Episode 114 of AI Ethics with Fexingo dives into a surprising new source of algorithmic bias: smart recycling bins. Lucas and Luna explore how a city's rollout of AI-powered recycling bins in 2024 led to unequal service — richer neighborhoods got better collection and data-driven rewards, while lower-income areas saw fewer bins and less accurate waste classification. They unpack the study from the University of Michigan that showed how the bins' computer vision models performed 30% worse on non-standard packaging common in ethnic grocery stores, and how this feedback loop reinforces environmental inequality. The hosts discuss the broader implications for smart city infrastructure, from traffic cameras to air quality sensors, and ask whether municipalities are auditing these systems for fairness before deployment. A sharp, specific look at how the Internet of Things can encode bias into public services.

#AIEthics #AlgorithmicBias #SmartCities #Recycling #EnvironmentalJustice #ComputerVision #IoT #BiasInTech #DataJustice #UrbanTech #WasteManagement #Inequality #Fairness #UniversityOfMichigan #FexingoBusiness #BusinessPodcast #Technology #AIAndSociety

Keep every episode free: buymeacoffee.com/fexingo

In this episode of AI Ethics with Fexingo, Lucas and Luna dive into the hidden biases embedded in predictive policing algorithms. They focus on the controversial case of the PredPol system, which was trained on decades of historical crime data from the Los Angeles Police Department. The hosts break down how such data reflects over-policing in minority neighborhoods, creating a feedback loop that reinforces racial disparities. Lucas explains the concept of 'biased ground truth' — when the data used to train AI itself contains systemic bias. Luna highlights a 2023 study showing that PredPol's predictions in Oakland were disproportionately concentrated in Black and Latino communities, even after controlling for actual crime rates. The conversation explores technical solutions like fairness constraints and community oversight, but questions whether algorithmic fixes alone can address the deeper structural issues. A must-listen for anyone interested in the real-world impact of AI on social justice.

#AIEthics #PredictivePolicing #PredPol #LosAngelesPoliceDepartment #RacialBias #AlgorithmicJustice #CrimeData #FairnessInAI #TechnicalDebt #FeedbackLoop #MachineLearning #Technology #BusinessPodcast #FexingoBusiness #DataBias #CommunityOversight #HumanInTheLoop #OaklandStudy

Keep every episode free: buymeacoffee.com/fexingo

In this episode of AI Ethics with Fexingo, Lucas and Luna explore how artificial intelligence is being used in hospital emergency departments to triage patients — and how those models can inherit racial and socioeconomic biases from historical data. They focus on a specific case: a 2025 study of an AI triage system deployed in three Chicago hospitals, which was found to systematically under-prioritize Black patients for critical care. Lucas walks through the data pipeline, from electronic health records to training labels, and explains why a model trained on 'time-to-treatment' encodes existing disparities. Luna questions whether regulators are equipped to audit these systems, and they discuss emerging solutions like 'fairness-aware' training objectives. The episode strikes a balance between technical depth and real-world stakes, giving listeners a concrete framework for understanding algorithmic bias in healthcare.

#AIEthics #HealthcareAI #MedicalBias #TriageAlgorithms #AlgorithmicFairness #HealthDisparities #RacialBias #ArtificialIntelligence #MachineLearning #EthicsInTech #PatientSafety #ChicagoHospitals #ClinicalDecisionSupport #ElectronicHealthRecords #HealthcareTechnology #FexingoBusiness #BusinessPodcast #Technology

Keep every episode free: buymeacoffee.com/fexingo

In this episode of AI Ethics with Fexingo, Lucas and Luna explore a surprising new source of bias in AI training data: voice chat from online video games. They discuss how millions of hours of toxic, gendered, and racially charged speech from gaming platforms are being scraped to train speech recognition and sentiment analysis models. The hosts examine a real 2025 study that found models trained on game voice data performed 22 percent worse on female and non-native English speakers. They also unpack the broader implications for AI safety when training data includes deliberate adversarial inputs. Lucas and Luna connect this to ongoing debates about content moderation, consent in data collection, and the challenge of filtering hostile speech without losing linguistic diversity. A must-listen for anyone working in AI ethics, game development, or speech technology.

#AIEthics #BiasInAI #VoiceChat #VideoGames #SpeechRecognition #ToxicLanguage #TrainingData #DataConsent #AISafety #GenderBias #RacialBias #GameDevelopment #NaturalLanguageProcessing #MachineLearning #Technology #FexingoBusiness #BusinessPodcast #EthicalAI

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna explore a surprising frontier of AI bias: satellite imagery. When computer vision models trained on overhead photos predict everything from property values to crop yields to crime hotspots, they absorb and amplify the human patterns embedded in those pixels — including redlining, segregation, and economic inequality. Lucas walks through a concrete case from a 2025 Stanford study that found AI models predicting neighborhood income from roof color and tree canopy, and how that creates feedback loops in insurance pricing and municipal services. Luna presses on whether satellite data is less biased than ground-level photos, and Lucas explains why 'geographic fairness' is becoming a new category in AI ethics auditing. The episode closes with a look at proposed regulations in the EU's AI Act that would classify satellite-based risk assessment as high-risk. Turns 31. Word count 1727.

