People see more of their biases in algorithms.

Algorithmic bias occurs when algorithms incorporate biases in the human decisions on which they are trained. We find that people see more of their biases (e.g., age, gender, race) in the decisions of algorithms than in their own decisions. Research participants saw more bias in the decisions of algorithms trained on their decisions than in their own decisions, even when those decisions were the same and participants were incentivized to reveal their true beliefs. By contrast, participants saw as much bias in the decisions of algorithms trained on their decisions as in the decisions of other participants and algorithms trained on the decisions of other participants. Cognitive psychological processes and motivated reasoning help explain why people see more of their biases in algorithms. Research participants most susceptible to bias blind spot were most likely to see more bias in algorithms than self. Participants were also more likely to perceive algorithms than themselves to have been influenced by irrelevant biasing attributes (e.g., race) but not by relevant attributes (e.g., user reviews). Because participants saw more of their biases in algorithms than themselves, they were more likely to make debiasing corrections to decisions attributed to an algorithm than to themselves. Our findings show that bias is more readily perceived in algorithms than in self and suggest how to use algorithms to reveal and correct biased human decisions.

A Feasible Estimation of a "Corrected" EQ-5D Social Tariff.

OBJECTIVES: To demonstrate the feasibility of estimating a social tariff free of utility curvature and probability weighting biases and to test transferability between riskless and risky contexts. METHODS: Valuations for a selection of EQ-5D-3L health states were collected from a large and representative sample (N = 1676) of the Spanish general population through computer-assisted personal interviewing. Two elicitation methods were used: the traditional time trade-off (TTO) and a novel risky-TTO procedure. Both methods are equivalent for better than death states, which allowed us to test transferability of utilities across riskless and risky contexts. Corrective procedures applied are based on rank-dependent utility theory, identifying parameter estimates at the individual level. All corrections are health-state specific, which is a unique feature of our corrective approach. RESULTS: Two corrected value sets for the EQ-5D-3L system are estimated, highlighting the feasibility of developing national tariffs under nonexpected utility theories, such as rank-dependent utility. Furthermore, transferability was not supported for at least half of the health states valued by our sample. CONCLUSIONS: It is feasible to estimate a social tariff by using interviewing techniques, sample sizes, and sample representativeness equivalent to prior studies designed to generate national value sets for the EQ-5D. Utilities obtained in distinct contexts may not be interchangeable. Our findings caution against routinely taking transferability of utility for granted.

Cognitive Bias in the Patient Encounter: Part II. Debiasing using an adaptive toolbox.

Cognitive bias may lead to medical error, and awareness of cognitive pitfalls is a potential first step to addressing the negative consequences of cognitive bias (see Part 1). For decision-making processes that occur under uncertainty, which encompass most physician decisions, a so-called "adaptive toolbox" is beneficial for good decisions. The adaptive toolbox is inclusive of broad strategies like cultural humility, emotional intelligence, and self-care that help combat implicit bias, negative consequences of affective bias, and optimize cognition. Additionally, the adaptive toolbox includes situational-specific tools such as heuristics, narratives, cognitive forcing functions, and fast and frugal trees. Such tools may mitigate against errors due to cultural, affective, and cognitive bias. Part 2 of this two-part series covers metacognition and cognitive bias in relation to broad and specific strategies aimed at better decision-making.

Efficient adversarial debiasing with concept activation vector - Medical image case-studies.

