. More specifically, violations of perfect duties have been shown to arouse both more negative affect and stronger correspondent inferences than do violations of imperfect duties (e.g., D. Trafimow, I. K. Bromgard, K. A. Finlay, & T. Ketelaar, 2005). But when it comes to making meta-attributions—that is, guessing the attributions others would make—is the affect differential an advantage or a disadvantage? In addition to the null hypothesis of no effect, the authors proposed and tested additional hypotheses about how negative affect might increase or decrease the effectiveness of peoples meta-attribution strategies and how even if there is no effect on strategy effectiveness, negative affect could increase or decrease the consistencies with which these strategies could be used.

Using PPT to correct for inconsistency in a speeded task.

ABSTRACT Two experiments employed Potential Performance Theory (PPT) to correct for inconsistency in a speeded task. In Experiment 1, a search task was employed whereby participants searched for a target letter among distracters. In Experiment 2, the search task was more complex and naturalistic—participants searched for enemy weapons in aerial photographs. Not surprisingly, the results revealed that longer search times led to greater accuracy. However, this improvement was due only in part to improved search strategies. Much, if not most, of the improvement was due to increased consistency. The authors demonstrate and discuss some of the advantages to be gained by considering a speeded task within the context of a clear and mathematically precise theory (PPT).

The Harder the Task, the More Inconsistent the Performance: A PPT Analysis on Task Difficulty

ABSTRACT Previous research has demonstrated that as task difficulty increases, task performance subsequently decreases. These decrements in task performance as difficulty increases have been attributed to the processes individuals use to complete tasks. Over a series of three experiments, Potential Performance Theory (PPT; Trafimow & Rice, 2008; 2009), was used to test the hypothesis that decreases in task performance are, in part, due to inconsistency rather than only systematic factors. Task difficulty was manipulated in three visual search tasks by increasing set size (Experiment 1), decreasing contrast (Experiment 2), and increasing background distracters (Experiment 3). Findings over the three studies indicated that decreases in observed task performance as task difficulty increases are primarily due to a decrease of consistency rather than systematic factors. The theoretical implications of these findings are discussed.

Generalizing Kant's Distinction between Perfect and Imperfect Duties to Trust in Different Situations

ABSTRACT Based on previous research that violations of perfect duties cause stronger correspondent inferences than violations of imperfect ones, the authors performed four experiments to generalize this effect to trust. In Experiment 1, abstract violations of perfect duties resulted in less trust than violations of imperfect ones for specific trust scenarios. In Experiments 2 and 3, the authors experimented with different levels of abstractness of the duty violations and obtained similar effects. Experiment 4 was concerned with generalizing further—from duty violations in one situation to trust in a different situation. Although mostly consistent with the findings from Experiments 1–3, the data also demonstrated partial generalization for violations of both perfect and imperfect duties.

Superior Visual Search Accuracy after Exposure to Natural Relative to Urban Environments

Recent studies have demonstrated a performance benefit when interacting in natural compared to urban environments. Presently, we are interested in testing whether visual search performance would differ depending on environmental exposure. We exposed participants to images of either natural scenes or urban scenes and asked them to complete a visual conjunction search (identifying the letter O embedded in an array of Q’s). Consistent with existing literature, findings revealed that participants were more accurate at identifying targets when they were exposed to images of nature relative to participants exposed to images of urban environments.

Analyzing Sub-Optimal Human-Automation Performance across Multiple Sessions

Objective: The purpose of the current study was to analyze the interaction between operator strategy and consistency over time. Background: Sub-optimal human-automation performance is a phenomenon whereby combining human operators with diagnostic aids results in performance that is less than desirable. Rice, Trafimow and Hunt (in press) used Potential Performance Theory (PPT) to show that strategies account for little of the variance in sub-optimal performance, and that most of the decrements are due to operators inconsistently using these strategies. The current study looks at these issues across multiple sessions. Method: 20 participants spent four sessions searching for enemy helicopters in aerial images of Baghdad; the task was augmented by a 70% reliable diagnostic aid that provided recommendations during each trial. Results: Consistency accounted for improvement in observed scores early on, despite no gains to strategy, while improvements in strategy accounted for increased observed performance in the later sessions, despite no gains to consistency. Individual data were also analyzed separately, showing various reasons for improvement across time. Conclusion: Both consistency and strategy play important roles in observed performance during human-automation interaction. Application: This study points to a topic that designers and users of automated systems should carefully consider.

Can priming your self lead to punishing others

ABSTRACT Previous research has shown that priming the individual self can increase the accessibility of individual self-cognitions. In turn, we hypothesized that blame for immoral behaviors also would increase, leading to higher assignments of punishment, but that the size of this effect would depend on whether the behaviors violated perfect or imperfect duties. To test these hypotheses, participants assigned yearly bonus penalties to employees who had performed dishonest, disloyal, unfriendly, or uncharitable acts. As expected, the individual prime increased punishments, and this effect was accentuated for violations of perfect duties relative to violations of imperfect duties.

Operator Aid-Switching in a Case of Redundant Automation with Single-Aid Failure

Research exploring factors that affect an operator’s use of automated systems has gained a great deal of attention over the years; however, it is only recently that an operator’s simultaneous use of multiple automated aids has been investigated. The current study employed a target search task to examine how individuals utilize two automated aids when one of the aids fails and begins to decrease in reliability. More specifically, we were interested in investigating whether people would be able to switch from utilizing the failing aid (originally highly reliable) to the other aid (originally less reliable) when the failing aid became less reliable than the other aid. The results show that participants were able to consistently use the more reliable aid, even if this meant switching from one aid to the other. Practical implications are discussed.

Damned if you do and damned if you don’t: Assigning blame to victims regardless of their choice

Abstract The just world literature implies that when someone is a victim of suffering, observers will somehow attribute the suffering to the behavior of the victim. In the current study, participants read a scenario about a person who had either converted or not converted to a new religion. This same person later either experienced no tragedy or was a victim of an unrelated brutal robbery which permanently disabled him. When the target person was victimized, participants were reluctant to attribute blame to the person or to his morality; however, they were quick to assign blame to the victims choice to convert. Interestingly, even when the victim had not converted, participants still assigned blame to the decision to not convert.