Feedback Delays: How Can Decision Makers Learn Not to Buy a New Car Every Time the Garage Is Empty?

Decision makers in dynamic environments (e.g., stock trading, inventory control, and firefighting) learn poorly in experiments where feedback about the outcomes of their actions is delayed. In searching for ways to mitigate these effects, this paper presents two computational models of learning with feedback delays and contrasts them against human decision-makers' performance. The no-memory model hypothesizes that decision makers always perceive feedback as immediate. The with-memory model hypothesizes that, over time, decision makers are able to develop internal representations of the task that help them to perform with delayed feedback. As borne out by human subjects, both models predict that a display of past history improves learning with delay and that increasing delay increasingly degrades performance. Even though the length of training in this task exceeds that used in many laboratory-based dynamic tasks, neither the two models nor the subjects are able to effectively learn without decision aids when faced with feedback delays. When given an amount of training that more closely approximates that provided in functioning dynamic environments, the with-memory model predicts that human decision makers may learn without decision aids over the long term if feedback delays are simple. These results raise several issues for continued theoretical investigation as well as potential suggestions for training and supporting decision makers in dynamic environments with feedback delays. Copyright 2000 Academic Press.     

Misperceptions of feedback in dynamic decision making

In recent years laboratory experiments have shed significant light on the behavior of economic agents in a variety of microeconomic and decision-theoretic contexts such as auction markets, portfolio choice, and preference elicitation. Despite the success of experimental techniques in the micro domain, there has been relatively little work linking the behavior of decision makers to the dynamics of larger organizations such as corporations, industries, or the macroeconomy. This paper presents a laboratory experiment in which subjects manage a simulated economy. Subjects must invest sufficient capital plant and equipment to satisfy demand. Subjects were given complete and perfect information regarding the structure of the simulated economy, the values of all variables, and the past history of the system. Nevertheless, the overwhelming majority of the subjects generate significant and costly oscillations. A simple decision rule based on the anchoring and adjustment heuristic is shown to simulate the subjects' decisions quite well. Several distinct sources of the subjects' poor performance are identified and termed “misperceptions of feedback.” The decision rule is related to various models of economic fluctuations; implications for experimental investigation of dynamic decision making in aggregate systems are explored.     

Effects of delayed time‐out on problem behavior of preschool children

Time‐out is a common negative punishment procedure in home and school settings. Although prior studies have shown time‐out is effective, more research is needed on its effects when implementation is imperfect. We evaluated delays to time‐out with 4 preschool children who engaged in some combination of aggression, property destruction, and rule breaking. Target behavior decreased for all subjects exposed to delayed time‐out, with 3 of 4 subjects displaying low levels of target behavior even when time‐out was delayed by 90‐120 s. These data suggest delayed time‐out might be effective in situations in which a caregiver or teacher cannot implement time‐out immediately.     

Why Are Judgments Less Consistent in Less Predictable Task Situations?

Subjects made a series of six forecasts for each of four 58-point time-series. The stimulus series were noisy sinusoids of various frequencies. Individual subjects′ forecasts were regressed on to the time period for which they were made. The residual variance in these regressions was greater when the stimulus series were noisier. This effect cannot have arisen from memory overload because simultaneous presentation of all points in the stimulus series meant that there were no memory demands. It cannot have been a side effect of a learning process because presenting each series just once excluded opportunities for learning. It cannot have been caused by deterministic rule switching consequent on feedback about a noisy process because no feedback was presented. Instead it appears that subjects tried to simulate the noise as well as the pattern in the series when making their forecasts. In support of this, similar results were obtained in a second experiment in which subjects were asked to simulate rather than to forecast the series. Two possible reasons for subjects′ simulation of series noise are discussed.     

