Manual Decision Research Articles

Optimizing Supply-Chain Planning

Semiconductor manufacturing is a very capital-intensive endeavor that can return substantial revenues. The production planning process must deliver a build schedule that makes efficient use of Intel’s capital resources while satisfying as much demand as possible. This schedule should comprehend the flexibility of production resources, the dynamic nature of supply and demand within Intel’s supply chain, as well as the timing of new product releases and production facility improvements. Previous planning processes relied on spreadsheets for heuristic manual decision making with localized data. With the growing complexity of Intel’s products and manufacturing processes, these methods had become inadequate and unsustainable. Upgrading the planning process required better decision algorithms, improved data management, as well as more automated and integrated planning processes. New tools based on Mathematical Programming were implemented in multiple divisions and stages of Intel’s supply chain. The development team worked closely with the users to understand their business and capture their operating logic to create automated decision systems. These tools balance requirements to satisfy demand, achieve inventory targets, and remain within production capacity to reduce costs and satisfy demand across Intel’s supply chain. They have been developed to evolve the planning process while maintaining visibility to the logic and data flow to facilitate continuous improvement. Advances in data management were required to complement decision algorithm improvements. The new tools integrate directly into source data systems while providing planningand optimization-specific functionality, including mechanisms to track parameter changes and supply dynamic reporting capabilities. These advances allow planners to more easily identify data issues and to better understand the planning recommendations from the tools. The robust data management infrastructure enables tighter integration of organizations, increased scalability, and more consistent implementation of solutions across business units. Advances in decision algorithms, data management, and system automation led to improvements in solution quality, data health, and productivity. The new applications allow planners to rapidly perform analyses on multiple business scenarios to produce better solutions and improve collaboration with other organizations. While results reported by the business users over the past four years have proven the stability and value of this decision support technology, there is still work to be done. Plans for extensions and continuous improvement are provided in the last section of this paper.

Screening for Vertebrobasilar Insufficiency in Patients With Neck Pain: Manual Therapy Decision Making in the Presence of Uncertainty

Screening for Vertebrobasilar Insufficiency in Patients With Neck Pain: Manual Therapy Decision Making in the Presence of Uncertainty

Data mining in retail credit

This article presents a real-world application of a data mining approach to credit scoring. It describes the development and the validation of a decision tree, which aims to discriminate between good and bad accounts of Littlewoods Home Shopping customers based on a sample of orders placed between January and November of 2000. This decision tree was constructed for the orders referred to the Authorisations Department. It showed a great improvement in performance compared to the current manual decisions taken for the orders referred to this Department. The implementation of this tree indicates that Authorisation Advisors should apply a set of simple rules in order to optimise their decision making process. The methodology of the decision tree construction is presented in detail. Furthermore the article discusses alternative approaches to credit scoring. Logistic regression is the most widely used technique and it can be used as a benchmarking to assess competing approaches in credit scoring. Using the Receiver Operating Characteristic (ROC) curve as a performance measure of predictive accuracy, the superiority of the decision tree model against the logistic regression model is indicated.

No go on neutrals? An interhemispheric account of semantic category priming

In a divided-visual-field priming paradigm, participants made manual lexical decisions performing either go no-go responses, or yes-no responses, to related prime-target word pairs that were nonassociated, category members (DEER-PONY). Half of each response group received unrelated (ATTIC-PONY) baselines and half received neutral (BLANK-PONY) baselines. In the go no-go response, small but reliable priming was obtained in both target visual fields and there was a reliable slowing of response time in the right visual field/left hemisphere for participants receiving the neutral baseline. In the yes-no response, greater priming was obtained from a neutral baseline, as compared to an unrelated baseline. We show that under appropriate conditions priming for these stimuli occurs in both visual fields. We also report that cross-visual-field advantages emerged as cognitive task load increased across combinations of baseline and response variables. We conclude that task load may be as important a factor in priming asymmetries as SOA or type of semantic relationship.

Designing Displays for Command and Control Supervision: Contextualizing Alerts and “Trust-But-Verity” Automation

Supervision of command and control tasks requires monitoring and maintaining situation awareness of multiple on-going tasks performed by other agents and systems. Automation may aid the user, but realistic automation is often not totally reliable. Sophisticated users may learn to use even unreliable automation to their advantage if 1) they employ a strategy such as “trust-but-verify”, and 2) the user interface supports this strategy by integrating information about the situation with the display of what the automation is doing. The trust-but-verify strategy can be broken into two parts: 1.1) conditional trust and 1.2) qualitative verification. Conditional trust involves knowing when the situation is conducive to trusting the automation. Qualitative verification involves using the automation as a guide or input into manual decision making.

Feedback in DSS as a Source of Control: Experiments with the Timing of Feedback

Well‐designed feedback can improve decision making but to date there has been no comprehensive study of feedback in decision support systems (DSS) that could guide developers in its design. This work examines the opportunities and means to enhance cognitive control in decision making by providing appropriate feedback with DSS. It concentrates on the timing of feedback, which has been shown to affect the use of feedback and also demonstrates the potential advantages of using information technology over and above manual decision environments. Two experiments test the effects of controlling the timing of cognitive feedback on the user's cognitive control. The first tests the effect of timing one source of cognitive feedback, and the second tests the effect of timing two sources of feedback. The findings suggest that the design of feedback should become a routine part of the development of DSS.

Allocation of tool-replacement decision-making responsibility in flexible manufacturing systems

Effective decision support system (DSS) design for the supervisory controller of a flexible manufacturing system (FMS) requires that functions be allocated between the human and computer subsystems in order to establish an effective human-computer cooperative arrangement. To better understand the nature of this arrangement for the tool-replacement decision function, a study was performed that investigated four different levels of decision-making responsibility between human and computer with respect to making tool-replacement decisions. The human interacted with a simulated FMS through an interface consisting of two screens. One screen provided menu information and the interactive dialogue and information support characteristics to the particular cooperative arrangement while the other screen displayed an ongoing graphical representation of the FMS. Twelve subjects performed the task under three computer-interactive conditions: manual decision making, manual decision making with decision support, and computer control with human override. Six of these subjects performed the task under conditions where the priorities of workparts changed at some point during the task. A computer control condition that was governed by a Bayesian decision policy served as the control condition. Subjects were required to minimize economic losses as specified by a loss function while maximizing part throughput. The overall results indicated that, as compared to the other conditions, the manual decision making allocation policy was significantly more effective on both performance objectives under no priority change conditions and on the loss function measure when priority changes occurred. Based on various analyses, it was hypothesized that the more effective performance demonstrated under manual decision making conditions was due to the potential for this arrangement to promote better consideration of global as opposed to local information when making decisions.

Bottleneck starvation indicators for shop floor control (semiconductor manufacturing process)

A graphical tool for inventory and production control, currently in use at a major US semiconductor manufacturer, is presented. The tool supports a bottleneck starvation avoidance policy. Equations for calculating the bottleneck in a clean room manufacturing environment are presented. A new constraint which must be satisfied to ensure starvation avoidance is introduced. A technique for helping a manual decision process conform to this constraint is presented. Many graphs are given to illustrate the concepts. >

Human Operator Decision-Making in Manual Control

To decide whether or not to attempt a given maneuver, an operator must consider the likelihood of his success as well as the costs or values of the possible outcomes. Detecting the future success of a maneuver from the physical circumstances that make it more or less difficult to achieve and from past experience is formally like detecting signals in noise. Two experiments are described, which indicate that the basic decision model and ROC analysis used for signal detection can also be applied to manual control decisions. Results are given comparing the decision performance of operators who execute the maneuvers with the performance of observers who only watch them.