BACKGROUND: A major hurdle for the real time deployment of the AI models is ensuring trustworthiness of these models for the unseen population. More often than not, these complex models are black boxes in which promising results are generated. However, when scrutinized, these models begin to reveal implicit biases during the decision making, particularly for the minority subgroups. METHOD: We develop an efficient adversarial de-biasing approach with partial learning by incorporating the existing concept activation vectors (CAV) methodology, to reduce racial disparities while preserving the performance of the targeted task. CAV is originally a model interpretability technique which we adopted to identify convolution layers responsible for learning race and only fine-tune up to that layer instead of fine-tuning the complete network, limiting the drop in performance RESULTS:: The methodology has been evaluated on two independent medical image case-studies - chest X-ray and mammograms, and we also performed external validation on a different racial population. On the external datasets for the chest X-ray use-case, debiased models (averaged AUC 0.87 ) outperformed the baseline convolution models (averaged AUC 0.57 ) as well as the models trained with the popular fine-tuning strategy (averaged AUC 0.81). Moreover, the mammogram models is debiased using a single dataset (white, black and Asian) and improved the performance on an external datasets (averaged AUC 0.8 to 0.86 ) with completely different population (primarily Hispanic patients). CONCLUSION: In this study, we demonstrated that the adversarial models trained only with internal data performed equally or often outperformed the standard fine-tuning strategy with data from an external setting. The adversarial training approach described can be applied regardless of predictor's model architecture, as long as the convolution model is trained using a gradient-based method. We release the training code with academic open-source license - https://github.com/ramon349/JBI2023_TCAV_debiasing.

How Can Debiasing Research Aid Efforts to Reduce Discrimination?

PUBLIC ABSTRACT: Scientists studying intergroup biases are often concerned with lessening discrimination (unequal treatment of one social group versus another), but many interventions for reducing such biased behavior have weak or limited evidence. In this review article, we argue one productive avenue for reducing discrimination comes from adapting interventions in a separate field-judgment and decision-making-that has historically studied "debiasing": the ways people can lessen the unwanted influence of irrelevant information on decision-making. While debiasing research shares several commonalities with research on reducing intergroup discrimination, many debiasing interventions have relied on methods that differ from those deployed in the intergroup bias literature. We review several instances where debiasing principles have been successfully applied toward reducing intergroup biases in behavior and introduce other debiasing techniques that may be well-suited for future efforts in lessening discrimination.

Easy-fix attentional focus manipulation boosts the intuitive and deliberate use of base-rate information.

In addressing human reasoning biases, "easy-fix" attentional focus interventions have shown that we can prompt reasoners to align responses with logico-mathematical principles. The current study aimed to test the impact of such interventions on both intuitive and deliberate responses on base-rate items. Using a two-response paradigm, participants provided initial intuitive responses under time constraints and cognitive load, followed by deliberate responses. During the intervention, we used attentional focus manipulations with base-rate items that aimed to redirect participants' attention toward the "logical" base-rate cue (i.e., the logical intervention) or toward the "heuristic" descriptive cue (i.e., the heuristic intervention). The results indicate that the logical intervention led to improved alignment with logico-mathematical principles in both intuitive and deliberate responses, albeit with a modest effect size. Conversely, the heuristic intervention had no discernible impact on accuracy. This indicates that our attentional focus manipulation is more effective at getting reasoners to respect rather than to override base-rates.

A strategy to improve expert technology forecasts.

Forecasts of the future cost and performance of technologies are often used to support decision-making. However, retrospective reviews find that many forecasts made by experts are not very accurate and are often seriously overconfident, with realized values too frequently falling outside of forecasted ranges. Here, we outline a hybrid approach to expert elicitation that we believe might improve forecasts of future technologies. The proposed approach iteratively combines the judgments of technical domain experts with those of experts who are knowledgeable about broader issues of technology adoption and public policy. We motivate the approach with results from a pilot study designed to help forecasters think systematically about factors beyond the technology itself that may shape its future, such as policy, economic, and social factors. Forecasters who received briefings on these topics provided wider forecast intervals than those receiving no assistance.

Removing Biases in Communication of Severity Assessments of Intimate Partner Violence: Model Development and Evaluation.

