We will discuss various types of bias and their impacts, how power dynamics are reflected in data and organizational practices, and practical methods to embed ethical practices into data projects.
Prior to this session, read and watch the following materials and write a short response to the reflection question 60 minutes total
This week, we’ll be diving into practical tools for an ethical approach to data projects, with a focus on documentation, which is a great tool that any individual data practitioner can use to help identify and address bias in any data project. Whether you’re a leader in charge of framing the whole project, or an entry-level data analyst working on a team, or anywhere in between, documentation is a powerful tool for you: you can read what exists (or ask if it exists at all 😅); author some new documentation or add to what exists; propose a framework for your organization to use for documenting all data-related projects; publish documentation alongside your project deliverables. These are all places where you can, looking through your data equity lens, identify gaps and opportunities to build better practices directly into your data work.
One strong example of a data documentation framework is *Datasheets for Datasets,* documentation to facilitate communication between dataset creators and consumers, by TImnet Gebru et al.
<aside> 💡 Key Point: You might have noticed that Datasheets for Datasets is designed for projects with data that either is currently or may in the future be fed into a machine learning or artificial intelligence algorithm. Even if not all of those 52 questions in the framework are relevant to your project, you can always extract a subset of those questions to use as a data documentation framework (hint hint, we’ll be talking about that more in our session!).
If this framework really doesn’t resonate for you, there are several similar frameworks for different levels of complexity that you could use or adapt: the Data Nutrition Project's Dataset Nutrition Label, the Open Data Institute’s Data Ethics Canvas, the We All Count Data Biography Template, and for machine learning developers, Model Cards for Model Reporting.
</aside>
How do you define ethics in your day-to-day data work? Humans have been thinking about ethics for thousands of years, and we can’t do justice to this vast topic in just three weeks! That’s why we are focused here on tools you can use and questions you can ask to get started. The Feminist Data Manifest-No prompts us to think about our responsibilities to one another as human beings, and how choosing how we do (and do not) use data can help us create a better future.
These next two videos, from the Canadian data equity consultancy We All Count, show us examples of how our decisions in data projects, such as how we compare outcomes or whose perspective we use when performing calculations, can dramatically change the outcomes of projects:
[ ] WATCH: Impact Reporting: How Averages Hide Equity Issues 4 min
[ ] WATCH: Not Your Average Average 4 min
This video excerpt provides an overview of documentation and some tools for equity in data projects:
[ ] WATCH: Documentation & Data Ethics – 3:00 to 9:20 TechChange 2021 Gender Data Impact + Innovation Series 6 min
[ ] WRITE: a post in our Slack channel