Specific Error Research Articles

An automated deep learning pipeline for detecting user errors in spirometry test

Spirometer is used as a major diagnostic tool for obstructive airway diseases and a monitoring tool for therapy response and disease staging over time. It is a sophisticated medical device employed to quantify flow and volume of air exhaled by a subject during a specific testing period. The essential metrics obtained from the spirometry test, play a crucial role in enabling healthcare professionals to thoroughly evaluate the respiratory health and condition of the individual under examination. Several spirometer measurements including Forced Vital Capacity (FVC) and Forced Expiratory Volume (FEV) serve as guidelines for diagnosis and prognosis of Chronic Obstructive Pulmonary Diseases (COPD) and asthma. However, user errors caused by different reasons, including improper handling of the equipment and poor performance during the maneuvers of the expiratory airflow, end up in incorrect treatment directions. To ensure accurate results, spirometry tests traditionally require the presence of a skilled professional to identify and address these errors promptly. A novel machine learning approach is proposed in this paper to automatically identify four such user errors based on Volume-Time and Flow-Volume graphs. By detecting specific errors and providing immediate feedback to patients, reliability and accuracy of spirometry results will be improved and the need for trained professionals will be reduced. The implementation facilitates the widespread adoption of spirometry, particularly in low-resource telemedicine settings. This work implements a binary classification model distinguishing between normal and error test samples, achieving a prediction accuracy of 93%. Additionally, a 4-way classification model is presented for identifying individual error sub-types, demonstrating a prediction accuracy of 94%.

Under-Fitting and Over-Fitting: The Performance of Bayesian Model Selection and Fit Indices in SEM

We extended current knowledge by examining the performance of several Bayesian model fit and comparison indices through a simulation study using the confirmatory factor analysis. Our goal was to determine whether commonly implemented Bayesian indices can detect specification errors. Specifically, we wanted to uncover any differences in detecting under-fitting or over-fitting a model. We examined a conventional Bayesian fit index (the posterior predictive p-value), approximate Bayesian fit indices (Bayesian RMSEA, CFI, and TLI), and model comparison indices (BIC and DIC). We varied the type and severity of model mis-specification, sample size, and priors. We focused on the ability of these indices to detect model under- or over-fitting. We provide practical advice for applied researchers regarding how to assess and compare models using these common indices implemented in the Bayesian framework.

A comparative study of fluorescence and Raman spectroscopy for discrimination of virgin olive oil categories: Chemometric approaches and evaluation against other techniques

Spectroscopic techniques are presented as a reliable and economical alternative for the discrimination of olive oil quality categories. In this work, fluorescence and Raman spectroscopic techniques have been compared by evaluating different characteristics focused on sample preparation and analysis time, as well as classification percentages by chemometric models (PLS-DA) to classify olive oils into three categories: extra virgin (EVOO), virgin (VOO), and lampante (LOO). A statistical evaluation of the quality of the chemometric models was performed by comparing the parameters of sensitivity in terms of the true positive rate (%TPR), specificity (or true negative rate %TNR), and classification error in calibration, cross-validation and prediction. Fluorescence spectroscopy was the screening technique with the best overall accuracy in the ternary models for prediction (78.60%), with higher results than those obtained by Raman (66.03%), the latter being below the average error in the tasting panels. Regarding the classification obtained in binary models (EVOO/Non-EVOO and LOO/Non-LOO), fluorescence spectroscopy continued to present better results than Raman, with classifications of 92.5% and 88.5 % for both models, respectively (Raman obtained percentages of 70.7% and 66.6%). A comparison of the results obtained with confirmation techniques (headspace-gas chromatography-ion mobility spectrometry, HS-GC-IMS) and other spectroscopic techniques (NIR and MIR) studied in previous research was carried out, where it was observed that fluorescence spectroscopy obtained the closest classification percentage to the HS-GC-IMS technique (86.4%) in ternary models. On the other hand, the NIR technique obtained the highest classification percentage in binary models, with 92.6% for EVOO/Non-EVOO and 88.9% for LOO/Non-LOO compared to the value obtained using GC-IMS (95.4 and 86.4 %). Evaluating the rest of the parameters of interest for the olive industry, IR (both NIR and MIR) was the most interesting technique because it is a cheap, simple sensor with fast chemometric treatments and is environmentally-friendly. The results of this comparative study can be used to provide solutions and alternatives for applying new analytical strategies to support the work of tasting panels.