BACKGROUND: To support a victim of violence and establish the correct penalty for the perpetrator, it is crucial to correctly evaluate and communicate the severity of the violence. Recent data have shown these communications to be biased. However, computational language models provide opportunities for automated evaluation of the severity to mitigate the biases. OBJECTIVE: We investigated whether these biases can be removed with computational algorithms trained to measure the severity of violence described. METHODS: In phase 1 (P1), participants (N=71) were instructed to write some text and type 5 keywords describing an event where they experienced physical violence and 1 keyword describing an event where they experienced psychological violence in an intimate partner relationship. They were also asked to rate the severity. In phase 2 (P2), another set of participants (N=40) read the texts and rated them for severity of violence on the same scale as in P1. We also quantified the text data to word embeddings. Machine learning was used to train a model to predict the severity ratings. RESULTS: For physical violence, there was a greater accuracy bias for humans (r2=0.22) compared to the computational model (r2=0.31; t38=-2.37, P=.023). For psychological violence, the accuracy bias was greater for humans (r2=0.058) than for the computational model (r2=0.35; t38=-14.58, P<.001). Participants in P1 experienced psychological violence as more severe (mean 6.46, SD 1.69) than participants rating the same events in P2 (mean 5.84, SD 2.80; t86=-2.22, P=.029<.05), whereas no calibration bias was found for the computational model (t134=1.30, P=.195). However, no calibration bias was found for physical violence for humans between P1 (mean 6.59, SD 1.81) and P2 (mean 7.54, SD 2.62; t86=1.32, P=.19) or for the computational model (t134=0.62, P=.534). There was no difference in the severity ratings between psychological and physical violence in P1. However, the bias (ie, the ratings in P2 minus the ratings in P1) was highly negatively correlated with the severity ratings in P1 (r2=0.29) and in P2 (r2=0.37), whereas the ratings in P1 and P2 were somewhat less correlated (r2=0.11) using the psychological and physical data combined. CONCLUSIONS: The results show that the computational model mitigates accuracy bias and removes calibration biases. These results suggest that computational models can be used for debiasing the severity evaluations of violence. These findings may have application in a legal context, prioritizing resources in society and how violent events are presented in the media.

How to Minimize the Impact of Experts' Non-rational Beliefs on Their Judgments on Autism.

The non-autistic majority often judges people on the autism spectrum through the prism of numerous stereotypes, prejudices, cognitive biases, or, generally speaking, non-rational beliefs. This causes problems in autistic people's everyday lives, as they often feel stigmatized, marginalized, and they internalize deficit-laden narratives about themselves. Unfortunately, experts, including health or law professionals, are not entirely immune to these non-rational beliefs, which affect their decision-making processes. This primarily happens when a mix of background knowledge, overconfidence, and haste co-occur. The resulting decisions may impact autistic people, e.g., by determining eligibility for the state's therapeutical and financial support. This paper shows how simplified reasoning and inference may influence experts' (medical examiners or court expert witnesses) decision-making processes concerning autistic people. It also proposes particular clues and strategies that could help experts cope with this risk and avoid making biased decisions.

How to activate intuitive and reflective thinking in behavior research? A comprehensive examination of experimental techniques.

Experiments comparing intuitive and reflective decisions provide insights into the cognitive foundations of human behavior. However, the relative strengths and weaknesses of the frequently used experimental techniques for activating intuition and reflection remain unknown. In a large-scale preregistered online experiment (N = 3667), we compared the effects of eight reflection, six intuition, and two within-subjects manipulations on actual and self-reported measures of cognitive performance. Compared to the overall control, the long debiasing training was the most effective technique for increasing actual reflection scores, and the emotion induction was the most effective technique for increasing actual intuition scores. In contrast, the reason and the intuition recall, the reason induction, and the brief time delay conditions failed to achieve the intended effects. We recommend using the debiasing training, the decision justification, or the monetary incentives technique to activate reflection, and the emotion induction, the cognitive load, or the time pressure technique to activate intuition.

Analysis and Accuracy Improvement of UWB-TDoA-Based Indoor Positioning System.

