Abstract:Background: Social media sites are becoming an increasingly important source of information about mental health disorders. Among them, eating disorders are complex psychological problems that involve unhealthy eating habits. In particular, there is evidence showing that signs and symptoms of anorexia nervosa can be traced in social media platforms. Knowing that input data biases tend to be amplified by artificial intelligence algorithms and, in particular, machine learning, these methods should be revised to mitigate biased discrimination in such important domains. Objective: The main goal of this study was to detect and analyze the performance disparities across genders in algorithms trained for the detection of anorexia nervosa on social media posts. We used a collection of automated predictors trained on a data set in Spanish containing cases of 177 users that showed signs of anorexia (471,262 tweets) and 326 control cases (910,967 tweets). Methods: We first inspected the predictive performance differences between the algorithms for male and female users. Once biases were detected, we applied a feature-level bias characterization to evaluate the source of such biases and performed a comparative analysis of such features and those that are relevant for clinicians. Finally, we showcased different bias mitigation strategies to develop fairer automated classifiers, particularly for risk assessment in sensitive domains. Results: Our results revealed concerning predictive performance differences, with substantially higher false negative rates (FNRs) for female samples (FNR=0.082) compared with male samples (FNR=0.005). The findings show that biological processes and suicide risk factors were relevant for classifying positive male cases, whereas age, emotions, and personal concerns were more relevant for female cases. We also proposed techniques for bias mitigation, and we could see that, even though disparities can be mitigated, they cannot be eliminated. Conclusions: We concluded that more attention should be paid to the assessment of biases in automated methods dedicated to the detection of mental health issues. This is particularly relevant before the deployment of systems that are thought to assist clinicians, especially considering that the outputs of such systems can have an impact on the diagnosis of people at risk.

Application of a web-based tool for quantitative bias analysis: the example of misclassification due to self-reported body mass index

Stratified Probabilistic Bias Analysis for Body Mass Index-related Exposure Misclassification in Postmenopausal Women

A Quantitative Bias Analysis Approach to Informative Presence Bias in Electronic Health Records

An Examination of Bias of Facial Analysis based BMI Prediction Models

Conceptualizing bias in EHR data: A case study in performance disparities by demographic subgroups for a pediatric obesity incidence classifier

Validation of self-reported height and weight in a large, nationwide cohort of U.S. adults

A framework for understanding selection bias in real-world healthcare data

Addressing Information Biases Within Electronic Health Record Data to Improve the Examination of Epidemiologic Associations With Diabetes Prevalence Among Young Adults: Cross-Sectional Study

Quantitative bias analysis methods for summary level epidemiologic data in the peer-reviewed literature: a systematic review

Core Concepts in Pharmacoepidemiology: Quantitative Bias Analysis

Adjustment for Biased Sampling Using NHANES Derived Propensity Weights

Gender Bias When Using Artificial Intelligence to Assess Anorexia Nervosa on Social Media: Data-Driven Study

People with a body mass index ⩾30 under-report their dietary intake: A systematic review

Applying machine learning approaches for predicting obesity risk using US health administrative claims database

Self-estimated BMI, but not self-perceived body size, accurately identifies unhealthy weight in US adults

Measuring algorithmic bias to analyze the reliability of AI tools that predict depression risk using smartphone sensed-behavioral data

Human biases in body measurement estimation

De-Biasing the Bias: Methods for Improving Disparity Assessments with Noisy Group Measurements

External Validation of an Electronic Phenotyping Algorithm Detecting Attention to High Body Mass Index in Pediatric Primary Care

Uncovering Bias in Personal Informatics

Impact of body mass index and examination type on utilization of screening programs: A big data study