Feedback Bias

Feedback bias is studied and defined differently by computer scientists, social scientists, and engineers. My definition is that when feedback is received in order to confirm or refine a decision made at an earlier time, the decision makers place a higher emphasis on feedback received from sources that are known to support their earlier claim more so than feedback that disputes their claim. 

In the book Project Sponsorship by Randall Englund and Alfonso Bucero, feedback bias may be alleviated by the use of a Feedback Action Plan and a Feedback Assessment Tool. A Feedback Action Plan allows decision makers to map out, early in the decision-making process, the sources of data to be sought, the media in which the feedback will be delivered, the timeliness of the reception of that feedback, and an action plan for use of the feedback received. A Feedback Assessment Tool is a list of feedback items that can be graded on a scale of 1 to 10 in order to allow the decision-making team to assess the quality and diversity of feedback items received, and then evaluate the changes in decisions made based on that feedback. 

In addition, bias played a role in the systemic management deficiencies desaibed in the ensuing investigative reports, hearings, and press conferences. These deficiencies contributed to the safety breakdown that caused this event. Specifically, feedback bias and pressure bias were clearly present. 

The Shuttle Challenger disaster represents a real-world example in which an organization, NASA, allowed the bridge between systems engineering and safety that had existed throughout NASA s history, from Mercury to Gemini to Apollo and on to the early development of the shuttle, to collapse. The Safety Breakdown Theory in this case was the result of several biases that contributed to the failure. As described in Chapter 6, The Glismann Effect—in the form of Pressure bias. Feedback bias, and Availability bias—was obviously present ... 

The "feedback loop in the analytical approach is maintained by a quality assurance program (Figure 15.1), whose objective is to control systematic and random sources of error.The underlying assumption of a quality assurance program is that results obtained when an analytical system is in statistical control are free of bias and are characterized by well-defined confidence intervals. When used properly, a quality assurance program identifies the practices necessary to bring a system into statistical control, allows us to determine if the system remains in statistical control, and suggests a course of corrective action when the system has fallen out of statistical control. 

Feedfoi ward control is usually combined with feedback control to eliminate any offset resulting from inaccurate measurements and calculations and unmeasured load components. The feedback controller can either bias or multiply the feedfoi ward calculation. 

Figure 2.9 (a) STM image of AI203/Ni3AI(l 1 l).The unit cell of the dot structure is shown as dashed line, (b) STS spectra taken at the points indicated in the STM image. The feedback loop was opened at a bias voltage of +3.5 and +3.0V (inset of b). 

In the case of TES, the joule heating of the superconducting film produces a negative thermal feedback which increases the thermal stability. The thermal equilibrium takes place when joule heating is balanced by the thermal leak to the substrate. If for some reason in a TES, biased by a voltage V at the centre of the transition, the temperature decreases, an increase of the TES electrical resistance R takes place. Consequently, the bias power V2/R increases, bringing back the TES at the centre of the transition. 

Fig. 12 (a) Current-distance retraction traces recorded with a goid STM tip for f mM 1,9-nonanedithiol in 1,3,5-trimethyibenzene on Au(lll)-(1 x 1), at bias = 0.10 V. The setpoint current before disabling the feedback was chosen at i0 = 0.1 nA. The retraction rate was 4 nm s-1. (b) Same conditions as in (a), except that the preamplifier iimit was chosen at 10 nA. The dotted lines represent characteristic regions of the low, mid, and high conductances... 

In any imaging and spectroscopic mode of the STM, a bias is required between the sample and the tip. In an electrochemical solvent, faradaic current between the tip and sample can interfere with, and sometimes completely obscure, the tunneling current. This undesirable situation makes it very difficult to control the feedback and to maintain a constant tunneling gap between the tip and the sample. For example, in our laboratory, we have found that feedback control is lost on our present microscope if the faradaic current, ip, assumes a value greater than one-half that of the tunneling current, it. Use of partially insulated tips alleviates this condition, but unfortunately, does not completely eliminate the problem (57). 

Lithography With the STM Electrochemical Techniques. The nonuniform current density distribution generated by an STM tip has also been exploited for electrochemical surface modification schemes. These applications are treated in this paper as distinct from true in situ STM imaging because the electrochemical modification of a substrate does not a priori necessitate subsequent imaging with the STM. To date, all electrochemical modification experiments in which the tip has served as the counter electrode, the STM has been operated in a two-electrode mode, with the substrate surface acting as the working electrode. The tip-sample bias is typically adjusted to drive electrochemical reactions at both the sample surface and the STM tip. Because it has as yet been impossible to maintain feedback control of the z-piezo (tip-substrate distance) in the presence of significant faradaic current (vide infra), all electrochemical STM modification experiments to date have been performed in the absence of such feedback control. 

As with all statistical methods, the mean-field estimate will have statistical error due to the finite sample size (X ), and deterministic errors due to the finite grid size (S ) and feedback of error in the coefficients of the SDEs Ui,p). Since error control is an important consideration in transported PDF simulations, we will now consider a simple example to illustrate the tradeoffs that must be made to minimize statistical error and bias. The example that we will use corresponds to (6.198), where the exact solution141 to the SDEs has the form ... 

In addition, the results of such reactions have suggested plausible models for the mechanism of abiotic generation of optical activity, including an autocatalytic feedback mechanism (261). The latter involves random development of chiral crystals from achiral starting material, and solid-state reaction leading to products in which one enantiomer is in excess and thus can bias subsequent further crystallization (262). 

The temperature sensor is located in the microhotplate center (J2x, 10 kO nominal). This polysilicon resistor is biased with a temperature-independent current source (/bias)- The voltage-drop across the polysilicon temperature sensor provides the feedback signal for the temperature controller. 

In line with perceptual distortions and mood congruent hallucinations, patients with affective disorders frequently demonstrate mood congruent biases in information processing. Depressed patients are oversensitive to negative feedback and perceived failure during memory recall tests, being more likely than matched controls to make an error following a previously identified error (Elliott et al., 1996). However, the most reliably reported bias is in memory... 

Some of the material for this book comes from a graduate course, Quality assurance in chemical laboratories, that I have taught for a number of years at the University of New South Wales in Sydney. I am indebted to the many students who have given excellent feedback and hope I have distilled their communal wisdom with appropriate care. I also extend many thanks to my co-teachers, Tareq Saed Al-Deen, Jianfeng Li, and Diako Ebrahimi. Thanks also to my present PhD student Greg O Donnell for his insights into the treatment of bias. 

If a resistance is placed in the feedback loop (Fig. 6.6b), the bias current ib will also create a difference between Ej and E0 by an amount ibRr. Even very inexpensive (< 1) OAs can have bias currents of less than 10 9 A, which means that the value of Rr will have to exceed 1 MO to create a 1-mV error. Amplifiers with bias currents of less than 0.1 pA (10 13 A) are available. Using the same criterion, Rr may then reach 1000 MQ, a value well beyond any resistance commonly encountered in dynamic electroanalytical techniques. Such amplifiers are, however, eminently useful for constructing pH meters and pH stats and measuring potentials in electrophysiology, where very small high-impedance electrodes are often used. 

A significant problem with the style of separator shown as Fig. 1 is that any feed bias to one location will result in a reduced local airflow rate and subsequent reduction in the cleaning efficiency an inherent negative feedback loop exists in the airflow path. 

---