Positioning systems are used in a wide range of applications which require determining the position of an object in space, such as locating and tracking assets, people and goods; assisting navigation systems; and mapping. Indoor Positioning Systems (IPSs) are used where satellite and other outdoor positioning technologies lack precision or fail. Ultra-WideBand (UWB) technology is especially suitable for an IPS, as it operates under high data transfer rates over short distances and at low power densities, although signals tend to be disrupted by various objects. This paper presents a comprehensive study of the precision, failure, and accuracy of 2D IPSs based on UWB technology and a pseudo-range multilateration algorithm using Time Difference of Arrival (TDoA) signals. As a case study, the positioning of a 4×4m2 area, four anchors (transceivers), and one tag (receiver) are considered using bitcraze's Loco Positioning System. A Cramér-Rao Lower Bound analysis identifies the convex hull of the anchors as the region with highest precision, taking into account the anisotropic radiation pattern of the anchors' antennas as opposed to ideal signal distributions, while bifurcation envelopes containing the anchors are defined to bound the regions in which the IPS is predicted to fail. This allows the formulation of a so-called flyable area, defined as the intersection between the convex hull and the region outside the bifurcation envelopes. Finally, the static bias is measured after applying a built-in Extended Kalman Filter (EKF) and mapped using a Radial Basis Function Network (RBFN). A debiasing filter is then developed to improve the accuracy. Findings and developments are experimentally validated, with the IPS observed to fail near the anchors, precision around ±3cm, and accuracy improved by about 15cm for static and 5cm for dynamic measurements, on average.

Sparse PLS-Based Method for Overlapping Metabolite Set Enrichment Analysis.

Metabolite set enrichment analysis (MSEA) has gained increasing research interest for identification of perturbed metabolic pathways in metabolomics. The method incorporates predefined metabolic pathways information in the analysis where metabolite sets are typically assumed to be mutually exclusive to each other. However, metabolic pathways are known to contain common metabolites and intermediates. This situation, along with limitations in metabolite detection or coverage leads to overlapping, incomplete metabolite sets in pathway analysis. For overlapping metabolite sets, MSEA tends to result in high false positives due to improper weights allocated to the overlapping metabolites. Here, we proposed an extended partial least squares (PLS) model with a new sparse scheme for overlapping metabolite set enrichment analysis, named overlapping group PLS (ogPLS) analysis. The weight vector of the ogPLS model was decomposed into pathway-specific subvectors, and then a group lasso penalty was imposed on these subvectors to achieve a proper weight allocation for the overlapping metabolites. Two strategies were adopted in the proposed ogPLS model to identify the perturbed metabolic pathways. The first strategy involves debiasing regularization, which was used to reduce inequalities amongst the predefined metabolic pathways. The second strategy is stable selection, which was used to rank pathways while avoiding the nuisance problems of model parameter optimization. Both simulated and real-world metabolomic datasets were used to evaluate the proposed method and compare with two other MSEA methods including Global-test and the multiblock PLS (MB-PLS)-based pathway importance in projection (PIP) methods. Using a simulated dataset with known perturbed pathways, the average true discovery rate for the ogPLS method was found to be higher than the Global-test and the MB-PLS-based PIP methods. Analysis with a real-world metabolomics dataset also indicated that the developed method was less prone to select pathways with highly overlapped detected metabolite sets. Compared with the two other methods, the proposed method features higher accuracy, lower false-positive rate, and is more robust when applied to overlapping metabolite set analysis. The developed ogPLS method may serve as an alternative MSEA method to facilitate biological interpretation of metabolomics data for overlapping metabolite sets.

The efficacy of mindful practice in improving diagnosis in healthcare: a systematic review and evidence synthesis.

Despite a variety of definitions of mindfulness, over the past 20 years there have been increasing claims that mindful practice is helpful in improving the accuracy of clinical diagnosis. We performed a systematic review and evidence synthesis in order to: determine the nature and definitions of mindful practice and associated terms; evaluate the quality of evidence for the benefits of mindful practice; and conclude whether mindful practice may reduce diagnostic error. We screened 14397 refereed reports from the five common literature databases, to include 33 reports related to the use of mindful practice in clinical diagnosis. Our evidence synthesis contained no randomised controlled trials (level I evidence) of mindful practice, the majority of supporting evidence (26 reports or 79%) comprised conceptual commentary or opinion (level IV evidence). However, 2 supporting reports constituted controlled studies without randomisation (level IIa), 1 report was quasi-experimental (level IIb), and 4 reports were comparative studies (level III). Thus, we may tentatively conclude that mindful practice appears promising as a method of improving diagnostic accuracy, but that further definitive studies of efficacy are required. We identified a taxonomy of 71 terms related to mindful practice, 7 of which were deemed core terms due to being each cited 5 times or more. The 7 core terms appear to be sufficient to describe the findings at higher levels of evidence in our evidence synthesis, suggesting that future definitive studies of mindful practice should focus on these common core terms in order to promote more generalisable findings.