Some affective consequences of social comparison and reflection processes: the pain and pleasure of being close

A self-evaluation maintenance (SEM) model of social behavior was described. According to the comparison process, when another outperforms the self on a task high in relevance to the self, the closer the other the greater the threat to self-evaluation. According to the reflection process, when another outperforms the self on a task low in relevance to the self, the closer the other the greater the promise of augmentation to self-evaluation. Affect was assumed to reflect threats and promises to self-evaluation. In three studies, subjects were given feedback about own performance and the performance of a close (friend) and distant (stranger) other on tasks that were either low in self-relevance (Study 2) or that varied in self-relevance (Studies 1 and 3). In Study 1 (N = 31), subjects' performance on simple and complex tasks after each feedback trial served as a measure of arousal. Being outperformed by a close other resulted in greater arousal than being outperformed by a distant other. In Study 2 (N = 30), evaluative ratings of words unrelated to task performance served as an indirect measure of affect. Results indicated that when relevance is low, more positive affect is associated with a friend's outperforming the self than either a friend's performing at a level equal to the self or being outperformed by a stranger. In Study 3 (N = 31), subjects received feedback while their facial expressions were monitored. Pleasantness of expression was an interactive function of relevance of task, relative performance, and closeness of comparison other. The results of all three studies were interpreted as being generally consistent with the SEM model.     

When Audience Presence Is Enjoyable: The Influence of Audience Awareness of Prior Success on Performance and Task Interest

Subjects received positive feedback on a practice anagram task and then performed a similar task in front of either an audience that was aware of their prior success or an audience that was unaware of their prior success. Two control conditions were included in which subjects performed alone after receiving positive feedback or performed alone without feedback. Results indicated subjects' task interest and performance increased under conditions in which they performed in front of unaware audiences. In addition, subjects' task interest and performance levels decreased in the presence of audiences that were aware of their prior success. These results point to an important role of audience awareness in setting performance standards and expectations. These findings are interpreted from a resource-investment analysis of task value and performance. Implications for surveillance in public settings and in the workplace (e.g., automated computer monitoring) are discussed.     

时间估计任务中的反馈加工和行为调节：来自ERP的证据

本研究采用时间估计任务, 利用高时间分辨率的ERP技术, 试图从时间进程上窥探行为调节的认知神经机制。该时间估计任务提供了指导行为调节的反馈信息有三类：时间估计的准确性、时间估计的不准确方向和不准确程度。行为结果表明, 被试根据前次时间估计的对错、不准确的方向和程度进行相应的调整。脑电成分发现, FRN对于时间估计的对、错敏感, 对时间估计的不准确的方向和程度不敏感, 显示FRN提供了粗略的、需要做出行为调节的早期预警信号; 相反, P300能够区别时间估计上不准确的方向和程度, 结合行为数据, 显示P300反映的是整合不同的信息来更新行为表征, 从而引导被试做出不同方向和不同程度的调节。这些结果表明, 行为调节的认知神经基础可能由两个阶段组成：早期传递需要做出行为调节的预警信号阶段和晚期整合信息来更新行为表征从而指导行为调节的阶段。     

Complex oscillations in a human motor system

Experiments were performed to investigate the oscillations arising in a human motor system with delayed visual feedback. Eight subjects were instructed to maintain a constant finger position relative to a stationary baseline. The finger displacement was measured using a microdisplacement transducer connected to the index finger, and was displayed on an oscilloscope. Time delays between 40 and 1,500 ms were inserted in the visual feedback loop for 100 s. Results show that, as the time delays increase, irregular rhythms appear with short intermittent periods of regular oscillations. These regular low-frequency oscillations have an amplitude that increases with the time delays and a period that is consistently about 2 to 4 times the time delay. Fast Fourier Transforms (FFT) show a peak between 8 and 12 Hz, corresponding to physiological tremor in half the subjects. No systematic variations in the FFT for the 2 to 15 Hz range were observed as time delay increased. In the 0 to 2 Hz range, the FFT show a consistent increase in power with the time delay. These results indicate that, under the conditions of this experiment, tremor is not affected by time delays in the visuomotor system, and time delays in the visuomotor feedback loop give rise to complex oscillatory behaviors.     

