Ordering

Most Digital platforms organize information and direct attention. In both of these functions, the order in which the information ultimately gets presented to the user has an outsized impact on what the user sees, engages with, absorbs, and thus the harm (or benefit) that the content ultimately causes.

One subcategory of ordering approaches can broadly be described as "downranking": pushing content that appears to be harmful lower in orderings for search and discovery than it otherwise would appear. While downranking approaches can be a potent force for preventing harm, they suffer from the same weaknesses as other forms of Content Analysis: they require automated process to detect potential violative content, and suffer from the challenges of subjectivity, gamification, and false-positives that any type of content-based feature intervention will entail.

Other forms of  content-neutral ordering interventions are also promising, if less frequently deployed today. Some platforms that give users explicit control over the order in which they see content, while others randomize the ordering of content. Each of these kinds of interventions has its own strengths and weaknesses, but the commonality is that ordering is a domain fraught with tradeoffs of harms, and that the mechanism (or mechanisms) used for ordering should be considered a cornerstone of a platform's integrity design.