High-performance repetition cat code using fast noisy operations

Bosonic cat qubits stabilized by two-photon driven dissipation benefit from exponential suppression of bit-flip errors and an extensive set of gates preserving this protection. These properties make them promising building blocks of a hardware-efficient and fault-tolerant quantum processor. In this paper, we propose a performance optimization of the repetition cat code architecture using fast but noisy CNOT gates for stabilizer measurements. This optimization leads to high thresholds for the physical figure of merit, given as the ratio between intrinsic single-photon loss rate of the bosonic mode and the engineered two-photon loss rate, as well as an improved scaling below threshold of the required overhead, to reach an expected level of logical error rate. Relying on the specific error models for cat qubit operations, this optimization exploits fast parity measurements, using accelerated low-fidelity CNOT gates, combined with fast ancilla parity-check qubits. The significant enhancement in the performance is explained by: 1- the highly asymmetric error model of cat qubit CNOT gates with a major component on control (ancilla) qubits, and 2- the robustness of the repetition cat code error correction performance in presence of the leakage induced by fast operations. In order to demonstrate these performances, we develop a method to sample the repetition code under circuit-level noise that also takes into account cat qubit state leakage.

Cross-gender Analysis of Iraqi EFL Learners Recurrent Grammatical Errors in English

The study attempts to investigate recurrent grammatical errors that hinder Iraqi undergraduate EFL learners` mastery of English grammar. If male-female learners of English are not competent in the structural and semantic bases of tense, they can hardly produce acceptable sentences in the L2/FL. For this purpose, an elicitation test is used to sort out the typical errors of Iraqi male-female undergraduate students (1st, 3rd and 4th stage) of the Department of English Language and Literature in Mustansiriyah University / Iraq, with respect to specific recurrent tense errors reflecting the structure of their transitional competence or what S. Pit Corder (1976, p.15) labels idiosyncratic dialect. The justification behind this study is those adopted following Corder (1976, p1), who mentions that two justifications are essentially relevant for the study of learners' errors, one of which is the pedagogical justification, which implies that eliminating any error requires a good understanding of the nature of that error. The research follows a comprehensive analysis methodology. It combines quantitative and qualitative approaches to analyze samples from Iraqi male-female EFL learners. The study found that there are several cross-gender differences concerning the recurrent errors, which might be attributed to several factors, i.e. linguistic and sociolinguistic. The study takes these factors as the building blocks for suggesting pedagogical solutions. The study was originally an attempt to help EFL learners eliminate such recurrent L1\L2 conflicting errors.

Is illiquidity priced in an international factor pricing model? A dynamic panel data application with robust IV

AbstractIn the setting of a dynamic panel data framework, we investigate the international five‐factor Fama–French (2017) model augmented with traditional illiquidity factors (Amihud, Journal of Financial Markets, 2002, 5, 31–56; Amihud, Critical Finance Review, 2019, 8, 203–221; Pástor and Stambaugh, Journal of Political Economy, 2003, 111, 642–685; Pástor and Stambaugh, Critical Finance Review, 2019, 8, 277–299) to determine if any of these factors are priced. Since illiquidity measures are endogenous, we propose an algorithm that generates robust instruments which are combined with a GMM estimator to cope with both the endogeneity issues surrounding illiquidity and other eventual specification errors. In this dynamic framework, we generally find that the most significant factors correspond to market and size but illiquidity may matter depending on the level of the beta. We find that illiquidity has more impact on returns in expansion than in recession. However, the bid‐ask spread seems to behave differently from the other illiquidity measures.