Paired associate learning in reading-disabled children: evidence for a rule-learning deficiency

Two experiments examined whether normal and disabled readers differed in the utilization of rules in a paired associate learning task. Experiment 1 required children to learn symbol-word associates. Children were assigned to one of three conditions: nonrule, consistent rule, or inconsistent rule. When present, the rule was based on semantic opposites. Subjects benefited from having the rule, but disabled readers showed less improvement across four test blocks than both chronological age (CA) controls and reading age (RA) controls, particularly in the inconsistent condition. Experiment 2 required subjects to learn symbol-symbol associations in one of three conditions: nonrule, consistent rule, or inconsistent rule. When present, the rule specified the locations of a subsidiary figure in each symbol according to the pattern top-right, bottom-left. Disabled and normal readers did not differ in the nonrule condition where reliance on visual memory would be an effective strategy. Normal readers were superior to disabled readers in both rule conditions. In addition, disabled readers in the inconsistent rule condition were less able than normal readers to apply the rule in a generalization task where memory demands were reduced. Results supported the hypothesis that disabled readers have greater difficulty than normal readers inducing and/or using rules, particularly when they are inconsistent. It is suggested that difficulties in acquiring or using complex and inconsistent rules may be one important source of problems learning spelling-sound correspondence rules, which in English are complex and inconsistent.     

The Impact of Biased Inquiry Strategies on Performance Judgments

This experiment explores the attributional consequences of asking constraining performance-relevant questions-biased to elicit favorable or unfavorable remarks from the person being evaluated. After hearing the performer deliver a speech, all subjects played the role of questioner in a feedback interview purportedly designed to help the speechmaker improve future performance. Subjects asked biased questions that focused on either the positive or negative aspects of the speech and heard answers that were either congruent or incongruent with the biasing implications of their questions. Results showed that speech performance ratings were influenced by the direction of the performer′s self-evaluative comments, even when these comments were in line with, and thus potentially constrained by, the evaluator′s own behavior. Implications of biased inquiry strategies in organizational contexts where the evaluator is the source of constraint are discussed.     

Conversation skills training for socially isolated children

A social skills training program consisting of instructions, modeling, behavior rehearsal, performance feedback, and programmed generalization set was used to teach conversational behavior to four socially isolated children. Effects of training were assessed in a multiple-baseline analysis across behavior. Subjects were selected on the basis of peer sociometric ratings, a behavioral role-playing task, naturalistic observation, and teacher ratings. All subjects demonstrated noticeable improvement in the target behavior as well as in overall conversational ability. However, these gains were maintained unequally in the children at follow-up probe administrations (4-and 8-weeks). Generalization assessment revealed some transfer of treatment effects to the school setting for all children. Results are discussed in terms of the complexity of social isolation in children and factors which influenced generalizability of the treatment program.     

The Impact of an Assigned Performance Goal on Feedback Seeking Behavior

This study was designed to test the impact of an assigned performance goal on feedback seeking behavior. Subjects brainstormed uses for two common objects. During the first brainstorming trial, all subjects performed without an explicit performance goal. For the second trial, half of the participants were assigned a goal. Feedback seeking behavior was defined as an intermedi- ate count of production taken by a subject as he or she worked on the task. Results showed that (a) subjects in the assigned goal condition were more likely to seek feedback about their performance than subjects who worked without a goal; and (b) among subjects in the goal condition, those who sought feedback were more likely to meet their goal than subjects who did not. These results support Ashford and Cummings's (1983) view that feed- back is sought because it is a resource that can be used to meet goals.     

The Effects of Feedback Equivocality on Escalation of Commitment: An Empirical Investigation of Decision Dilemma Theory1

The effects of feedback equivocality on escalation of commitment were examined in a laboratory study. Subjects had multiple opportunities to allocate money to market a software product. During the initial phase, subjects received feedback that was of either low or high equivocality. Half of the subjects in each equivocality level were given a standard by which to judge the feedback. In the second phase, all subjects received negative feedback. Consistent with Bowen's (1987) decision dilemma theory, subjects who received low equivocality feedback did not escalate allocations, and subjects who received high equivocal feedback escalated allocations. The presence of a standard or goal attenuated escalation. These results are consonant with a dynamic, temporally based account of escalation and are discussed in a synthesis of the literatures in behavior analysis, economics, forecasting, and psychology.     

