AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence

Episode 29 of AI Ethics with Fexingo explores the hidden dangers of AI fitness coaching apps. Lucas and Luna dive into a 2025 study from the University of Colorado Boulder that found popular AI-powered workout apps frequently recommend exercises with poor form or excessive intensity, leading to a 40% increase in user-reported injuries over a six-month period. The hosts examine the case of FitAI, a...

This episode explores how AI hiring tools can inherit and amplify gender bias from historical training data. Lucas and Luna dig into a 2024 study by the EU's Fundamental Rights Agency that found an AI recruiting system penalized resumes with female-coded language from women's colleges, reducing callback rates by 12 percent compared to male-coded equivalents. They discuss why gender-blind algorithm...

Lucas and Luna explore the growing use of AI in legal research and document review, focusing on a 2025 incident where an AI-powered legal assistant failed to cite a controlling Supreme Court precedent in a federal appeal. The episode examines how large language models hallucinate case law, the ethical obligations of attorneys who rely on them, and the push for 'certified AI' in litigation. With sp...

Episode 26 of AI Ethics with Fexingo dives into the growing use of AI systems that recommend medical treatments—like sepsis management or cancer therapy—without the rigorous clinical trials normally required for drugs or devices. Lucas and Luna examine the case of Epic Systems' Sepsis Model, which was deployed in hundreds of hospitals before a 2021 JAMA study showed it failed to detect sepsis in o...

In episode 25 of AI Ethics with Fexingo, Lucas and Luna explore the hidden world of AI content moderation on social media platforms. They zoom in on a 2025 study from the Algorithmic Justice League that found automated moderation tools disproportionately flag speech from minority communities—misinterpreting dialect, slang, and cultural context as hate speech or misinformation. Lucas breaks down th...

Episode 24 of AI Ethics with Fexingo investigates the growing use of AI-powered interview analysis tools like HireVue and Retorio. Lucas and Luna dive into a 2025 academic study revealing that many of these systems don't just evaluate what you say, but infer personality traits, emotional states, and even political leanings from your voice tone and facial expressions—without your knowledge or conse...

Episode 23 of AI Ethics with Fexingo explores the hidden world of applicant tracking systems (ATS) that automatically reject 75% of résumés before a human ever sees them. Lucas and Luna dive into how these AI filters, trained on historical hiring data from companies like Amazon and Unilever, can encode subtle biases against women, older workers, and non-traditional career paths. We break down the...

Most AI ethics debates assume your data has to leave your device. But a growing number of AI models now run entirely on your phone or laptop — no cloud, no server, no company seeing your selfies or search history. In this episode, Lucas and Luna look at Apple's on-device foundation model, Google's Gemini Nano, and how Qualcomm's latest mobile chips change the privacy equation. They ask whether 'ed...

Lucas and Luna explore the hidden crisis in AI-powered education: tutoring systems that claim to personalize learning but lack any pedagogical training. They examine the case of Khan Academy's Khanmigo, an AI tutor rolled out in 2023 that sometimes gives students confident but wrong answers. The conversation digs into why AI tutors fail at the basics of teaching—like knowing when a student is stuc...

This episode examines the rise of AI-generated educational content in art classrooms. Lucas and Luna dig into a case where a school district replaced most of its art curriculum with a generative AI system that creates lesson plans, project prompts, and even grading rubrics. The problem: the AI was trained on a dataset of art criticism and history that heavily favored Western, male, and canonical a...

In this episode, Lucas and Luna explore the growing use of AI in rental housing decisions — from tenant screening to eviction predictions. They focus on the case of SafeRent, a scoring algorithm that assigns rental risk scores to tenants, and how it has been challenged in court for alleged racial bias. The hosts discuss a 2023 lawsuit by the Massachusetts Attorney General against SafeRent, the com...

In this episode, Lucas and Luna explore the growing use of AI in child welfare assessments, focusing on the Allegheny Family Screening Tool (AFST) in Pennsylvania. They discuss how predictive models score families for risk of child maltreatment, the documented racial disparities in these systems, and the tension between efficiency and fairness. The hosts examine a 2023 study showing that AFST can...