Analyzing the Interactions between Environmental Parameters and Cardiovascular Diseases Using Random Forest and SHAP Algorithms.

Cardiovascular diseases (CVD) remain the predominant global cause of mortality, with both low and high temperatures increasing CVD-related mortalities. Climate change impacts human health directly through temperature fluctuations and indirectly via factors like disease vectors. Elevated and reduced temperatures have been linked to increases in CVD-related hospitalizations and mortality, with various studies worldwide confirming the significant health implications of temperature variations and air pollution on cardiovascular outcomes. A database of daily Emergency Room admissions at the Giovanni XIII Polyclinic in Bari (Southern Italy) was developed, spanning from 2013 to 2019, including weather and air quality data. A Random Forest (RF) supervised machine learning model was used to simulate the trend of hospital admissions for CVD. The Seasonal and Trend decomposition using Loess (STL) decomposition model separated the trend component, while cross-validation techniques were employed to prevent overfitting. Model performance was assessed using specific metrics and error analysis. Additionally, the SHapley Additive exPlanations (SHAP) method, a feature importance technique within the eXplainable Artificial Intelligence (XAI) framework, was used to identify the feature importance. An of 0.97 and a Mean Absolute Error of 0.36 admissions were achieved by the model. Atmospheric pressure, minimum temperature, and carbon monoxide were found to collectively contribute about 74% to the model's predictive power, with atmospheric pressure being the dominant factor at 37%. This research underscores the significant influence of weather-climate variables on cardiovascular diseases. The identified key climate factors provide a practical framework for policymakers and healthcare professionals to mitigate the adverse effects of climate change on CVD and devise preventive strategies.

Asymptotically robust permutation-based randomization confidence intervals for parametric OLS regression

Randomization inference provides exact finite sample tests of sharp null hypotheses which fully specify the distribution of outcomes under counterfactual realizations of treatment, but the sharp null is often considered restrictive as it rules out unspecified heterogeneity in treatment response. However, a growing literature shows that tests based upon permutations of regressors using pivotal statistics can remain asymptotically valid when the assumption regarding the permutation invariance of the data generating process used to motivate them is actually false. For experiments where potential outcomes involve the permutation of regressors, these results show that permutation-based randomization inference, while providing exact tests of sharp nulls, can also have the same asymptotic validity as conventional tests of average treatment effects with unspecified heterogeneity and other forms of specification error in treatment response. This paper extends this work to the consideration of interactions between treatment variables and covariates, a common feature of published regressions, as well as issues in the construction of confidence intervals and testing of subsets of treatment effects.

Intracranial Tumor Detection using Magnetic Resonance Imaging and Deep Learning

An intracranial tumor is a malignant Central Nervous System (CNS) cancer that significantly contributes to global mortality. Timely prediction of these brain tumors can improve the survival rates of a patient. Magnetic Resonance Imagining (MRI) and Computed Tomography (CT) have emerged as effective non-invasive ways to extract 2D or 3D images of human internal organs, eliminating any pain or surgical procedures. However, analyzing and distinguishing the normal and abnormal tissue is a challenging task. Due to this implying Datadriven approaches, could be a pragmatic way to efficiently classify and detect regions of malignancy of a tumor. The scope of this study is to efficiently predict Brain tumors and their location by employing sophisticated optimized Deep learning models like Convolutional neural Networks (CNN) and Long-Short Term Memory (LSTM) through limited MRI images dataset of human brain. The dataset acquired from kaggle is comprised of 253 MRI images covering different angles of the human brain. The images were of different sizes and shapes, to standardize them data preprocessing simulations were made including resizing, normalizing, etc. For better categorization of model, the images were also converted to binary format (0, 1) by the One Hot Encoding technique. The ratio of training and testing data was taken as 90:10. For CNN and LSTM model, suitable hyperparameters were selected through the trial-and-error technique to ensure the best optimization of the model on the implied training data. The loss and accuracy graph depicts optimized validation and accuracy losses indicating the model to call back early stopping to save computational cost. The predicted labels for both the models and actual label were compared by a confusion matrix which showed accuracy, specificity, recall, and misclassification error to be 91.51, 92.85, 90.12, and 8.49 on average for CNN,while 95.54, 92.86, 94.2, and 5.805 for LSTM model, respectively.

