How Not to Use Data Like a Racist: A Seven-Step Framework for Ethics and Equity in Data

The presentation was part of a Data on Purpose (DoP) conference presented by the Stanford Social Innovation Review (SSIR) in Feb 2021.

Heather Krause, founder of We All Count and Datassist, explains how we can unintentionally contribute to inequity in our projects. We assume that data can provide a “silver bullet” against bias, prejudice, human error and injustice. But we make choices that can impact how we interpret our results, and it depends on who’s perspective we are viewing the data.

Here are a few key takeaways:

  • All models have assumptions, so there is no perfect model, but we need to make sure our assumptions reflect those we are trying to serve.
  • It’s also not a binary decision, it’s a process to move towards more equity.
  • We should also move away from terms like statistically significance and instead use terms like uncertainty.
  • Use confidence intervals to reflect the sample size of different groups, instead of pooling many groups together

She discusses the seven frameworks of data equity:

  1. Funding
  2. Motivation
  3. Project Design
  4. Data Collection
  5. Analysis
  6. Interpretation
  7. Communication

There are also six principles for making robust and reliable data science choices in our analysis:

  1. Recognize that we are making subjective, human choices every day in our data work
  2. Identify as many choice points in the data process as possible
  3. Try to make choices around data that reflect the equity that you want to see
  4. Expand the group of people who get to make these meaningful choices and your data projects
  5. Talk about your data choices (be transparent) and stand by them.
  6. Be ready to try and make even better choices next time

You can watch the full webinar presentation by registering here