Artificial Bias Research Articles

A cell marker‐based clustering strategy (cmCluster) for precise cell type identification of scRNA‐seq data

BackgroundThe precise and efficient analysis of single‐cell transcriptome data provides powerful support for studying the diversity of cell functions at the single‐cell level. The most important and challenging steps are cell clustering and recognition of cell populations. While the precision of clustering and annotation are considered separately in most current studies, it is worth attempting to develop an extensive and flexible strategy to balance clustering accuracy and biological explanation comprehensively.MethodsThe cell marker‐based clustering strategy (cmCluster), which is a modified Louvain clustering method, aims to search the optimal clusters through genetic algorithm (GA) and grid search based on the cell type annotation results.ResultsBy applying cmCluster on a set of single‐cell transcriptome data, the results showed that it was beneficial for the recognition of cell populations and explanation of biological function even on the occasion of incomplete cell type information or multiple data resources. In addition, cmCluster also produced clear boundaries and appropriate subtypes with potential marker genes. The relevant code is available in GitHub website (huangyuwei301/cmCluster).ConclusionsWe speculate that cmCluster provides researchers effective screening strategies to improve the accuracy of subsequent biological analysis, reduce artificial bias, and facilitate the comparison and analysis of multiple studies.

Simulation of ultimatum game with artificial intelligence and biases

In this research we have developed experimental designs of the ultimatum game with supervised agents. This agents have unbiased and biased thinking depending on the case. We used Reinforcement Learning and Bucket Brigade to program the artficial agentes. We used simulations and behavior comparison to answer the following questions: Does artificial intelligence reach a perfect subgame equilibrium in the ultimatum game experiment? How would Artificial Intelligence behave in the Ultimatum Game experiment if biased thinking is included in it? This exploratory analysis showed one important result: artificial inteligence by itself doesn´t reach a perfect subgame equilibrium. Whereas, the experimental designs with biased thinking agents quickly converge to an equilibrium. Finally, we demonstrated that the agents with envy bias behaves the same as the ones with altruistic bias.

Artificial Intelligence Bias and Ethics in Retinal Imaging

Our website uses cookies to enhance your experience. By continuing to use our site, or clicking "Continue," you are agreeing to our Cookie Policy | Continue JAMA Ophthalmology HomeNew OnlineCurrent IssueFor Authors Podcast Journals JAMA JAMA Network Open JAMA Cardiology JAMA Dermatology JAMA Health Forum JAMA Internal Medicine JAMA Neurology JAMA Oncology JAMA Ophthalmology JAMA Otolaryngology–Head & Neck Surgery JAMA Pediatrics JAMA Psychiatry JAMA Surgery Archives of Neurology & Psychiatry (1919-1959) JN Learning / CMESubscribeJobsInstitutions / LibrariansReprints & Permissions Terms of Use | Privacy Policy | Accessibility Statement 2023 American Medical Association. All Rights Reserved Search All JAMA JAMA Network Open JAMA Cardiology JAMA Dermatology JAMA Forum Archive JAMA Health Forum JAMA Internal Medicine JAMA Neurology JAMA Oncology JAMA Ophthalmology JAMA Otolaryngology–Head & Neck Surgery JAMA Pediatrics JAMA Psychiatry JAMA Surgery Archives of Neurology & Psychiatry Input Search Term Sign In Individual Sign In Sign inCreate an Account Access through your institution Sign In Purchase Options: Buy this article Rent this article Subscribe to the JAMA Ophthalmology journal

Numerical discreteness errors in multispecies cosmological N-body simulations

ABSTRACT We present a detailed analysis of numerical discreteness errors in two-species, gravity-only, cosmological simulations using the density power spectrum as a diagnostic probe. In a simple set-up where both species are initialized with the same total matter transfer function, biased growth of power forms on small scales when the solver force resolution is finer than the mean interparticle separation. The artificial bias is more severe when individual density and velocity transfer functions are applied. In particular, significant large-scale offsets in power are measured between simulations with conventional offset grid initial conditions when compared against converged high-resolution results where the force resolution scale is matched to the interparticle separation. These offsets persist even when the cosmology is chosen so that the two particle species have the same mass, indicating that the error is sourced from discreteness in the total matter field as opposed to unequal particle mass. We further investigate two mitigation strategies to address discreteness errors: the frozen potential method and softened interspecies short-range forces. The former evolves particles under the approximately ‘frozen’ total matter potential in linear theory at early times, while the latter filters cross-species gravitational interactions on small scales in low-density regions. By modelling closer to the continuum limit, both mitigation strategies demonstrate considerable reductions in large-scale power spectrum offsets.

