Social Media Recommendation Algorithms Need Values
Biden recently called out social media platforms for “killing people” by spreading disinformation. Yet, this is just an instance of a larger problem of the attention economy, where platforms have been optimized for engagement or to maximize ad revenue. While minor tweaks have been done for “harmful content”, we need more systematic rethinking of how social media algorithms work. We need algorithms that think about values.
When we think of values, we can loo k to the more universal values of our culture and society. These can be things like “life, liberty, and the pursuit of happiness” from the declaration of independence, or the UN’s Sustainable Development Goals that includes things like eliminating poverty, eliminating poverty, addressing climate change, and improving health. While platforms could choose a particular set of values, it could also be possible to allow users choose which sets of values they want their feed to use. Some might want to emphasize or de-emphasize certain things depending upon their ideology or preferences.
Social media sites should use data about these values to train their platforms to meet these values. If we went with the UN SDGs, we would optimize for what social media feeds lead to people improving their economic well-being, becoming healthier, reducing gender inequality, and building more ecologically sustainable societies. While there may be some challenges in measuring these things, it doesn’t seem beyond the capacity of social media companies. For example, I’m sure Facebook already has data about the physical health of their users, which could be turned into an indicator.