Debiasing Training Improves Decision Making in the Field.

The primary objection to debiasing-training interventions is a lack of evidence that they improve decision making in field settings, where reminders of bias are absent. We gave graduate students in three professional programs (N = 290) a one-shot training intervention that reduces confirmation bias in laboratory experiments. Natural variance in the training schedule assigned participants to receive training before or after solving an unannounced business case modeled on the decision to launch the Space Shuttle Challenger. We used case solutions to surreptitiously measure participants' susceptibility to confirmation bias. Trained participants were 19% less likely to choose the inferior hypothesis-confirming solution than untrained participants. Analysis of case write-ups suggests that a reduction in confirmatory hypothesis testing accounts for their improved decision making in the case. The results provide promising evidence that debiasing-training effects transfer to field settings and can improve decision making in professional and private life.

Finding the intended use population for a new treatment.

Genomic tools are demonstrating that many human diseases are molecularly heterogeneous and likely to respond differently to molecularly targeted therapeutics. For many widely used treatments, the number of patients needed to treat (NNT) for each patient who benefits is large indicating that many patients are being exposed to the risks of serious adverse effects although they do not benefit from the drug. Consequently, more accurately determining the intended use population for new therapeutics is of increased importance. In this paper, we describe a new paradigm for identifying and internally validating an estimate of the intended use population in randomized phase III clinical trials. The approach preserves the type I error of the trial and approaches determination of the intended use population as a classification problem, not a multiple hypothesis testing problems.

Graphical Depiction of Statistical Information Improves Gambling-Related Judgments.

The domain of gambling is rife with both diagnostic and non-diagnostic information. Previous studies examining scratch card gambling have demonstrated that people are often biased by intuitively appealing, yet non-diagnostic information (i.e., unclaimed prize information). The current study investigated how varying the presentation format of a diagnostic piece of information (i.e., payback percentage) could influence participants' use of this information when in conflict with unclaimed prize information. We hypothesized that when payback percentage information was presented in a graphical, as opposed to a numerical format, participants would be better at ignoring unclaimed prize information and correspondingly have their preferences become congruent with the true value of the presented scratch cards. In Experiment 1 (N = 201), with payback percentage presented in a numerical format, participants displayed a non-optimal preference for cards with greater numbers of unclaimed prizes and lower payback percentages. This preference was reversed in Experiment 2 (N = 201) when payback percentage was presented in a graphical format. In conclusion, the results of the current study demonstrate how judgments in a scratch card gambling domain can be improved by simply changing the presentation format of a single piece of information.

DISTRIBUTED TESTING AND ESTIMATION UNDER SPARSE HIGH DIMENSIONAL MODELS.

This paper studies hypothesis testing and parameter estimation in the context of the divide-and-conquer algorithm. In a unified likelihood based framework, we propose new test statistics and point estimators obtained by aggregating various statistics from k subsamples of size n/k, where n is the sample size. In both low dimensional and sparse high dimensional settings, we address the important question of how large k can be, as n grows large, such that the loss of efficiency due to the divide-and-conquer algorithm is negligible. In other words, the resulting estimators have the same inferential efficiencies and estimation rates as an oracle with access to the full sample. Thorough numerical results are provided to back up the theory.

On cognitive debiasing and the judgment of suicide-attackers.

Suicide-attacks are possibly increasing in frequency all over the world. To date, these attacks are not considered as a manifestation of a particular mental illness. However, the process of radicalization of suicide-attackers has to interest the field of mental health. One plausible explanation for the radicalization of individuals is the use of biased cognitive schemes by the indoctrinator. Among these cognitive schemes could figure the causal attribution bias in which the subject cannot distinguish in front of two factors that operate simultaneously, the share of each factor in achieving a certain goal. Another cognitive bias would be the confirmation bias during which the subject would tend to adhere to ideas from his/her own thinking or the thinking of subjects who share some cultural values with him/her and refute any other ideas. Finally, the bias of polarization or splitting could also be incriminated. Through this bias, the subject would either be proud of being a member of a cultural group or ashamed when he/she feels that this group is being attacked and that he/she is unable to rescue it. Approaches to increase the awareness of individuals to the adverse effects of these biased cognitive schemes may theoretically reduce the risk of committing suicide-attacks. However, despite numerous attempts of "deradicalization" involving technological means of communication as well as social "reintegration" centers, all approaches aiming at raising awareness of cognitive biases need to be studied in a scientific manner before they become widespread.