Automated recognition and analysis of body bending behavior in C. elegans

BackgroundLocomotion behaviors of Caenorhabditis elegans play an important role in drug activity screening, anti-aging research, and toxicological assessment. Previous studies have provided important insights into drug activity screening, anti-aging, and toxicological research by manually counting the number of body bends. However, manual counting is often low-throughput and takes a lot of time and manpower. And it is easy to cause artificial bias and error in counting results.ResultsIn this paper, an algorithm is proposed for automatic counting and analysis of the body bending behavior of nematodes. First of all, the numerical coordinate regression method with convolutional neural network is used to obtain the head and tail coordinates. Next, curvature-based feature point extraction algorithm is used to calculate the feature points of the nematode centerline. Then the maximum distance between the peak point and the straight line between the pharynx and the tail is calculated. The number of body bends is counted according to the change in the maximum distance per frame.ConclusionExperiments are performed to prove the effectiveness of the proposed algorithm. The accuracy of head coordinate prediction is 0.993, and the accuracy of tail coordinate prediction is 0.990. The Pearson correlation coefficient between the results of the automatic count and manual count of the number of body bends is 0.998 and the mean absolute error is 1.931. Different strains of nematodes are selected to analyze differences in body bending behavior, demonstrating a relationship between nematode vitality and lifespan. The code is freely available at https://github.com/hthana/Body-Bend-Count.

Estimating GDP growth using VIIRS night-time light data

ABSTRACT The accuracy of GDP per capita growth rate in countries with low statistical capacity has been questioned, which hinders understanding progress of Sustainable Development Goals (SDGs) accurately. This study aims to estimate GDP per capita growth rate in countries with poor statistical quality. To achieve this goal, the study builds an econometric model with processed Visible Infrared Imaging Radiometer (VIIRS) night-time light data and other economic variables and the study further assesses the accuracy of the model by comparing the estimated GDP growth to high-quality economic data from Organization for Economic Co-operation and Development (OECD) countries between 2014 and 2018. First, due to the abnormal fluctuations in the time series generated from yearly composite night-time light images, we propose a trend analysis method for the monthly night-time light data to synthesize a more stable yearly time series. Then, we construct artificially low-quality GDP simulations and combine VIIRS data and official statistics data to construct regression models to correct the simulated GDP. Finally, to assess the model accuracy, we compare its estimates of GDP per capita growth rates to official data from high-quality OECD countries, looking for any deviations. We have two major findings as follows: first, the VIIRS data, which comprise information on human activity, are a valuable resource for estimating GDP growth; second, the GDP estimation model can reduce the artificial bias of the simulated GDP per capita growth rate by using VIIRS night-time light data and the average accuracy of the model with simulated data can reach 89%.

Human-induced weakening of the Northern Hemisphere tropical circulation.

By accounting for most of the poleward atmospheric heat and moisture transport in the tropics, the Hadley circulation largely affects the latitudinal patterns of precipitation and temperature at low latitudes. To increase our preparednesses for human-induced climate change, it is thus critical to accurately assess the response of the Hadley circulation to anthropogenic emissions1-3. However, at present, there is a large uncertainty in recent Northern Hemisphere Hadley circulation strength changes4. Not only do climate models simulate a weakening of the circulation5, whereas atmospheric reanalyses mostly show an intensification of the circulation4-8, but atmospheric reanalyses were found to have artificial biases in the strength of the circulation5, resulting in unknown impacts of human emissions on recent Hadley circulation changes. Here we constrain the recent changes in the Hadley circulation using sea-level pressure measurements and show that, in agreement with the latest suite of climate models, the circulation has considerably weakened over recent decades. We further show that the weakening of the circulation is attributable to anthropogenic emissions, which increases our confidence in human-induced tropical climate change projections. Given the large climate impacts of the circulation at low latitudes, the recent human-induced weakening of the flow suggests wider consequences for the regional tropical-subtropical climate.

