January 1, 2019
Subject tag: Algorithmic systems | Data Access | Privacy and data protection
In this paper, I discuss the opportunities and challenges of integrating algorithm auditing into a data sharing organization. I begin by briefly presenting background information on algorithm auditing as a field, examples of successful audits, and methods that are commonly leveraged to conduct audits. Next, I discuss how algorithm audits could be used to enrich donated datasets by (1) analyzing the algorithmic contexts that may have influenced the collection and production of the data, and (2) enabling independent validation that the data is complete and representative. Finally, I discuss several key challenges, including: tradeoffs between cooperating with and adversarially auditing data partners; maintaining the privacy of data collected during audits; and legal risks associated with auditing.
[This entry was sourced with minor edits from the Carnegie Endowment’s Partnership for Countering Influence Operations and its baseline datasets initiative. You can find more information here: https://ceip.knack.com/pcio-baseline-datasets]