PENGARUH CORPORATE GOVERNANCE TERHADAP TRIPLE BOTTOM LINE REPORTING

The importance of the triple bottom line (TBL) concept has been debated among academicians and practitioners. The discussion about the extent of implementing these sustainable initiatives and what factor determines its reporting. Corporate governance is one of the influencing factors of TBL reporting. However, limited studies investigate corporate governance's role in TBL reporting, especially utilizing data from different continental European corporate governance systems like Indonesia. Therefore, this paper investigates the effect of the commissioner board size and gender on TBL reporting using the coal-mining industry. In addition, this study also collected the company's characteristic data and included it in the research model to maintain the model specification error. Twenty coal mining company has participated in this study. The analysis of multivariate regression is employed, and the finding indicates that there is no corporate governance role proxied by the size and gender of the commissioner's board. However, profitability is positively related to all dimensions of TBL. In addition, the TBL (social and environmental) is influenced positively by the company size. This study has practical and theoretical implications and is discussed in detail. 
 
 Keywords: Triple Bottom Line, Board Of Commissioners’ Size, Existence Of Women

Development of an augmented Damerau–Levenshtein method for correcting spelling errors in Kazakh texts

The presented paper is devoted to the development of a method for identifying and correcting spelling errors in Kazakh texts. In this paper, the study object is methods for more accurate correction of spelling errors in Kazakh texts. The aim of the study is to develop an augmented version of the Damerau-Levenshtein method for correcting spelling errors in Kazakh language texts. Automatic detection and correction of spelling errors have become a default feature in modern text editors for working with text data, in text messaging applications such as chatbots, messengers, etc. However, although this task is well solved in geographically widespread languages, it has not been fully solved in languages with a small audience, such as the Kazakh language. The methods developed so far cannot correct all spelling errors found in Kazakh texts. Therefore, the development of a method with specific algorithms for spelling error correction in Kazakh texts is considered. As a result of the research work, algorithms for correcting errors found in Kazakh language texts were developed, and the developed algorithms were included in the Damerau-Levenshtein method. The experimental testing results of the augmented Damerau- Levenshtein method showed 97.2 % accuracy in correcting specific errors found only in Kazakh words and 92.8 % accuracy in correcting common errors from letter symbols. The standard Damerau-Levenshtein method testing results showed 76.4 % accuracy in correcting specific errors found only in Kazakh words. The results of the tests in correcting common errors from letter symbols with the standard Damerau-Levenshtein were approximately the same with the augmented Damerau-Levenshtein method, the accuracy is 92.2 %. The extent and conditions of practical application of the results are implemented by including them in text editors, messengers, e-mails and similar applications that work with text data.

Personality traits and risk of eating disorders among Polish women: the moderating role of self-esteem.