Manipulating Google's Knowledge Graph Box to Counter Biased Information Processing During an Online Search on Vaccination: Application of a Technological Debiasing Strategy.

BACKGROUND: One of people's major motives for going online is the search for health-related information. Most consumers start their search with a general search engine but are unaware of the fact that its sorting and ranking criteria do not mirror information quality. This misconception can lead to distorted search outcomes, especially when the information processing is characterized by heuristic principles and resulting cognitive biases instead of a systematic elaboration. As vaccination opponents are vocal on the Web, the chance of encountering their non‒evidence-based views on immunization is high. Therefore, biased information processing in this context can cause subsequent impaired judgment and decision making. A technological debiasing strategy could counter this by changing people's search environment. OBJECTIVE: This study aims at testing a technological debiasing strategy to reduce the negative effects of biased information processing when using a general search engine on people's vaccination-related knowledge and attitudes. This strategy is to manipulate the content of Google's knowledge graph box, which is integrated in the search interface and provides basic information about the search topic. METHODS: A full 3x2 factorial, posttest-only design was employed with availability of basic factual information (comprehensible vs hardly comprehensible vs not present) as the first factor and a warning message as the second factor of experimental manipulation. Outcome variables were the evaluation of the knowledge graph box, vaccination-related knowledge, as well as beliefs and attitudes toward vaccination, as represented by three latent variables emerged from an exploratory factor analysis. RESULTS: Two-way analysis of variance revealed a significant main effect of availability of basic information in the knowledge graph box on participants' vaccination knowledge scores (F2,273=4.86, P=.01), skepticism/fear of vaccination side effects (F2,273=3.5, P=.03), and perceived information quality (F2,273=3.73, P=.02). More specifically, respondents receiving comprehensible information appeared to be more knowledgeable, less skeptical of vaccination, and more critical of information quality compared to participants exposed to hardly comprehensible information. Although, there was no significant interaction effect between the availability of information and the presence of the warning, there was a dominant pattern in which the presence of the warning appeared to have a positive influence on the group receiving comprehensible information while the opposite was true for the groups exposed to hardly comprehensible information and no information at all. Participants evaluated the knowledge graph box as moderately to highly useful, with no significant differences among the experimental groups. CONCLUSION: Overall, the results suggest that comprehensible information in the knowledge graph box positively affects participants' vaccination-related knowledge and attitudes. A small change in the content retrieval procedure currently used by Google could already make a valuable difference in the pursuit of an unbiased online information search. Further research is needed to gain insights into the knowledge graph box's entire potential.

Estimated time of arrival and debiasing the time saving bias.

The time saving bias predicts that the time saved when increasing speed from a high speed is overestimated, and underestimated when increasing speed from a slow speed. In a questionnaire, time saving judgements were investigated when information of estimated time to arrival was provided. In an active driving task, an alternative meter indicating the inverted speed was used to debias judgements. The simulated task was to first drive a distance at a given speed, and then drive the same distance again at the speed the driver judged was required to gain exactly 3 min in travel time compared with the first drive. A control group performed the same task with a speedometer and saved less than the targeted 3 min when increasing speed from a high speed, and more than 3 min when increasing from a low speed. Participants in the alternative meter condition were closer to the target. The two studies corroborate a time saving bias and show that biased intuitive judgements can be debiased by displaying the inverted speed. Practitioner Summary: Previous studies have shown a cognitive bias in judgements of the time saved by increasing speed. This simulator study aims to improve driver judgements by introducing a speedometer indicating the inverted speed in active driving. The results show that the bias can be reduced by presenting the inverted speed and this finding can be used when designing in-car information systems.