The Impact of Artificial Intelligence Bias on Human Resource Management Functions: Systematic Literature Review and Future Research Directions

Artificial intelligence (AI) has become a valuable tool for facilitating Human Resource Management (HRM) functions. Although, it should be noted that AI has a specific character side away from other technology. Publications covering this knowledge area have grown sharply, however the scholarly covering the impact of AI bias inHRM is scarce. This paper studies this area and goes deeper to explore the future research areas by conducting a systematic literature review for 598 papers from Scopes and Emerald insight databases of which 34 articles were selected after implementing the PRISMA tool and quality evaluation stage. Results generated revealed that biased AI applications are negatively affecting performance management, compensation, staffing and training and development. Apart from that future research domains and questions have been outlined and identified from organizations’ and employees’ perspectives.

Sources of bias in observational studies of covid-19 vaccine effectiveness.

Sources of bias in observational studies of covid-19 vaccine effectiveness.

Radiometric Re-Compensation of Sentinel-1 SAR Data Products for Artificial Biases due to Antenna Pattern Changes

SAR data products for Sentinel-1 have been freely available and delivered operationally on behalf of the European Space Agency since the routine operation of Sentinel-1A in 2014. These products were delivered using the best knowledge at their processing time, in particular with respect to the radiometric calibration. As reprocessing of SAR data products is not foreseen in the nominal processing chain of Sentinel-1, changes of applied processing parameters impact the SAR data quality and can be a disturbing factor for long-term monitoring of radiometric features. In particular, antenna pattern updates produce artificial radiometric steps which are visible in radar backscatter time series, especially in case of monitoring radiometric stable reference targets. This paper introduces a method for correcting changes due to such updates without the need of reprocessing SAR data products. The method was applied to long-lasting time series of data acquisitions which are used to monitor the radiometric performance with reference targets at the DLR calibration site. It has been shown that artificial steps due to antenna pattern updates disappear in backscatter timelines after correct application. Furthermore, the derived absolute radiometric accuracy was improved for the joint observation period of S1A and S1B for almost five years until December 2021.

Fair detection of poisoning attacks in federated learning on non-i.i.d. data.

Reconciling machine learning with individual privacy is one of the main motivations behind federated learning (FL), a decentralized machine learning technique that aggregates partial models trained by clients on their own private data to obtain a global deep learning model. Even if FL provides stronger privacy guarantees to the participating clients than centralized learning collecting the clients' data in a central server, FL is vulnerable to some attacks whereby malicious clients submit bad updates in order to prevent the model from converging or, more subtly, to introduce artificial bias in the classification (poisoning). Poisoning detection techniques compute statistics on the updates to identify malicious clients. A downside of anti-poisoning techniques is that they might lead to discriminate minority groups whose data are significantly and legitimately different from those of the majority of clients. This would not only be unfair, but would yield poorer models that would fail to capture the knowledge in the training data, especially when data are not independent and identically distributed (non-i.i.d.). In this work, we strive to strike a balance between fighting poisoning and accommodating diversity to help learning fairer and less discriminatory federated learning models. In this way, we forestall the exclusion of diverse clients while still ensuring detection of poisoning attacks. Empirical work on three data sets shows that employing our approach to tell legitimate from malicious updates produces models that are more accurate than those obtained with state-of-the-art poisoning detection techniques. Additionally, we explore the impact of our proposal on the performance of models on non-i.i.d local training data.

Analysis of the Artificial Intelligence Bias Risks in the Implementation of Public Legal Relations: Foreign Countries' Experience and Russian Administrative Law Perspectives

Analysis of the Artificial Intelligence Bias Risks in the Implementation of Public Legal Relations: Foreign Countries' Experience and Russian Administrative Law Perspectives

Clinically Consistent Prostate Cancer Outcome Prediction Models with Machine Learning

