DataFramed #57 The Credibility Crisis in Data Science
Mar 18, 2019
Skipper Seabold, Director of Data Science at Civis Analytics and statsmodels creator, and Emily Robinson, an expert in A/B testing, delve into the credibility crisis plaguing data science. They discuss mismatched expectations across industries and the risks this poses to the labor market. The conversation highlights the importance of robust methodologies, effective communication with stakeholders, and the vital role of randomized control trials. Listeners gain insights into improving the integrity of data science and setting realistic goals in experimentation.
AI Snips
Chapters
Books
Transcript
Episode notes
Credibility Crisis in Economics
- Economists struggled to impact policy because their arguments focused on minutiae and lacked data.
- Studies often used limited datasets, leading to debates on estimation techniques rather than substantial findings.
Data Science Credibility Crisis
- Data science, like economics, faces a credibility crisis due to a focus on methods over impact.
- Decision-makers often question the value of data scientists' work, highlighting a communication gap.
Apply the Scientific Method
- Data scientists should follow the scientific method: asking relevant questions, forming falsifiable hypotheses, and designing experiments.
- Focus on providing value and translating findings into product improvements or actionable decisions.
