English: AI technologies have the potential to benefit the Wikimedia Movement, but they come with risks. The Wikimedia Foundation has begun to build AI products around these technologies. The emerging domain of ethical AI proposes new approaches for addressing the discrimination, disruption, and damage that AI can cause. The established discipline of human centered design provides guidance on how to maintain a focus on human needs and wellbeing throughout product development.

This white paper is intended to help Wikimedia ensure ethical and human-centered outcomes in AI product development given our current and anticipated goals, needs, capacities, and workflows. The paper makes two contributions: 1) it motivates a set of risk scenarios intended to define the problem space and promote reflective decision-making, and 2) it presents a set of process proposals for improving AI product development. Taken together, these scenarios and proposals can help Wikimedia address anticipated challenges and identify emerging opportunities to leverage AI technologies to further our mission.

This document is oriented towards Wikimedia’s 2017 Strategic Direction. It complements the strategic priorities described in the white papers Knowledge gaps, Knowledge integrity, and Foundations by Wikimedia Research and Augmentation by Wikimedia Audiences. Background material and additional resources are available on meta.wikimedia.org.

See also: the project page on Meta and the citeable version of the white paper on Figshare