<h3>Purpose/Objective(s)</h3> State-of-the-art prediction modeling relies on machine learning (ML). To provide accurate out-of-sample predictions, such as when the number of available covariates is large, these models trade-off variance by adding bias. However, these artificial biases frequently create anomalies that are clinically inconsistent, hindering their use in clinical practice. We illustrate such biases in outcome prediction ML models using the National Cancer Data Base (NCDB) and propose a solution that leverages clinical expertise. We hypothesize that integrating clinical expertise into outcome prediction ML models will correct significant problems present in ML modeling without compromising the model accuracy, enabling models to be more clinically intuitive. <h3>Materials/Methods</h3> We use data from 92,990 prostate cancer patients in the NCDB (death rate: 28%) to build a gradient boosted machine (GBM) model to predict 10-year survival post-treatment. We then investigate how the model output depends on each covariate and identify inconsistencies with clinical knowledge. To correct for these biases, we collect clinical intuition from five oncologists through a survey that asks how increasing a covariate impacts (increase/decrease/variable effect) a patient's survival. We then encode the survey results as clinically informed monotonic constraints into a LASSO model and compare the AUC with and without constraints. <h3>Results</h3> We find that while the GBM model has a relatively high out-of-sample AUC of 0.79 (compared to similar studies, with AUC between 0.65-0.83, although patient cohort, methodology, and outcomes may differ), the way some covariates are used is not clinically consistent. For example, keeping other covariates equal, the probability of survival should consistently be higher for patients with lower prostate specific antigen (PSA) levels compared to patients at higher levels. But the model predicts (with high confidence) a sporadic, non-monotonic pattern. This is because the GBM model biases its predictions for patients that comprise a small fraction of the patient cohort (i.e., patients with very low PSA values) to avoid high variance, but the artificially injected bias is towards lower survival which is not clinically consistent. Covariates named by oncologists to have a monotonic effect on the probability of survival include PSA and age. We find that by adding clinically informed constraints to a LASSO model, the problem of clinical inconsistency is addressed without compromising model performance (AUC is 0.77 with and without constraints). <h3>Conclusion</h3> We illustrate that even in a large-scale regime, state-of-the-art clinical risk prediction models can suffer from biases. Our work shows that innovative integration of clinical intuition can make models clinically consistent and verifies that this method is promising research to resolve critical challenges in ML-based decision support.

Tools adapted to Ethical Analysis of Data Bias

Data Bias, a bias embedded in data during collection, storing, and use, and in the apps used by a human, is an emerging issue of data privacy exemplified by Artificial Intelligence bias (AI bias). This issue is becoming gradually an added vulnerability to data ethics, an added threat to data security, and an added burden to data protection. It has an effect to induce a reduction in data protection expenditure, and is crucial to the success of any creative endeavours in the data-driven technology-intensive era, including Engineering, exemplified by AI bias, and AI bias is bias created when biased data creeps in during design, development, and training of AI algorithms. AI indeed culminates in a phenomenon in which the populace jumps, yet a sober minority steers away from because of the pervasive cyber-threats that AI bias raises. At issue is not data bias per se, nor the multi-dimensional issues induced by human bias, which are usually complex and slippery, but a need for a method to enable a holistic view covering the technical, financial, legal, social, ethical, and ecological aspects of a given problem, action, policy, or decision. Recommendable is a method composed of the Ethical Matrix Algorithm and Hexa-dimension Metric Algorithm (Lee, forthcoming) based respectively on the Ethical Matrix and Hexa-dimension Metric (Lee, 2021).

Validating Automatic Concept-Based Explanations for AI-Based Digital Histopathology.

Digital histopathology poses several challenges such as label noise, class imbalance, limited availability of labelled data, and several latent biases to deep learning, negatively influencing transparency, reproducibility, and classification performance. In particular, biases are well known to cause poor generalization. Proposed tools from explainable artificial intelligence (XAI), bias detection, and bias discovery suffer from technical challenges, complexity, unintuitive usage, inherent biases, or a semantic gap. A promising XAI method, not studied in the context of digital histopathology is automated concept-based explanation (ACE). It automatically extracts visual concepts from image data. Our objective is to evaluate ACE’s technical validity following design science principals and to compare it to Guided Gradient-weighted Class Activation Mapping (Grad-CAM), a conventional pixel-wise explanation method. To that extent, we created and studied five convolutional neural networks (CNNs) in four different skin cancer settings. Our results demonstrate that ACE is a valid tool for gaining insights into the decision process of histopathological CNNs that can go beyond explanations from the control method. ACE validly visualized a class sampling ratio bias, measurement bias, sampling bias, and class-correlated bias. Furthermore, the complementary use with Guided Grad-CAM offers several benefits. Finally, we propose practical solutions for several technical challenges. In contradiction to results from the literature, we noticed lower intuitiveness in some dermatopathology scenarios as compared to concept-based explanations on real-world images.

InterpolatedXY: a two-step strategy to normalize DNA methylation microarray data avoiding sex bias.

