Data Skeptic

💼 AI-powered job matching sounds great... but can you trust the recommendations?Roan Schellingerhout discusses explainable recommender systems for recruitment—and why "healthy friction" might actually help users make better decisions.Listen 🎧 open.spotify.com/episode/5aXhhs…

Václav Blahut from seznam.cz explains "inverse recommendation"—finding the right users for niche content instead of the usual approach.A clever repurposing of two-tower models that gives long-tail content a fighting chance.Dive in 🎧 open.spotify.com/episode/6zB80r…

@ek8terina (@MIT) talks strategic learning in rec sys 🎯 The paradox: algorithmic "protest movements" can actually HELP platforms by providing clearer signals We explore game theory, coordinated user behavior, and the platform vs. user arms race open.spotify.com/episode/358hGx…

@ervindervishaj open.spotify.com/episode/2vu304…

Can recommender systems be both powerful AND interpretable? 🔍 @ervindervishaj (@UniCopenhagen) shares research on disentanglement in RecSys Key finding: strong correlation between disentanglement & interpretability, but not always with performance

🎵 How can music recommendations be fairer? @Rebeccasalganik, @UofR, presents LARP, a framework tackling popularity and multi-interest bias in playlist continuation. Her Music Semantics dataset captures how ppl describe music—atmosphere, context, vibes. 🎯open.spotify.com/episode/0eIvXG…

@ek8terina dataskeptic.com/blog/episodes/…

🧠 What happens when users coordinate to game recommendation algorithms? @ek8terina reveals her findings: algorithmic "protest movements" can paradoxically benefit platforms by providing clearer preference signals.

Can ML be unfair without bad intent? Yes. 🤔 @dayvidliu (@Cornell) shows how PCA over-specializes on popular content, neglecting niche users. But his item-weighted PCA improves BOTH fairness AND performance! ⚖️ #AlgorithmicFairness #ResponsibleAI open.spotify.com/episode/6IsPN6…

What if we tracked eyes, not just clicks? 👁️ Santiago reveals how eye tracking uncovers what users actually see in recommendations. Introducing RecGaze—the 1st eye tracking dataset for rec systems! Changes everything about positional bias. 🎬 #RecSys open.spotify.com/episode/15ZZsL…

Study recommendation algorithms without direct data access! Our guests present a "recommender neutral user model" to deduce algorithmic impact when exposure data is missing. This breakthrough aids in understanding complex social media systems. #RecSys 🎯 tinyurl.com/mw53hu53

📊 @alberto_mancino of @sisinflab talks DataRec—a Python library solving dataset management headaches in recommender systems research. Automated downloads, checksum verification, standardized filtering. Game-changer for reproducibility! 🚀 #RecSys #MLOps open.spotify.com/episode/1r4nFw…

Can we build greener AI without losing performance? 🌱 @AntonioPuri00 of @SapienzaRoma talks EcoAware Graph Neural Networks—measuring and reducing the enviro impact of recommender systems. Learn to make your ML models more sustainable! 🌍 #GreenAI #RecSys open.spotify.com/episode/5ew0Nj…

How do fake profiles game recommendation algorithms? 🎯 @aditya_chichani from Walmart breaks down shilling attacks—from Spotify playlist manipulation to fake product reviews. Essential listening for anyone building rec systems! #MachineLearning #RecSys open.spotify.com/episode/5ecvTT…

🌍 Can AI make tourism more sustainable? @ashmi_banerjee from @TU_Muenchen talks using recommender systems to promote responsible travel. Learn how LLMs make synthetic tourism data & how algorithms can balance traveler satisfaction with enviro impact. 🎯 open.spotify.com/episode/0jcswV…

@Vida19231939 open.spotify.com/episode/0OIYxU…

🚀 The cold start problem is REAL! @Vida19231939 breaks down how hybrid recommender systems combine collaborative filtering + embeddings + bandit learning to make great recommendations for brand new users 🎧️ Listen👇 #MachineLearning #RecommenderSystems #DataScience #AI

🎙️ Check out Kyle's recent interview with @center4inquiry! He dives into AI, skepticism, and how technology intersects with critical thinking—covering everything from recommendation systems to the future of AI development. 🤖🔍

🏡 How do you find the next hot neighborhood? Graph neural networks may be the answer. @kunmukh from @Virginia_Tech talks about Z-REx, a GNN approach that recommends real estate regions and explains why—crucial for transparency in property search. 🎯 open.spotify.com/episode/6hjAdC…

🏛️ How do you build recommender systems for medieval manuscripts? @f_atzenhofer from @TUGraz1827 shares how they're using multi-modal AI to make Europe's largest charter collection discoverable for historians & researchers! 🎧Link in thread