Value Alignment of Social Media Ranking Algorithms

A Stanford-led research team has published a paper proposing a fundamental shift in how social media feeds are ranked. Instead of optimizing purely for engagement, their new algorithm is based on Schwartz's theory of Basic Human Values, a psychological framework of 10 universal values like benevolence, self-direction, and power. The system first models how these values are expressed in social media posts, then provides users with interactive controls to assign personal weights to each value. This creates a custom ranking formula that respects a user's articulated trade-offs between, for example, content promoting tradition versus content encouraging stimulation.

Through two controlled experiments with 141 and 250 participants, the researchers demonstrated that users could effectively use these controls to "architect" feeds reflecting their desired values. The resulting value-aligned feeds showed a substantial divergence from traditional engagement-driven feeds, which the paper argues are not value-neutral but instead prioritize individualistic values like achievement and hedonism by default. This work, accepted to the CHI 2026 conference, presents a concrete technical approach for platform designers to move beyond monolithic, engagement-optimized algorithms and toward user-steerable, value-aware systems.