#AIEthics #SatelliteBias #ComputerVision #GeographicFairness #RedliningAI #PropertyAssessment #InsuranceAI #StanfordStudy #EUAIAct #AlgorithmicBias #RemoteSensing #UrbanAI #FairnessInAI #Technology #FexingoBusiness #BusinessPodcast #ResponsibleAI #DataBias

Keep every episode free: buymeacoffee.com/fexingo

In this episode of AI Ethics with Fexingo, Lucas and Luna examine a specific case: an AI model trained on breast cancer clinical trial data that performed poorly on Black women because the original trials enrolled only nine percent Black participants. They discuss how regulatory guidelines from the FDA and NIH have tried to fix trial diversity since the 1990s, but real enrollment still lags. The hosts trace the problem from trial design to model deployment, asking whether fixing the data pipeline is enough when the underlying healthcare system has biases built in. They also touch on a recent preprint showing how an AI trained on 'diverse' synthetic data still failed for a subgroup it was designed to include. This episode is a focused look at one structural flaw in medical AI, not a broad survey.

#AIEthics #ClinicalTrials #MedicalAI #AlgorithmicBias #HealthcareDisparities #BreastCancer #RaceInMedicine #FDA #NIH #DataDiversity #SyntheticData #HealthEquity #Technology #FexingoBusiness #BusinessPodcast #ResponsibleAI #BiasedAlgorithms #LucasAndLuna

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna explore how YouTube's comment sections are training AI models to absorb and amplify toxic language, political polarization, and racial bias. They focus on a 2025 study from the University of Washington that analyzed 10 million YouTube comments and found that AI models trained on them picked up hate speech, conspiracy theories, and gendered insults at rates far higher than models trained on curated text datasets. The hosts discuss why YouTube comments are uniquely dangerous training data — unmoderated, algorithmically boosted, and filled with bot-driven content — and what it means for companies building chatbots and content moderation tools on top of Google's data. They also touch on Google's own attempts to filter biased comments, the limits of synthetic data as a fix, and why this problem is harder to solve than biased Wikipedia or Reddit data.

#YouTubeComments #AIBias #TrainingData #HateSpeech #Google #AIethics #MachineLearning #NaturalLanguageProcessing #ContentModeration #Polarization #ToxicLanguage #Bots #UniversityOfWashington #LanguageModels #SyntheticData #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Episode 107 of AI Ethics with Fexingo explores a surprising source of algorithmic bias: dating apps. Lucas and Luna unpack how recommendation algorithms on platforms like Tinder and Hinge can encode racial, socioeconomic, and appearance-based preferences, reinforcing stereotypes and affecting real-world match outcomes. They discuss a 2025 study from Stanford's Human-Centered AI Lab that found users' swiping patterns train models to favor certain demographics, creating feedback loops of exclusion. The hosts also examine the ethical implications of 'desirability scores' and whether regulation could address these hidden biases. This episode offers a concrete look at how AI shapes human connection—and what it means for fairness in digital spaces.

#AIEthics #DatingApps #AlgorithmicBias #MachineLearning #RecommendationSystems #StanfordHAI #DesirabilityScores #RacialBias #SocioeconomicBias #AppearanceBias #Tinder #Hinge #FeedbackLoops #Regulation #Technology #FexingoBusiness #BusinessPodcast #ResponsibleAI

Keep every episode free: buymeacoffee.com/fexingo

In this episode of AI Ethics with Fexingo, Lucas and Luna dive into a growing concern: biased AI analysis of police body camera footage. They examine a 2025 study from the University of Michigan that found speech-to-text models performed 20% worse on officers from underrepresented backgrounds, leading to inaccurate incident reports. The hosts discuss how this bias can compound in evidence analysis, affect public trust, and what steps researchers are taking to improve datasets. A concrete look at how technology meant to increase accountability can inadvertently embed new forms of bias.

#AIEthics #PoliceBodyCameras #SpeechRecognitionBias #AlgorithmicBias #CriminalJusticeTech #UniversityOfMichigan #FacialRecognition #NaturalLanguageProcessing #DataBias #Accountability #Technology #BiasInAI #LawEnforcementTech #VoiceRecognition #EvidenceAnalysis #FexingoBusiness #BusinessPodcast #ResponsibleAI

Keep every episode free: buymeacoffee.com/fexingo

Episode 105 of AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence. Hosts Lucas and Luna examine how AI diagnostic tools absorb racial and demographic bias from medical imaging datasets. They focus on a 2025 Nature Medicine study showing that chest X-ray models trained on predominantly white populations misdiagnose pneumonia at 40 percent higher rates for Black patients. The conversation covers the root cause — historical disparities in healthcare access reflected in training data — and explores solutions like federated learning and demographic rebalancing that researchers are testing at institutions like MIT and Stanford. Lucas and Luna also discuss why the FDA's current approval framework for AI medical devices doesn't require demographic performance audits, and whether the 2026 AI Bill of Rights could close that gap. If today's discussion gave you a new way to think about responsible tech, support the show at buy me a coffee dot com slash fexingo.

#AIEthics #MedicalAI #AlgorithmicBias #HealthTech #ArtificialIntelligence #RacialBiasInMedicine #DataJustice #FDA #TechRegulation #NatureMedicine #FexingoBusiness #BusinessPodcast #Technology #Podcast #MachineLearning #HealthEquity #DiagnosticAI #FederatedLearning

Keep every episode free: buymeacoffee.com/fexingo