Wisconsin Card Sorting in adolescents: analysis of performance, response times and heart rate

Forty-nine adolescents performed the Wisconsin Card Sorting Test (WCST). A main PCA component of WCST performance was identified as 'efficiency of reasoning'. This factor was related to feedback processing. From the WCST, a perseveration score can be derived. Perseveration is the continued application of a rule, after it has been disconfirmed. We compared more and less perseverating subjects in relation to stimulus-response (SR) time, feedback inspection time and cardiac acceleration and deceleration. Less perseverating subjects responded faster, and had longer and more adaptive inspection times of error feedback. We examined the switch from rule application to rule search, and the difference between correct and error responses. A transient cardiac deceleration at the initiation of rule search was interpreted as a change in supervisory attention. An error-related deceleration to negative feedback was interpreted as a disturbance of higher control processing. Previous trial feedback influenced current processing time, feedback inspection time, and the cardiac acceleration and deceleration responses.     

Response patterns and cardiovascular effects during response sequence acquisition by humans.

The effects of temporal delays imposed between successive responses and of vitamin C administration were examined on the acquisition of response sequences and on cardiovascular reactivity during sequence acquisition. Thirteen adult subjects (6 female, 7 male), in good health, gave written consent prior to participating in 12 weekly 45-min sessions. Points, exchanged for money after each session, were presented when subjects completed 15-response sequences on a touch-sensitive three-response keypad. A position counter increased from 0 to 14 as subjects emitted correct responses in the sequence. Four novel 15-response sequences were presented each session. No delays were imposed between successive responses during the acquisition of one sequence; delays were imposed immediately following each response during the acquisition of a second sequence, thereby delaying response feedback; delays were imposed following feedback during acquisition of a third sequence, resulting in the removal of the stimulus correlated with sequence position; and, as a control condition, delays were imposed following feedback, but stimuli correlated with sequence position were reinstated prior to the next response during acquisition of a fourth sequence. Subjects were exposed to one of two delay durations (0.2 and 0.5 or 0.5 and 1.0 s) each session, and delay durations alternated every session. During Weeks 5 to 8, subjects received 3 grams of vitamin C per day, whereas during Weeks 1 to 4 and 9 to 12, subjects received placebo under single-blind conditions. All subjects acquired the sequences, as evidenced by decreasing percentages of incorrect responses across trials. When temporal delays were imposed between successive responses during sequence acquisition, acquisition efficiency was enhanced. Examination of response latencies suggested that the status of preceding responses (i.e., correct or incorrect) rather than the status of the position counter influenced subsequent responding. Cardiovascular effects were inversely related to the length of the temporal delay. Neither cardiovascular reactivity or sequence acquisition were related to vitamin C administration.     

Optimizing the Use of Vision in Manual Aiming: The Role of Practice

The purpose of this study was to determine how subjects learn to adjust the characteristics of their manual aiming movements in order to make optimal use of the visual information and reduce movement error. Subjects practised aiming (120 trials) with visual information available for either 400 msec or 600 msec. Following acquisition, they were transferred to conditions in which visual information was available for either more or less time. Over acquisition, subjects appeared to reduce target-aiming error by moving to the target area more quickly in order to make greater use of vision when in the vicinity of the target. With practice, there was also a reduction in the number of modifications in the movement. After transfer, both performance and kinematic data indicated that the time for which visual information was available was a more important predictor of aiming error than the similarity between training and transfer conditions. These findings are not consistent with a strong “specificity of learning” position. They also suggest that, if some sort of general representation or motor programme develops with practice, that representation includes rules or procedures for the utilization of visual feedback to allow for the on-line adjustment of the goal-directed movement.