MotivationData normalization is an essential step to reduce technical variation within and between arrays. Due to the different karyotypes and the effects of X chromosome inactivation, females and males exhibit distinct methylation patterns on sex chromosomes; thus, it poses a significant challenge to normalize sex chromosome data without introducing bias. Currently, existing methods do not provide unbiased solutions to normalize sex chromosome data, usually, they just process autosomal and sex chromosomes indiscriminately.ResultsHere, we demonstrate that ignoring this sex difference will lead to introducing artificial sex bias, especially for thousands of autosomal CpGs. We present a novel two-step strategy (interpolatedXY) to address this issue, which is applicable to all quantile-based normalization methods. By this new strategy, the autosomal CpGs are first normalized independently by conventional methods, such as funnorm or dasen; then the corrected methylation values of sex chromosome-linked CpGs are estimated as the weighted average of their nearest neighbors on autosomes. The proposed two-step strategy can also be applied to other non-quantile-based normalization methods, as well as other array-based data types. Moreover, we propose a useful concept: the sex explained fraction of variance, to quantitatively measure the normalization effect.Availability and implementationThe proposed methods are available by calling the function ‘adjustedDasen’ or ‘adjustedFunnorm’ in the latest wateRmelon package (https://github.com/schalkwyk/wateRmelon), with methods compatible with all the major workflows, including minfi.Supplementary information Supplementary data are available at Bioinformatics online.

Deep Learning Paradigm for Cardiovascular Disease/Stroke Risk Stratification in Parkinson's Disease Affected by COVID-19: A Narrative Review.

Background and Motivation: Parkinson’s disease (PD) is one of the most serious, non-curable, and expensive to treat. Recently, machine learning (ML) has shown to be able to predict cardiovascular/stroke risk in PD patients. The presence of COVID-19 causes the ML systems to become severely non-linear and poses challenges in cardiovascular/stroke risk stratification. Further, due to comorbidity, sample size constraints, and poor scientific and clinical validation techniques, there have been no well-explained ML paradigms. Deep neural networks are powerful learning machines that generalize non-linear conditions. This study presents a novel investigation of deep learning (DL) solutions for CVD/stroke risk prediction in PD patients affected by the COVID-19 framework. Method: The PRISMA search strategy was used for the selection of 292 studies closely associated with the effect of PD on CVD risk in the COVID-19 framework. We study the hypothesis that PD in the presence of COVID-19 can cause more harm to the heart and brain than in non-COVID-19 conditions. COVID-19 lung damage severity can be used as a covariate during DL training model designs. We, therefore, propose a DL model for the estimation of, (i) COVID-19 lesions in computed tomography (CT) scans and (ii) combining the covariates of PD, COVID-19 lesions, office and laboratory arterial atherosclerotic image-based biomarkers, and medicine usage for the PD patients for the design of DL point-based models for CVD/stroke risk stratification. Results: We validated the feasibility of CVD/stroke risk stratification in PD patients in the presence of a COVID-19 environment and this was also verified. DL architectures like long short-term memory (LSTM), and recurrent neural network (RNN) were studied for CVD/stroke risk stratification showing powerful designs. Lastly, we examined the artificial intelligence bias and provided recommendations for early detection of CVD/stroke in PD patients in the presence of COVID-19. Conclusion: The DL is a very powerful tool for predicting CVD/stroke risk in PD patients affected by COVID-19.

ADataViewer: exploring semantically harmonized Alzheimer’s disease cohort datasets

BackgroundCurrently, Alzheimer’s disease (AD) cohort datasets are difficult to find and lack across-cohort interoperability, and the actual content of publicly available datasets often only becomes clear to third-party researchers once data access has been granted. These aspects severely hinder the advancement of AD research through emerging data-driven approaches such as machine learning and artificial intelligence and bias current data-driven findings towards the few commonly used, well-explored AD cohorts. To achieve robust and generalizable results, validation across multiple datasets is crucial.MethodsWe accessed and systematically investigated the content of 20 major AD cohort datasets at the data level. Both, a medical professional and a data specialist, manually curated and semantically harmonized the acquired datasets. Finally, we developed a platform that displays vital information about the available datasets.ResultsHere, we present ADataViewer, an interactive platform that facilitates the exploration of 20 cohort datasets with respect to longitudinal follow-up, demographics, ethnoracial diversity, measured modalities, and statistical properties of individual variables. It allows researchers to quickly identify AD cohorts that meet user-specified requirements for discovery and validation studies regarding available variables, sample sizes, and longitudinal follow-up. Additionally, we publish the underlying variable mapping catalog that harmonizes 1196 unique variables across the 20 cohorts and paves the way for interoperable AD datasets.ConclusionsIn conclusion, ADataViewer facilitates fast, robust data-driven research by transparently displaying cohort dataset content and supporting researchers in selecting datasets that are suited for their envisioned study. The platform is available at https://adata.scai.fraunhofer.de/.

