towards editorial transparency in computational journalism
Post on 15-Apr-2017
Embed Size (px)
Towards Editorial Transparency in Computational Journalism
Towards Editorial Transparency in Computational JournalismJennifer A StarkNick DiakopoulosThe University of Maryland, College of Journalism, Computational Journalism Lab
What do we mean by Transparency?the ways in which people both inside and external to journalism are given a chance to monitor, check, criticize and even intervene in the journalistic process. Deuze, M. 2005. What is journalism?: Professional identity and ideology of journalists reconsidered. Journalism. 6, 4 (2005), 442464
What do we mean by Transparency?Storytelling: Make the steps / data used to create your story visible to the audience.
Tool making:Sharing the code with thorough documentation.
Why Share Our Work?Benefits to yourself, fellow journalists, audience
AccountabilityDocument ProcessStimulate Alternative Stories / viewpointsDouble check data, code, analysis, and conclusions / interpretationFacilitate future work / future you / fellow journalists / fieldNovel work, or extensions to your original work.
Case Study 1: Storytelling (Uber)How?
Transparency promotes Accountability, Documentation, Further Storytelling Share raw collected data: GitHub, Google Drive (consider size)Open Source code sharing platform: GitHub, Jupyter
Transparency promotes Accountability, Documentation, Further Storytelling Share raw collected data: GitHub, Google Drive (consider size)Open Source code sharing platform: GitHub, Jupyter Project and Code Documentation: README.md
Transparency promotes Accountability, Documentation, Further Storytelling Share raw collected data: Google Drive (consider size)Open Source code sharing platform: GitHub, Jupyter Project and Code Documentation: README.mdAccountability: share data collection / processing / wrangling and analysisInterim processed data: .csv filesReplicability: programmatic steps where possible
How?Case Study 2: Tool Making(Twitter Bot)
Twitter Bot: Transparancy promotes accessibilityOpen Source code sharing platform: GitHub, JupyterProject and Code Documentation: README.mdLanguage / platform agnostic: configuration fileHow much to parameterize?Case-by-case uniqueness? Instructions within code and README documentation
Documentation!Takes longer than you think
Consider it an investment
Documentation within codeDocumentation in GitHub repository (README.md)Reciprocal links between news article and GitHub repositoryLinks to reference material (eg APIs, preceding work)
LicencesNobody should use your Code or Data if it is not licencedCode licences https://opensource.org/licenses
Data licences http://opendatacommons.org/about/
Multiple licences http://choosealicense.com/non-software/
Why Share Our Work?Evidence difficult to measure at this time IRL
Sunlight LabsPolicy makers (eg Transport, AARP)
Hobbyists / Individuals
Kate Rabinowitz Civic data scientisthttp://www.datalensdc.com/index2.html
About: DataLensDC has been featured in The Washingtonian, The Atlantic's CityLab,Washington City Paper, WJLA ABC 7 News, and more
Reinventing the wheel | Reuse codeStack overflow for sharing code / solutions? http://area51.stackexchange.com/proposals/103335/data-journalism/ Data or file repository?: https://quiltdata.com (or something similar?? I have not tried this tool)