Personality traits should be taken into account when diagnosing individuals with disordered eating behaviors in the hope of better understanding their etiology and symptom progression and when planning treatment. The objective of this study was to attempt to determine the moderating role of self-esteem in the relationships between personality traits included in the Big Five model among Polish women and estimated risk of eating disorders. The study was conducted among 556 Polish women from Zachodniopomorskie Voivodeship. The average age of the women under study was 34 years. A diagnostic survey was used as the research method, and the empirical data were collected using the following research tools: The NEO Five-Factor Inventory (NEO-FFI), Rosenberg Self-Esteem Scale (SES), ORTO - 15 Questionnaire, The Three-Factor Eating Questionnaire (TFEQ-13), and the authors' original questionnaire. A multivariate linear regression analysis was conducted to examine the influence of selected independent variables. The assumptions for the linear regression model were satisfied, as indicated by the Ramsey Regression Equation Specification Error Test, White's test, and the Jarque-Bera test. Only the personality trait of neuroticism exhibits a statistically significant effect on the "Cognitive Restraint of Eating," "Uncontrolled Eating," and "Emotional Eating" scores (p < 0.001). The moderation effect was demonstrated between self-esteem and the personality trait of conscientiousness on the "Cognitive Restraint of Eating" scale score. There is a moderation effect between self-esteem and the personality trait of extraversion on the "Uncontrolled Eating" subscale score. There is a moderation effect between self-esteem and the personality trait of conscientiousness on the "Uncontrolled Eating" scale score. Self-esteem was not a predictor of the occurrence of risk of eating disorders while playing a moderating role in the relationship between certain personality traits and estimated risk of eating disorders. A higher level of neuroticism was identified as an important predictor of higher results for orthorexia, Cognitive Restraint of Eating, Uncontrolled Eating, and Emotional Eating. It was also demonstrated that the orthorexia risk index decreased with increased extraversion and openness to experience. The results of this study suggest that eating behaviors and psychological factors should be included in psychological interventions in the treatment of eating disorders. The clinical goal can be considered to be an improvement in non-normative eating behaviors, such as a reduction in overeating episodes or eating less frequently in the absence of a hunger feeling. In order to assist these individuals in their attempts to achieve healthy behaviors, variables related to mental functioning can be then identified as important goals to support individuals in their efforts to change health behaviors by achieving better mental well-being.

Carbon-based light addressable potential sensor based on nitrogen-doped graphene quantum dots for detection of low-density lipoprotein

Cardiovascular diseases have become the number one killer of human health. Early diagnosis and treatment of cardiovascular diseases can effectively improve the treatment effect and reduce the disease death rate. As an important marker of cardiovascular disease, Low-Density Lipoprotein (LDL) can be used for the early detection of cardiovascular disease. In this study, a novel carbon-based Light Addressable Potentiometric Sensor (C-LAPS) for determination LDL has been constructed based on o-phenylenediamine modified nitrogen-doped graphene quantum dots/reduce graphene oxide (OPD@NGQDs/rGO) photosensitive layer. The OPD@NGQDs/rGO photosensitive layer has excellent photoelectric conversion efficiency, large specific surface area and good biocompatibility to enhance the response signal. For the determination LDL by the C-LAPS system, LDL binding to LDL aptamer causes a change in the surface potential of the photosensitive layer, producing a response signal that can be detected by the I-V curve. Under optimal conditions, the C-LAPS had a good linear relationship with LDL concentration in the range of 0.01 μg/mL- 0.15 μg/mL. The sensitivity was 1819 mV/µg·mL−1 and the limit of detection (LOD) was 6.43 ng/mL (S/N = 3). In addition, the C-LAPS showed good specificity, reproducibility, stability and satisfactory relative error for detecting LDL in human serum samples. Therefore, the C-LAPS has good performance and demonstrates potential clinical value.

A text mining approach to categorize patient safety event reports by medication error type

Patient safety reporting systems give healthcare provider staff the ability to report medication related safety events and errors; however, many of these reports go unanalyzed and safety hazards go undetected. The objective of this study is to examine whether natural language processing can be used to better categorize medication related patient safety event reports. 3,861 medication related patient safety event reports that were previously annotated using a consolidated medication error taxonomy were used to develop three models using the following algorithms: (1) logistic regression, (2) elastic net, and (3) XGBoost. After development, models were tested, and model performance was analyzed. We found the XGBoost model performed best across all medication error categories. ‘Wrong Drug’, ‘Wrong Dosage Form or Technique or Route’, and ‘Improper Dose/Dose Omission’ categories performed best across the three models. In addition, we identified five words most closely associated with each medication error category and which medication error categories were most likely to co-occur. Machine learning techniques offer a semi-automated method for identifying specific medication error types from the free text of patient safety event reports. These algorithms have the potential to improve the categorization of medication related patient safety event reports which may lead to better identification of important medication safety patterns and trends.