This episode of AI Ethics with Fexingo looks at how AI tutoring platforms like Khan Academy's Khanmigo and others are being deployed in classrooms without the pedagogical training human teachers receive. We walk through a specific incident from spring 2026 where an AI tutor in a Los Angeles middle school gave a student incorrect and potentially misleading information about the French Revolution. W...

In episode 16 of AI Ethics with Fexingo, Lucas and Luna examine how artificial intelligence is being used to recommend sentencing lengths in criminal courts. They focus on a 2025 study from the University of Michigan analyzing over 200,000 sentencing recommendations from a tool used in four states. The hosts discuss how the model was trained on historical sentencing data that already reflected rac...

Episode 15 of AI Ethics with Fexingo explores the hidden world of automated grading in higher education. Lucas and Luna unpack a 2025 Stanford study that found AI grading systems penalize non-native English speakers by up to 15% in essay evaluations. They dive into the specific case of Arizona State University's 2024 pilot of an automated essay scoring tool, which led to a 23% higher fail rate amo...

Lucas and Luna investigate how artificial intelligence has quietly taken over consumer credit scoring in the United States, focusing on the case of a 2023 lawsuit against a major credit bureau that used a machine learning model to set credit limits. They walk through how the model penalized customers who used certain types of retail credit cards, why regulators are struggling to audit these system...

Episode 13 of AI Ethics with Fexingo digs into the growing use of AI in insurance claims processing. Lucas and Luna examine a specific case from May 2026: a major US insurer's automated system that denied flood damage claims after Hurricane Delta with an error rate of 34 percent. They discuss how algorithmic bias against older homeowners and non-English speakers persists even when regulators deman...

In this episode, Lucas and Luna examine how algorithms are increasingly used to set bail amounts and predict flight risk in the U.S. criminal justice system. They focus on the specific case of New Jersey, which reformed its bail system in 2017 to rely on a risk-assessment tool called the Public Safety Assessment (PSA). The hosts walk through how the PSA works, the data it uses, and the troubling f...

Episode 11 of AI Ethics with Fexingo examines the unregulated world of AI-powered mental health chatbots. Lucas and Luna dig into the case of a popular app that gave crisis users generic advice instead of emergency resources, and explore how the FDA's reluctance to classify these tools as medical devices leaves vulnerable users at risk. They discuss the difference between a therapeutic chatbot and...

Lucas and Luna examine how AI-driven worker management systems in the gig economy create unfair outcomes for drivers and delivery workers. They zero in on a 2025 study from the UC Berkeley Labor Center showing that a major ride-hailing platform's automated rating algorithm disproportionately deactivates drivers from low-income neighborhoods. Lucas explains how the AI evaluates 'performance' using...

Episode 9 of AI Ethics with Fexingo examines automated child welfare risk assessment algorithms. Lucas and Luna walk through the Allegheny Family Screening Tool (AFST), one of the earliest and most studied predictive models used by child protection agencies. They discuss how the tool scores calls to hotlines, the bias critiques that emerged, and what the data actually shows about false positives a...

Lucas and Luna explore the growing field of explainable AI, focusing on the tension between model accuracy and interpretability. They use the case of LIME and SHAP in loan approval algorithms, referencing a 2025 study from the Federal Reserve Bank of Philadelphia that found 23% of fintech lending decisions could not be explained by the models' stated factors. The hosts discuss why regulators are p...

Lucas and Luna explore the ethics of training AI on health records without explicit patient consent. They dive into the case of a major research hospital that gave a tech company access to millions of patient notes to train a diagnostic model, only to face a class-action lawsuit. The episode breaks down the trade-off between medical innovation and patient privacy, explaining concepts like de-ident...

Lucas and Luna examine the growing trend of AI transparency reports, which major tech companies like OpenAI, Google, and Meta publish to disclose how their models are built, tested, and monitored. But do these reports actually hold companies accountable? Lucas drills into the case of Anthropic's Model Card for Claude, showing that even the most detailed reports omit critical information about trai...

Episode 5 of AI Ethics with Fexingo tackles a next-generation bias problem: training data generated by AI itself. Lucas explains how synthetic data — used to save costs and protect privacy — can amplify existing prejudices when models learn from their own outputs. The conversation centers on one concrete case: a medical imaging model that inherited skin-tone bias from a synthetic dataset derived f...