The general #incarceration-trends D4D project is concerned with overall incarceration patterns (currently US-focused). Two active projects have been under development
Working with the Colorado ACLU we have been asked to analyze and visualize data on bail in Colorado to inform proposed legislation in the upcoming state house legislative session.
We are also exploring the recently released (Incarceration Trends dataset)[https://github.com/vera-institute/incarceration_trends] for an overall view of incarceration patterns in the US.
Fairness - We acknowledge the perspectives and biases we bring to the analysis and understand that no data analysis exists in a vacuum independent from the social and political context in which it was created and shared.
Openness - We will make every effort to make all data open and all analysis as transparent and clear as possible. All project work will be publicly available on GitHub with a clear and descriptive README.
Reliability - We will clearly communicate the origins of our data and document each step of our work in order to make our work completely reproducible by a third party, and therefore more reliable.
Trust - We will use this project in part to demonstrate the role of data practitioners in informing policy change and will create open, transparent, and accessible work in order to build public trust in data and algorithms.
Social Benefit - We hope this project will inform criminal justice reform policy in the state of Colorado and beyond and will shed light on criminal justice practices which harm marginalized members of our communities.
- Consider (if not collect) informed and purposeful consent of data subjects for all projects, and discard resulting data when that consent expires.
N/A, aggregate data.
- Make best effort to guarantee the security of data, subjects, and algorithms to prevent unauthorized access, policy violations, tampering, or other harm or actions outside the data subjects’ consent.
N/A, aggregate data.
- Make best effort to protect anonymous data subjects, and any associated data, against any attempts to reverse-engineer, de-anonymize, or otherwise expose confidential information.
N/A, aggregate data.
- Practice responsible transparency as the default where possible, throughout the entire data lifecycle.
All data and analysis code will be publicly available on the project GitHub repository. Where technologies/frameworks are obscure or uncommon, efforts will be made to provide the tools for analysis in a dockerfile.
- Foster diversity by making efforts to ensure inclusion of participants, representation of viewpoints and communities, and openness. The data community should be open to, welcoming of, and inclusive of people from diverse backgrounds.
All interested parties will be welcome to join the project. We also hope to find ways to reach out to prisoner’s rights/grassroots criminal justice reform organizations in order to include perspectives of affected individuals into our work.
- Acknowledge and mitigate unfair bias throughout all aspects of data work.
In our initial discussion, many project members shared their initial thoughts, perspectives and biases in an effort to openly acknowledge expectations before starting the analysis. We hope that by publicly acknowledging our initial thoughts and perspectives, we can hold each other accountable and challenging unspoken assumptions in the analysis.
- Hold up datasets with clearly established provenance as the expected norm, rather than the exception.
All datasets used in the analysis will include a clear data provenance description in the repo readme with links to download the data wherever possible. The data received from the Colorado ACLU will be described in detail, including collection mechanism.
- Respect relevant tensions of all stakeholders as it relates to privacy and data ownership.
All of the data we plan to use in our analysis is open source and is publicly available. The data from the ACLU was retrieved through open records requests and we have confirmed they are happy to make it publicly available.
- Take great care to communicate responsibly and accessibly.
In communicating our results, we will carefully consider how the analysis could be misconstrued or misused to harm vulnerable communities. All complex statistical principles will be explained in clear and accessible language to ensure that the audience for the results is as inclusive as possible. Care will be taken to confirm the accessibility of what is produced (e.g., use of color-blind-friendly color palettes for visualization).
- Ensure that all data practitioners take responsibility for exercising ethical imagination in their work, including considering the implication of what came before and what may come after, and actively working to increase benefit and prevent harm to others.
This discussion will be sent in the project channel and all project members will be asked to read and discuss our plans for ethical work.