Investigating the Use of Online Bilingual Resources by Translation Students: The Vertical/Sequential Model Revisited

Four experiments were carried out in a university translation course after noticing the recurrence of specific errors identical in most target texts produced by students. The purpose was, to identify the errors and verify whether they could be attributed to using online bilingual resources before comprehension of the source text, and, second, to implement a teaching strategy for addressing this problem and later assess its effectiveness. Participants in the study were twenty-seven female students in two groups registered for the course in two semesters: a group in one semester and the other in the following term. They translated two experimental texts from English L2 into Arabic L1. Each text was translated twice: the first time by one group at the beginning of their course, where students were allowed to access all types of resources, and the second by the other group towards the end of their course, after implementation of the teaching strategy. Comparison between the target texts produced before the implementation of the strategy and those made following the strategy reveals that the latter category contained f fewer translation errors -regrouped in the study as inappropriate renderings, with almost no occurrence of the type termed excessive literality. The findings suggest that the use of bilingual resources by translation students before full-text comprehension is the main cause of most otherwise avoidable errors. Consequently, it should be banned at the first phase of the translation process. The teaching strategy, based on the vertical/sequential translating model, is recommended in translation teaching.

Fast contact detection for ellipsoids using optimization approaches

AbstractThe paper is concerned with fast and robust contact detection methods for arbitrary ellipsoids. An iterative procedure, namely the Levenberg‐Marquardt method, based on the common normal concept for parametric ellipsoids, is employed together with an implementation of the widely used GJK algorithm for comparison. The performance and accuracy of both are analysed and compared to each other on the basis of two test sets, each containing a total of 70 000 pairs of prolates or oblates. Emphasis is placed on the specific error measure relating the iterative solution to the exact one, which was chosen to be the maximum angle between the normal vector and the distance vector between two ellipsoids. The results indicate increased performance when using the Levenberg‐Marquardt method over the GJK algorithm with no loss of accuracy.

A Bayesian model for quantifying errors in citizen science data: application to rainfall observations from Nepal

Abstract. High-quality citizen science data can be instrumental in advancing science toward new discoveries and a deeper understanding of under-observed phenomena. However, the error structure of citizen scientist (CS) data must be well-defined. Within a citizen science program, the errors in submitted observations vary, and their occurrence may depend on CS-specific characteristics. This study develops a graphical Bayesian inference model of error types in CS data. The model assumes that (1) each CS observation is subject to a specific error type, each with its own bias and noise, and (2) an observation's error type depends on the static error community of the CS, which in turn relates to characteristics of the CS submitting the observation. Given a set of CS observations and corresponding ground-truth values, the model can be calibrated for a specific application, yielding (i) number of error types and error communities, (ii) bias and noise for each error type, (iii) error distribution of each error community, and (iv) the single error community to which each CS belongs. The model, applied to Nepal CS rainfall observations, identifies five error types and sorts CSs into four static, model-inferred communities. In the case study, 73 % of CSs submitted data with errors in fewer than 5 % of their observations. The remaining CSs submitted data with unit, meniscus, unknown, and outlier errors. A CS's assigned community, coupled with model-inferred error probabilities, can identify observations that require verification and provides an opportunity for targeted re-training of CSs based on mistake tendencies.

