Making Better Decisions with Data
summary
Examples show where data fails: Netflix Prize: 10% offline improvement but Netflix declines to implement for cost reasons, recommendation system still relies on human labeling; Google Flu Trends misses spikes in 2011–2013 and overestimates later seasons; a Tennessee welfare drug test miscalculates prevalence by dividing 37 positives by 16,000 to claim 0.2% overall; a UCLA screen-time experiment uses nonrandom camps, small samples and reports pre/post error rates (~14 → 9.41 vs 9.81) with no significance testing.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 2:25Data alone doesn’t ensure decisions
- 6:59Big data myths debunked
- 18:35Flaws in data visualization
- 28:19Biases in data interpretation
- 35:24Mitigating biases
- 44:53Q&A
related talks
44mKeep it Clean: Why Bad Data Ruins Projects and How to Fix it
47mData Analytics: Realizing the Full Potential
49mKeep it Clean: Why Bad Data Ruins Projects and How to Fix it
20mDon't Look. See! Are We Blinded by Data (Visualization)?
42mAugmenting Human Decision Making with Data Science
51mFairytale vs. Narrative: Telling stories and making decisions with data
32mIntroducing Reality into Data Metrics
21mGarbage In, Garbage Out: Elevating AI with High-Quality Data