Calendar effects on surface air temperature and precipitation based on model-ensemble equilibrium and transient simulations from PMIP4 and PACMEDY

Abstract. Numerical modeling enables a comprehensive understanding not only of the Earth's system today, but also of the past. To date, a significant amount of time and effort has been devoted to paleoclimate modeling and analysis, which involves the latest and most advanced Paleoclimate Modelling Intercomparison Project phase 4 (PMIP4). The definition of seasonality, which is influenced by slow variations in the Earth's orbital parameters, plays a key role in determining the calculated seasonal cycle of the climate. In contrast to the classical calendar used today, where the lengths of the months and seasons are fixed, the angular calendar calculates the lengths of the months and seasons according to a fixed number of degrees along the Earth's orbit. When comparing simulation results for different time intervals, it is essential to account for the angular calendar to ensure that the data for comparison are from the same position along the Earth's orbit. Most models use the classical calendar, which can lead to strong distortions of the monthly and seasonal values, especially for the climate of the past. Here, by analyzing daily outputs from multiple PMIP4 model simulations, we examine calendar effects on surface air temperature and precipitation under mid-Holocene, Last Interglacial, and pre-industrial climate conditions. We came to the following conclusions. (a) The largest cooling bias occurs in boreal autumn when the classical calendar is applied for the mid-Holocene and Last Interglacial, due to the fact that the vernal equinox is fixed on 21 March. (b) The sign of the temperature anomalies between the Last Interglacial and pre-industrial in boreal autumn can be reversed after the switch from the classical to angular calendar, particularly over the Northern Hemisphere continents. (c) Precipitation over West Africa is overestimated in boreal summer and underestimated in boreal autumn when the classical seasonal cycle is applied. (d) Finally, month-length adjusted values for surface air temperature and precipitation are very similar to the day-length adjusted values, and therefore correcting the calendar based on the monthly model results can largely reduce the artificial bias. In addition, we examine the calendar effects in three transient simulations for 6–0 ka by AWI-ESM, MPI-ESM, and IPSL-CM. We find significant discrepancies between adjusted and unadjusted temperature values over continents for both hemispheres in boreal autumn, while for other seasons the deviations are relatively small. A drying bias can be found in the summer monsoon precipitation in Africa (in the classical calendar), whereby the magnitude of bias becomes smaller over time. Overall, our study underlines the importance of the application of calendar transformation in the analysis of climate simulations. Neglecting the calendar effects could lead to a profound artificial distortion of the calculated seasonal cycle of surface air temperature and precipitation.

Evaluation of the Homogenization Adjustments Applied to European Temperature Records in the Global Historical Climatology Network Dataset

The widely used Global Historical Climatology Network (GHCN) monthly temperature dataset is available in two formats—non-homogenized and homogenized. Since 2011, this homogenized dataset has been updated almost daily by applying the “Pairwise Homogenization Algorithm” (PHA) to the non-homogenized datasets. Previous studies found that the PHA can perform well at correcting synthetic time series when certain artificial biases are introduced. However, its performance with real world data has been less well studied. Therefore, the homogenized GHCN datasets (Version 3 and 4) were downloaded almost daily over a 10-year period (2011–2021) yielding 3689 different updates to the datasets. The different breakpoints identified were analyzed for a set of stations from 24 European countries for which station history metadata were available. A remarkable inconsistency in the identified breakpoints (and hence adjustments applied) was revealed. Of the adjustments applied for GHCN Version 4, 64% (61% for Version 3) were identified on less than 25% of runs, while only 16% of the adjustments (21% for Version 3) were identified consistently for more than 75% of the runs. The consistency of PHA adjustments improved when the breakpoints corresponded to documented station history metadata events. However, only 19% of the breakpoints (18% for Version 3) were associated with a documented event within 1 year, and 67% (69% for Version 3) were not associated with any documented event. Therefore, while the PHA remains a useful tool in the community’s homogenization toolbox, many of the PHA adjustments applied to the homogenized GHCN dataset may have been spurious. Using station metadata to assess the reliability of PHA adjustments might potentially help to identify some of these spurious adjustments.