Doppler Echocardiographic Phenotypes in Suspected ‘Severe’ Aortic Stenosis: Matrix-Based Approach to Diagnosis and Management

Among patients with suspected severe aortic stenosis (AS), Doppler echocardiographic (DE) data are often discordant, and further analysis is required for accurate diagnosis and optimal management. In this study, an automated matrix-based approach was applied to an echocardiographic database of patients with AS that identified 5 discrete echocardiographic data patterns, 1 concordant and 4 discordant, each reflecting a particular pathophysiology/measurement error that guides further workup and management. A primary/discovery cohort of consecutive echocardiographic studies with at least 1 DE parameter of severe AS and analogous data from an independent secondary/validation cohort were retrospectively analyzed. Parameter thresholds for inclusion were aortic valve area (AVA) <1.0cm2, transaortic mean gradient (MG) ≥ 40mmHg, and/or transaortic peak velocity (PV) ≥ 4.0m/sec. Doppler velocity index (DVI) was also determined. Logic provided by an in-line SQL query embedded within the database was used to assign each patient to 1 of 5 discrete matrix patterns, each reflecting 1 or more specific pathophysiologies. Feasibility of automated pattern-driven triage of discordant cases was also evaluated. In both cohorts, data from each patient fitted only 1 data pattern. Of the 4,643 primary cohort patients, 39% had concordant parameters for severe AS and DVI <0.30 (pattern 1); 35% had AVA < 1.0cm2, MG<40mm Hg, PV < 4 m/sec, DVI < 0.30 (pattern 2); 9% had MG≥ 40mmHg and/or PV ≥ 4 m/sec, DVI > 0.30 (pattern 3); 10% had AVA < 1.0cm2, MG<40mmHg, PV < 4 m/sec, DVI >0.30 (pattern 4); and 7% had MG> 40mmHg and/or PV ≥ 4 m/sec, AVA > 1.0cm2, DVI < 0.30 (pattern 5). Findings were validated among the 387 secondary cohort patients in whom pattern distribution was remarkably similar. Matrix-based pattern recognition permits automated in-line identification of specific pathophysiology and/or measurement error among patients with suspected severe AS and discordant DE data.

Higher social tolerance is associated with more complex facial behavior in macaques

The social complexity hypothesis for communicative complexity posits that animal societies with more complex social systems require more complex communication systems. We tested the social complexity hypothesis on three macaque species that vary in their degree of social tolerance and complexity. We coded facial behavior in &gt;3000 social interactions across three social contexts (aggressive, submissive, affiliative) in 389 animals, using the Facial Action Coding System for macaques (MaqFACS). We quantified communicative complexity using three measures of uncertainty: entropy, specificity, and prediction error. We found that the relative entropy of facial behavior was higher for the more tolerant crested macaques as compared to the less tolerant Barbary and rhesus macaques across all social contexts, indicating that crested macaques more frequently use a higher diversity of facial behavior. The context specificity of facial behavior was higher in rhesus as compared to Barbary and crested macaques, demonstrating that Barbary and crested macaques used facial behavior more flexibly across different social contexts. Finally, a random forest classifier predicted social context from facial behavior with highest accuracy for rhesus and lowest for crested, indicating there is higher uncertainty and complexity in the facial behavior of crested macaques. Overall, our results support the social complexity hypothesis.

Higher social tolerance is associated with more complex facial behavior in macaques.

The social complexity hypothesis for communicative complexity posits that animal societies with more complex social systems require more complex communication systems. We tested the social complexity hypothesis on three macaque species that vary in their degree of social tolerance and complexity. We coded facial behavior in >3000 social interactions across three social contexts (aggressive, submissive, affiliative) in 389 animals, using the Facial Action Coding System for macaques (MaqFACS). We quantified communicative complexity using three measures of uncertainty: entropy, specificity, and prediction error. We found that the relative entropy of facial behavior was higher for the more tolerant crested macaques as compared to the less tolerant Barbary and rhesus macaques across all social contexts, indicating that crested macaques more frequently use a higher diversity of facial behavior. The context specificity of facial behavior was higher in rhesus as compared to Barbary and crested macaques, demonstrating that Barbary and crested macaques used facial behavior more flexibly across different social contexts. Finally, a random forest classifier predicted social context from facial behavior with highest accuracy for rhesus and lowest for crested, indicating there is higher uncertainty and complexity in the facial behavior of crested macaques. Overall, our results support the social complexity hypothesis.